Activation asks you to stay on the page for 40 seconds because that visit is the signal that connects your website to your SEATEXT account. The script has to load, recognize a genuine page view, and send that event back to the server. If you leave before that happens, the AI stays inactive and your site name won't appear in the dashboard.

The 40-second wait also works as a lightweight human check. Installing a script is easy to automate, but a real person is more likely to stay, refresh, and watch for the connection. That makes the activation step harder to fake.

What the 40-second wait is really checking

During the wait, the system can confirm four things:

• The JavaScript code is actually present on your page.
• The page is running on a domain that SEATEXT can identify.
• The script can reach the SEATEXT service from your site.
• The visit belongs to the account that owns that domain.

You don't see those checks. You just see a normal page and a timer in the instructions. The 40 seconds gives each check time to finish.

An expert's view: why 40 seconds?

From a developer's perspective, 40 seconds is a buffer, not a magic number. JavaScript often waits for the page to finish loading, and browsers may delay third-party scripts. A 40-second window gives the script time to execute, send its signal, and retry if the first attempt fails.

The number also matters for security. Browsers already use the idea of “user activation”: certain APIs only work after a real click or tap from a human. The 40-second stay is similar. A human can wait 40 seconds without trying to automate anything; a simple bot often cannot or will not.

There is no public formula that says why 40 is exactly the right number. In practice, it's long enough to separate a real setup from an accidental visit, and short enough not to feel like a time-out.

Account and domain linking

The 40-second visit does more than start the AI. It links that specific URL to your SEATEXT account. According to the official setup instructions, each SEATEXT AI account is linked to a single primary URL.

That means the page you stay on matters. If you stay on a staging domain that happens to be on the same server, the account may not link correctly. Development URLs such as localhost are restricted for security reasons. You need a valid, real domain.

If you run several websites, create one account for each website. You can't use the same account to activate multiple distinct domains.

What changes if you skip or interrupt the wait

Leave too early and the signal may never complete. The site name won't appear next to the SEATEXT logo, and the system won't know the page is yours. You'll also be stuck before the next step, which is activating AI in the Main AI Hub.

There's a safety benefit here. The AI remains inert until activation, so your content isn't changed while the script is still unlinked. Skipping the wait doesn't cause damage; it just keeps activation from happening.

Limitations: when the 40-second rule does not apply

The instruction assumes a normal production website. These common setups need different handling:

• localhost: restricted for security reasons. Use a real domain.
• Dynamic development domains: may not work reliably because SEATEXT AI might not be able to associate traffic with your account.
• WP Engine: install the WP Engine plugin that lets you add custom JavaScript, then apply it across all pages.
• Multiple domains: create a separate account for each primary URL.

The 40-second wait is also a setup step, not a rule for every visitor. Once the site is linked and activated, the AI runs on its own.

How to tell activation worked

After your 40-second visit, wait at least five minutes. The main signal is your website name appearing next to the SEATEXT logo at the top of the setup page. That tells you the URL is connected and you can move on to the Main AI Hub.

If you see your site name, activation succeeded. From there, go to the Main AI Hub, choose the AI you need, and click Configuration to adjust settings. You can also edit the first round of translations and variants under Variants Edit.

If the website name still doesn't appear

Try these checks in order:

1. Confirm you created the SEATEXT AI account before installing the script.
2. Confirm the script was added to every page, not just the home page.
3. Use a valid, real domain; localhost and dynamic development domains are restricted.
4. Stay on the page for the full 40 seconds after refreshing, several times.
5. Wait the full five minutes before checking.
6. If 10 minutes pass with no website name, contact SEATEXT support. That usually points to an installation problem on your platform.

What activation unlocks

Activation is the gate before the AI agents can run. Once the site is linked, you can turn on agents for conversion-rate optimization, Google Ads landing page rewrites, translation into 125 languages, bot-click refund reports, and more. The setup page says to proceed to the Main AI Hub to activate the AI you need.

Before activation, those agents have nothing to work on. The script is inert. That's why the 40-second wait is worth getting right.

Key facts

Terminology you might see in the dashboard

Activation: the step that turns an installed script into an active AI connection.

Link to your account: binding the primary URL to your SEATEXT account.

Inert: the script is installed but is not doing anything yet.

Primary URL: the main domain associated with one account.

Main AI Hub: where you activate the AI agents and adjust configuration after the site is connected.

Frequently asked questions

What if I leave after 20 seconds instead of 40?

Activation may not finish, and your website name won't show up next to the SEATEXT logo. Go back, refresh the page, and stay for the full 40 seconds.

Why do I need to wait five minutes after the 40 seconds?

The 40 seconds lets the script send its signal. The next window gives the SEATEXT service time to update the account page with your website name. If you still don't see it after 10 minutes, contact support.

Can I activate SEATEXT on localhost?

No. Development URLs such as localhost are restricted for security reasons. Use a valid, real domain.

Can I use one SEATEXT account for two websites?

No. Each SEATEXT AI account is linked to a single primary URL. Create a separate account for each website.

What does “AI remains inert until activated” mean?

The script is safely installed on your page but does not change content, write variants, or run agents until the activation step links it to your account.

Does every visitor need to stay for 40 seconds?

No. The 40-second instruction is part of the setup and account-linking process, not a rule for visitors after the site is activated.

Need the actual code and the current setup steps? Go to the General Integration page.

AI translation outperforms manual translation for scaling multilingual SEO because it maintains the technical structure search engines rely on, updates automatically when your source content changes, and expands to dozens of languages without the cost and delay of hiring a separate translation team for each market. A tool like Seatext, for example, translates your site into 125 languages, preserves brand context, and optimizes localized pages for conversion. Manual translation simply cannot match that speed or consistency when you're targeting more than a few languages.

When you're scaling, the bottleneck isn't translation quality alone—it's the workflow around it. Manual translation forces you to manage per-language workflows, keep hreflang tags in sync, and redo work every time you update a page. AI translation automates those steps, so scaling from two languages to twenty doesn't multiply your workload.

Choose AI translation if you need to reach 5+ languages quickly, keep updates in sync, or manage a large content library. Choose manual translation if you need deep cultural nuance for a single high-value market, or if your content is legally or medically sensitive and requires human certification.

Why scaling manually breaks down

Manual translation works fine for one or two languages, but the moment you add a third, the operational costs explode. You need coordinators, glossary files, style guides, and separate publishing pipelines for each locale. Every homepage update becomes a multi-week project involving multiple vendors, and by the time the translation lands, your source content has changed again.

That lag creates broken experiences: visitors in Spanish see old offers, French pages have mismatched CTAs, and German URLs point to outdated product specs. Search engines notice the inconsistency and drop your rankings for those locales.

What AI translation changes under the hood

AI translation flips the workflow. Instead of exporting content, sending it to a translator, and re-importing the result, the AI agent watches your live website and translates on demand. Seatext's translation agent, for instance, translates your pages into 125 languages while preserving brand context, then optimizes the localized copy so visitors understand the product and convert without waiting on a manual project.

That means every new product page, blog post, or pricing change is immediately available in all target languages. No queued tasks, no email threads, no version mismatches.

The SEO mechanics: hreflang, content, and consistency

Multilingual SEO fails when the technical signals are wrong. Search engines rely on hreflang attributes to know which language version to show to a user in a specific region. If those tags are missing or contradictory, your pages compete against each other or get ignored altogether.

AI translation keeps those tags synchronized because it works directly on the page structure. The tool doesn't copy and paste text into a separate file; it creates the localized version within the same URL hierarchy, so the hreflang pairings stay intact. It also ensures that content changes—like a new product name or a promotional headline—appear in every language at the same time, which is what Google expects from a healthy multilingual site.

The cost and speed trade-off

Manual translation costs per word, per document, and per round of revision. AI translation runs on a subscription that covers all languages. Seatext's premium plan, for example, gives you every AI agent—including the translation agent—for $59 per month, with no separate charge for additional languages. The free starter plan even lets you try eight agents at no cost.

Speed is an even bigger differentiator. A manual project that takes six weeks can be done by AI in minutes. That speed doesn't just save time; it changes your strategy. You can test a new market with a full translated site in a week instead of a quarter, then double down on the languages that actually pull traffic.

When manual translation still makes sense

AI translation isn't always the right answer. For high-stakes content—medical instructions, legal contracts, or compliance documents—a human translator with domain expertise is still essential. Similarly, if you're producing a highly creative brand campaign that depends on cultural wordplay, a human copywriter in the target market will deliver more nuance than any machine.

The realistic approach is hybrid: use AI to cover the 95% of your site that is informational and transactional, then bring in human review for the pages that carry the most risk or brand equity.

Key facts about AI translation at scale

Limitations and common pitfalls

AI translation can't handle nuanced legal or medical copy without human oversight. It also can't adapt to regional dialects automatically—Spanish for Spain differs from Spanish for Mexico. You may need to manually set locale-specific preferences or run a second pass with a local reviewer.

A common mistake is treating AI translation as a one-time task. The real value comes from continuous synchronization. If you only translate once and never update, you'll get the same stale-content problem as manual translation. Choose a tool that re-translates whenever the source changes.

An expert's perspective

From a technical SEO standpoint, the biggest advantage of AI translation is that it keeps the localization layer attached to the original content. When a source page changes, the translated version changes with it. That prevents split signals, duplicate content penalties, and hreflang conflicts. Manual translation, by contrast, breaks that tie because the translated version is a separate asset that only changes when someone remembers to update it.

Expert teams also use AI translation to expand beyond just page content. The same technology can generate localized FAQ pages, blog posts, and product descriptions that target long-tail keywords in each language. Seatext's AI SEO agent, for example, builds FAQ pages that answer buyer questions in the local language, which helps you appear in both traditional search results and AI-generated answers.

FAQ

How many languages can I realistically scale to with AI translation?

Tools like Seatext support up to 125 languages without additional setup effort. The practical limit depends on your ability to maintain content quality across all locales, not on the tool.

Does AI translation rank as well as manually translated content?

Often yes, because AI tools can optimize for search intent and maintain technical SEO structure. But ranking also depends on your domain authority and link profile. The translation needs to be natural and accurate, which modern AI models deliver for most business content.

What does AI translation cost compared to manual translation?

AI translation typically runs on a flat subscription. Seatext's premium plan costs $59 per month for all agents, including translation. Manual translation usually costs ranges from $0.08 to $0.25 per word, plus project management overhead.

How do I handle regional dialects with AI translation?

You can either manually set locale-specific variants (e.g., Spanish-Mexico vs. Spanish-Spain) or run a quick human review for key pages. Most AI tools let you edit the translated copy directly if you need to adjust phrasing.

Is AI translation safe for my brand voice?

Yes, if you configure the glossary and tone in the tool. Seatext preserves brand context, which means it keeps your preferred terminology and product names consistent across languages.

What happens when I update my source content?

With AI translation, the tool automatically re-translates the updated sections. Manual translation requires you to resubmit the content and wait for a new round of revisions.

The root cause: Google can't tell intent from code

Google's crawler doesn't read your content the way a human does. It reads HTML structure, image file names, alt text, internal link patterns, and the actual text strings. When you publish a French version and an English version of the same page, the underlying code is often 90% identical. The images are the same. The layout is the same. The headings are the same — just in a different language.

To Google, that looks like a copy-paste job with a few words swapped. Unless you explicitly tell it otherwise, it will treat the pages as duplicates and pick one as the canonical version. The other pages get flagged as duplicate content and may be removed from the index entirely.

The diagnostic sequence: check these five signals in order

If you're seeing duplicate content flags on translated pages, work through this checklist in order. Each step rules out one possible cause.

1. Check for hreflang tags

Hreflang is the primary signal that tells Google "these pages are translations of each other, not copies." Without it, Google has no way to know that /fr/ and /en/ are intentional language variants.

Look at the HTML head of each translated page. You should see something like:

<link rel="alternate" hreflang="fr" href="https://example.com/fr/page/" />
<link rel="alternate" hreflang="en" href="https://example.com/en/page/" />
<link rel="alternate" hreflang="x-default" href="https://example.com/" />

If these are missing, that's your first problem. Add them and resubmit your sitemap.

2. Check for canonical tags pointing to the wrong URL

A canonical tag tells Google which version of a page is the "master" copy. If your French page has a canonical tag pointing to the English URL, Google will ignore the French page entirely — even if hreflang is correct.

This is a common mistake with WordPress plugins. The canonical tag should point to the page itself, not to a different language version.

3. Check your URL structure

Google uses URL patterns as a secondary signal. If your translated pages live at completely different paths with no logical connection — like /blog/hello-world and /actualites/bonjour-le-monde — Google has a harder time connecting them.

Best practice is to use a consistent structure: /fr/, /en/, /de/ subdirectories, or subdomains like fr.example.com. This makes the relationship obvious to both Google and users.

4. Check for browser language redirects

If your site uses JavaScript or server-side redirects based on the visitor's browser language, Google may follow those redirects and see only one version of the page. This can cause it to think the other versions are permanently redirected — and therefore not indexable.

Google follows JavaScript redirects as of 2015. If your redirect logic is aggressive, Google may never crawl the French version at all.

5. Check your sitemap

Your XML sitemap should list every language version as a separate URL. If you only list the English version, Google may never discover the French and German pages — or may assume they're not important enough to index.

Make sure each translated URL appears in the sitemap with its correct hreflang annotation.

Why Google treats translations differently from original content

Google's duplicate content detection isn't a simple text comparison. It uses a fingerprinting system that looks at multiple signals:

• HTML structure and DOM similarity
• Image file names and alt text
• Internal link patterns
• Meta descriptions and title tags
• Text overlap (even in different languages, machine translations share common phrases)

When two pages share 80-90% of these fingerprints, Google flags them as duplicates. The language difference alone isn't enough to override that signal — you need explicit markup to tell Google what's happening.

The trade-off: hreflang vs. canonical

There's a common confusion between hreflang and canonical tags. They serve different purposes:

You should never use a canonical tag to point a French page to an English page. That tells Google to ignore the French version. Instead, use hreflang to say "these are siblings, not copies."

What happens if you ignore the problem

If you don't fix the signal gap, the consequences compound over time:

• Your translated pages get deindexed, so you lose all international search traffic
• Google picks one language version as canonical, and users in other markets see the wrong language
• Your international SEO efforts become invisible — you're spending money on translation but getting zero return
• Competitors in those markets rank instead of you

The fix is usually straightforward, but it requires checking all five signals. Missing even one can keep the problem alive.

Practical scenarios: when this happens in the real world

Scenario 1: The WordPress plugin conflict

You use WPML for translations and Yoast for SEO. Both plugins add hreflang tags, but they conflict — Yoast's canonical tag overrides WPML's hreflang. The result: Google sees the canonical tag and ignores the hreflang.

Fix: Configure one plugin to handle hreflang and disable the other's hreflang output. Check that the canonical tag points to the page itself.

Scenario 2: The missing x-default tag

You have French and English versions but no x-default hreflang tag. Google doesn't know which version to show to users who don't match either language. It may pick one arbitrarily and treat the other as duplicate.

Fix: Add <link rel="alternate" hreflang="x-default" href="https://example.com/" /> pointing to your main page or a language selector page.

Scenario 3: The aggressive browser redirect

Your site redirects French visitors to /fr/ automatically. Googlebot crawls from a US IP, so it gets redirected to /en/ and never sees the French page. The French page stays in your sitemap but never gets crawled.

Fix: Use a language selector instead of automatic redirects, or make sure Googlebot can access all versions without redirects.

Limitations: when this advice doesn't apply

There are a few edge cases where the duplicate content flag is legitimate:

• Machine-translated pages with no human review: If your translations are low quality, Google may flag them as spam rather than duplicates. The fix is quality, not markup.
• Same-language duplicates: If you have two English pages with the same content, hreflang won't help. You need a canonical tag or a 301 redirect.
• Thin translated pages: If your translated pages have very little unique content (just a few sentences), Google may not index them regardless of hreflang. The content needs to be substantial.

Key facts at a glance

FAQ: Your next questions answered

Why does Google flag my translated pages even though I have hreflang?

Hreflang alone isn't always enough. Check for conflicting canonical tags, missing x-default tags, or browser redirects that prevent Googlebot from seeing all versions. The hreflang must be paired with correct canonical tags and accessible URLs.

How long does it take for Google to reindex after I fix the issue?

Typically 1-4 weeks after you resubmit your sitemap. Google needs to recrawl the pages and process the new signals. If the problem persists after a month, check for other conflicts.

Should I use subdirectories or subdomains for translated pages?

Subdirectories (/fr/, /en/) are generally easier to manage and pass more link equity. Subdomains (fr.example.com) work but require separate verification in Search Console. Choose based on your site architecture.

Does machine translation cause duplicate content flags?

Machine translation can trigger flags if the output is too similar to the original. Google's fingerprinting looks at text overlap, and machine translations often share common phrases. Human review or post-editing helps differentiate the content.

What's the difference between duplicate content and thin content?

Duplicate content means two pages have the same content. Thin content means a page has too little content to be valuable. Translated pages can be both — if your translation is short and similar to the original, it may get flagged for either reason.

Can I use a canonical tag to point to the original language version?

No. That tells Google to ignore the translated page entirely. Use hreflang to indicate translations, and keep canonical tags pointing to each page itself.

Most plugin installation failures on WP Engine are not random. They happen because WordPress cannot complete one of the steps it runs after you click Install: download the plugin package, unpack it, write the files to /wp-content/plugins/, verify the plugin header, and activate the code. WP Engine is a managed WordPress host, so it also applies stricter security and performance rules than a typical shared server. That combination can turn a small incompatibility into a visible error.

If you see an error like 'The package could not be installed' or 'Could not create directory', the root cause is usually one of four things: a PHP version mismatch, missing dependencies, file permission problems, or a WP Engine security restriction.

Common mistake: assuming the host is broken before checking the plugin package. Many installation failures come from an invalid plugin zip or a missing main PHP file, not from WP Engine.

What actually happens during a plugin installation

WordPress does not copy a plugin into place in one step. The sequence is: upload the zip file, unpack it, read the main plugin file to find the header, copy the files into the plugins folder, run any installation routines, and then activate.

Each step can fail with a different message. For example, if WordPress cannot unpack the zip, you may see a generic 'package could not be installed' error. If the main file is missing or lacks a proper header, you may see 'No valid plugins were found.' If the plugins folder is not writable, you may see 'Could not create directory.'

This is useful because the exact error tells you which step failed. General advice like clearing your cache rarely helps.

Cause 1: PHP version and extension mismatch

Most modern plugins state a minimum PHP version in a file called readme.txt. If your WP Engine environment runs an older PHP version than the plugin needs, installation can fail or the plugin can activate and then trigger a fatal error.

Check the PHP version for your environment in the WP Engine portal. If the version is older than the plugin requires, ask support about upgrading PHP before you install. But first confirm that your theme and all active plugins support the newer version. Upgrading PHP can fix one plugin and break another.

Missing PHP extensions can also cause failures. Some plugins require extensions such as curl, mbstring, zip, or intl. On managed hosting you usually cannot install these yourself. Ask WP Engine support whether the extension is available, or choose a plugin that does not need it.

Cause 2: Missing dependencies and plugin conflicts

Some plugins are not standalone. They are add-ons for a parent plugin. If the parent plugin is not installed and active, the add-on may fail to install or activate. Read the plugin description before downloading.

WordPress also checks the minimum WordPress version and PHP version. If the plugin requires a newer version of either, it may refuse to install.

Plugin conflicts usually show up after activation, not during installation. A security or caching plugin may prevent the new plugin from writing its files. In that case, deactivate the security plugin temporarily and try again. If the install works with it off, you have found the conflict.

Cause 3: File permission and ownership problems

The 'Could not create directory' error is the classic permission problem. WordPress tries to create a new folder under /wp-content/plugins/, and the file system says no. On WP Engine, permissions are usually configured correctly by the platform. If you see this error, check whether the plugins directory was changed by a migration or a security plugin.

Another common issue is manual installation done wrong. When you install via SFTP, you cannot upload the zip file and expect WordPress to unpack it. You need to create a folder like /wp-content/plugins/example-plugin/ and upload the plugin files into that folder. If you upload the zip file by mistake, or if the main PHP file is inside an extra subfolder, WordPress will report that no valid plugins were found.

Cause 4: WP Engine security rules and platform restrictions

WP Engine runs a managed platform tuned for speed and stability. That means some plugin behaviors are restricted. Plugins that write to core WordPress files, use server commands that managed hosts disable, or interfere with caching can be blocked.

These restrictions are a trade-off. You lose the freedom to install every plugin, but you gain a more predictable and secure environment. If a plugin is critical to your business and WP Engine blocks it, ask support if there is an approved workaround. The usual answer is to choose an alternative plugin.

Keep in mind that 'No valid plugins were found' is not a security restriction. It means the file you uploaded is not a valid plugin. This can happen with a corrupted zip, a zip that contains a nested folder with no main file, or a file that was never actually a plugin.

How to diagnose any failure: a practical order

1. Write down the exact error text. Screenshot it if you can.
2. Check the plugin's requirements on the WordPress.org page or the developer's site.
3. Check your PHP version and WordPress version in the WP Engine portal.
4. Look at the plugin folder in /wp-content/plugins/ via SFTP to see if a partial install is left over.
5. Try a manual SFTP install: place the unzipped plugin files in /wp-content/plugins/plugin-name/.
6. Deactivate caching or security plugins temporarily and try again.
7. Contact WP Engine support with the exact error message.

Common mistake: skipping step 1 and trying fixes blindly. The error message is the fastest way to tell which step failed. If you do not understand it, paste it into a support ticket rather than guessing.

Key facts: SEATEXT on WP Engine

These facts are specific to installing SEATEXT on WP Engine. They come from SEATEXT's integration documentation.

SEATEXT's installation page also notes that the installation process is secure and the AI remains inert until activated.

Limitations: when this advice does not apply

This article focuses on plugin installation failures in a WordPress site hosted on WP Engine. If you are on shared hosting, VPS, or another managed host, the causes and fixes can vary.

Some plugins are installed as must-use plugins in /wp-content/mu-plugins/. The normal WordPress installer does not handle those, so the errors described here do not apply to them.

If a plugin has been abandoned and does not support current PHP, no permission or version change will make it work properly. The reliable fix is to find a maintained alternative.

Finally, if WP Engine itself is having an outage, plugin installation can fail for reasons outside your site. Check WP Engine's system status before troubleshooting further.

FAQ

Why does 'No valid plugins were found' appear?

That error means WordPress could not find a valid plugin header in the package. The main PHP file may be missing, the file header may be incomplete, or the zip file may be corrupted. Re-download the plugin and re-upload it. If you are using SFTP, make sure the main plugin PHP file is in the plugin folder, not inside an extra subfolder.

Can I install a plugin manually on WP Engine?

Yes. Use SFTP to connect to your site, go to /wp-content/plugins/, create a folder for the plugin, upload the unzipped files, then activate the plugin from the WordPress admin. WP Engine security rules still apply.

Does WP Engine block certain plugins?

WP Engine's managed platform restricts plugin behaviors that can harm performance or security. If your plugin is blocked, contact support or choose an alternative plugin. Not every installation failure is caused by a block, though.

Why does a plugin install on another host but not on WP Engine?

Because the two hosts use different PHP versions, file permissions, caching layers, and security rules. A plugin written for a permissive environment may assume it can write files or run commands that WP Engine does not allow.

How do I check the PHP version on WP Engine?

Log in to the WP Engine portal, open the site's settings, and look for the PHP version. If you cannot find it, ask WP Engine support. Compare that version to the plugin's stated requirement.

What should I do after fixing the error?

Try the install again. If it succeeds, activate the plugin and test the front end and admin. If a fatal error appears, disable the plugin via SFTP by renaming its folder, then check for conflicts with your theme or other plugins.

Translation scripts break Webflow designs because they modify the DOM directly — inserting translated text nodes, wrapping elements in new containers, or replacing fonts — which disrupts the precise pixel calculations Webflow's layout engine relies on. The result is often text overflow, collapsed whitespace, shifted breakpoints, and misaligned flex or grid children.

Most third-party translation tools (widgets, proxy-based services, or DNS-level rewriters) operate by rewriting HTML before it reaches the browser or by injecting translation markup after render. Both approaches change element dimensions, line heights, and font metrics in ways Webflow's responsive breakpoints don't anticipate. SeaText takes a different path: it reads your live HTML and CSS, then serves translated text through an overlay layer that preserves your original classes, breakpoints, and layout engine entirely.

Common Symptoms of Translation-Related Layout Breaks

When a translation script conflicts with Webflow, the signs appear quickly:

• Text spills out of buttons, cards, or nav items because translated strings are longer
• Line height changes push adjacent elements down, breaking vertical rhythm
• Font fallback or @font-face loading shifts cause flash of unstyled text (FOUT) or layout shift (CLS)
• Grid and flex children wrap unexpectedly at specific breakpoints
• Fixed-height containers cut off translated content
• Custom code interactions (sliders, tabs, accordions) stop working because DOM structure changed

These symptoms usually appear only in translated languages, not in the original language, making them easy to miss during QA.

Why Translation Scripts Conflict With Webflow's Layout Engine

Webflow builds responsive layouts using a combination of CSS Grid, Flexbox, absolute positioning, and breakpoint-specific style overrides. Each element's size and position are calculated at render time based on content, viewport, and cascade order. When a translation script alters that content — or the elements containing it — after Webflow's initial render, the browser recalculates layout without Webflow's breakpoint logic reapplied.

Three common translation architectures cause this:

• DOM injection widgets (e.g., Google Translate widget) insert <font> tags, <span> wrappers, and inline styles directly into your markup, breaking CSS selectors and inheritance.
• Proxy-based translation rewrites HTML on the server before delivery, often stripping or modifying Webflow's generated classes (.w-dyn-item, .w-container, etc.) and inline breakpoint styles.
• DNS-level rewriters serve a fully translated copy of your site from a subdomain, but the translated HTML may not match Webflow's exact output, causing style drift.

All three approaches change the very thing Webflow's layout engine depends on: the relationship between content, container, and breakpoint.

Diagnostic Sequence: Identify the Root Cause in 5 Steps

1. Reproduce in the original language. Disable translation and verify the layout works perfectly at all breakpoints. If it's already broken, the issue isn't translation.
2. Enable translation for one language. Use browser dev tools to inspect a broken element. Check if new wrapper elements (<font>, <span class="goog-te-gadget">, etc.) appear around your text nodes.
3. Compare computed styles. In dev tools, toggle between original and translated views. Look for changes in line-height, font-family, display, white-space, or overflow on the element or its ancestors.
4. Test at each breakpoint. Resize the viewport across Webflow's breakpoints (desktop, tablet, mobile landscape, mobile portrait). Note where the layout fails — this often reveals which breakpoint's styles are being overridden.
5. Check for font loading issues. Open the Network tab, filter for fonts. Translated languages may trigger fallback fonts or additional @font-face requests that delay render and cause layout shift.

If steps 2–4 show injected markup or style overrides, the translation method is modifying your DOM. If step 5 shows font delays, the translation service is serving different font subsets. Both require a translation approach that doesn't touch your rendered output.

How Different Translation Approaches Affect Layout Stability

The overlay approach reads your rendered page, extracts text content, translates it, and displays the translation in a layer that sits visually over the original without altering the DOM. Your Webflow classes, breakpoint styles, interactions, and font loading remain exactly as designed.

SeaText's Approach: Automatic Translation Without Layout Disruption

SeaText's Website Translation Agent translates every page, headline, button, and offer into up to 125 languages while preserving your original layout. It detects each visitor's language, translates Webflow pages instantly, and keeps new posts, products, and updates translated in the background. When you publish a new Webflow page, product, post, or headline, SeaText sees it and translates it automatically — you do not need to remember to send every update through a translation workflow.

Unlike tools that make you manage language limits, page limits, word counts, DNS changes, or manual translation requests, SeaText focuses on letting AI do the translation job automatically after one install. The system watches the page for new text and translates it in the background, preserving brand context and optimizing localized pages for conversion.

Key Facts

Limitations and When This Advice Applies

This diagnostic applies specifically to Webflow sites where a third-party translation script was added after design completion. It does not cover:

• Webflow's native Localize feature (which manages translations within Webflow's own layout system)
• Sites built on other platforms (WordPress, Shopify, Framer) where layout engines differ
• Translation implemented during initial build where design accounted for multilingual content
• Cases where layout breaks exist in the original language before translation is enabled

If you're using Webflow Localize and still see breaks, the cause is likely content-length differences across locales — a design challenge, not a script conflict. For script-induced breaks on other platforms, the same DOM-injection principles apply but the specific class names and breakpoint structures will differ.

FAQ

Why does my design look fine in English but break in German or French?

German and French strings are typically 20–35% longer than English. When a translation script injects longer text into fixed-width containers or elements with overflow: hidden, the content spills or pushes adjacent elements. Webflow's breakpoints don't re-evaluate because the DOM change happened after initial render.

Can I fix layout breaks by adding custom CSS?

Sometimes, but it's fragile. You'd need to target the translation script's injected markup (which may change) and override styles for each language. A more stable fix is switching to a translation method that doesn't modify your DOM.

Does SeaText require me to change my Webflow hosting or DNS?

No. SeaText activates with a single script install on your existing Webflow site. No DNS changes, subdomain setup, or hosting migration needed.

What happens when I publish a new page in Webflow?

SeaText watches the page for new text and translates it in the background automatically. You do not need to remember to send every update through a translation workflow.

Can I still manually correct important translations like legal copy or brand names?

Yes. SeaText lets you control important translations while AI handles the rest automatically.

Will SeaText slow down my page load or affect Core Web Vitals?

The overlay approach adds minimal client-side overhead. Your original HTML, CSS, and font loading remain unchanged, so LCP, CLS, and FID are not impacted by the translation layer.

How does SeaText differ from Weglot, DeepL, or Google Translate widget?

The main difference is automation plus unlimited free activation. Many tools make you manage language limits, page limits, word counts, DNS changes, or manual translation requests. SeaText focuses on letting AI do the translation job automatically after one install.

Learn more about this service

See how this page can help with your next step.

Learn more

Why does SeaText AI require a Webflow site plan for Full Functionality?

SeaText AI mistranslates certain phrases because machine translation predicts language from patterns instead of understanding meaning. Short Shopify copy makes it harder: a button label, product title, or sale banner has almost no context for the model to work with. Idioms, brand names, and niche product terms are the usual casualties.

The good news is that this is not a random failure you have to accept. Mistranslations have identifiable causes, and most can be caught with a short review of high-risk phrases. This article shows you those causes, gives a simple diagnostic, and explains how to make SeaText translation more reliable in your Shopify store.

Why machine translation guesses (and why that matters)

Translation models like the one inside SeaText do not read a sentence the way a human does. They calculate the probability of the next word based on billions of examples from the web, books, and product catalogs. The output is the statistically most likely translation, not a guaranteed accurate one.

For a long product description, the surrounding sentences supply context. For a Shopify menu item or a call-to-action button, the model only has a few words. A phrase like 'one-time' can mean 'once' or 'single' depending on context. Without context, the model picks whichever pattern was more common in its training data.

This property affects every machine translation system. It is not a defect specific to SeaText. It is a trade-off between speed and coverage on one side and the risk of an odd phrase on the other.

The three main causes of mistranslation

Missing context

Product titles and menu labels rarely explain what they refer to. The word 'hook' in a hardware store means one thing; 'hook' in a marketing course means another. A translation model that sees the word alone may translate it literally. The safest fix is to write source copy with enough context: use 'camera lens' instead of 'lens' if that is what you sell.

Idioms and cultural expressions

Phrases like 'best-in-class,' 'set for life,' or 'made to last' do not translate word-for-word. In some languages they become nonsense or take on a different meaning. Marketing language makes this harder because metaphors are everywhere.

If a phrase depends on a pun or a cultural reference, no current model can guarantee a clean translation. A human translator would rewrite the phrase for the target market. An AI will usually attempt a literal version.

Training data gaps for your language pair

Some language pairs have far more training data than others. English to Spanish, French, or German works well because those languages appear everywhere online. English to a low-resource language, or a pair used mostly in a niche community, has fewer examples. Less data means more guesswork and more visible errors.

Technical product vocabulary makes the gap larger. A store selling medical devices or specialized tools will have terms that rarely appear in training data. The translation agent can handle general ecommerce copy, but your custom product names may need a manual check.

What happens when a mistranslation slips through

A mistranslated page in a storefront is not just embarrassing. A wrong button label can stop a customer at checkout. A mistranslated shipping policy can create support tickets. A product name that sounds offensive in the target language can damage trust quickly.

The cost grows when the mistranslated page gets indexed by search engines. Search users see the wrong meaning in a snippet and click away. You do not need to track a complex metric to understand the pattern: unclear copy adds friction, and friction lowers conversions.

That is why translation is worth treating as a review task, not a fully automated output.

Diagnostic: is it accuracy, fluency, or tone?

Before you fix a phrase, decide what kind of problem you are looking at. The fix is different in each case.

• Accuracy - the translated word means something different from the source. Example: 'free shipping' turns into 'no shipping'. Fix it by simplifying the source copy or editing the rendered translation.
• Fluency - the meaning is mostly right, but the sentence sounds awkward. A native speaker would understand it but would not write it like that. This usually needs a human edit.
• Tone - the meaning is correct, but the formality or marketing energy is wrong. Example: 'You'll love it' becomes too casual or too stiff. Tone is a judgement call, and no automatic system fully matches your brand voice.

If the word is simply wrong, it is an accuracy error. If the translation is understandable but clumsy, it is a fluency issue. If you are annoyed by the style, that is a tone issue.

A fix-and-prevent workflow

Use this order whenever you see a suspicious translation.

1. Keep a list of high-risk phrases. Product titles, button labels, category names, holiday promos, and phrases that use imagery.
2. Simplify the source copy. Short, concrete words usually translate better than long metaphors. Write 'exactly as shown' instead of 'you will not believe your eyes.'
3. Activate the translation agent and review its output. Pay attention to the languages that matter most for your customers.
4. Correct anything that confuses you. If the target sentence does not make sense in a language you know, ask a native speaker or a trusted colleague to review.
5. Test the live page. Switch the storefront to the translated language and check the main flows: search, product page, cart, checkout.
6. Reuse the same source phrasing. If the same product term appears over and over, the model is more likely to settle on one consistent translation.

What SeaText offers for Shopify stores

SeaText's Website Translation Agent is built for stores that want broad language coverage without a manual localization project. The company describes it as translating and optimizing your website and product content into 125 languages. Translation is one agent among many; you activate only what you need.

For Shopify, setup happens inside your theme. You go to Online Store, choose your theme, select Edit code, and paste the SeaText snippet into theme.liquid. The AI stays inactive until you activate it, so installing the code will not change your store by itself. The Shopify integration page also shows how to wrap product descriptions with a data attribute so the AI knows where it is allowed to work.

SeaText translation at a glance: key facts

The table is a snapshot, not a full feature list. For implementation details, check the Shopify integration page.

Limitations and honest expectations

No machine translation, including SeaText AI, can guarantee a perfect result for every phrase and language pair. Plan for these limits:

• Short marketing phrases will sometimes sound flat or literal.
• Puns, rhymes, and brand slogans may lose their effect.
• Low-resource language pairs can show more errors than common ones.
• Product-specific terms unique to your store need a human review.
• For regulated or highly sensitive markets, professional localization is still a safer option.

These limits do not make AI translation useless. The speed and coverage are real benefits. The useful mental model is: use the AI for scale, then review the pieces that carry big meaning.

When a mismatch is not a mistake

Sometimes the translated phrase is technically correct but feels different because languages are different. Some languages use longer greetings. Some put adjectives after nouns. Some have no exact word for 'checkout.'

Use this test: would a native speaker of the target language understand the phrase without confusion? If yes, the translation is fine even if it differs from English. If no, it is a mistranslation and worth fixing.

Frequently asked questions

Why does SeaText translate long product descriptions better than short buttons?

Longer text gives the model more context. A button like 'Go' or 'Get' has almost nothing to anchor on, so the model has to guess.

Which Shopify phrases are most likely to be mistranslated?

Short button labels, sale banners, slogans, product feature bullets, and shipping or policy terms. These rely on context, idioms, or specific industry vocabulary.

Can I control where SeaText translation runs in my Shopify store?

Yes. The Shopify setup uses a tag around product descriptions, and the translation agent is described as giving control. You choose how to scope the code and can review the live result before relying on it.

Will installing SeaText change my store before I activate it?

No. The company states that the AI remains inert until activated, so the code itself does not alter your content.

Does SeaText work on development domains?

Not reliably. Development URLs such as localhost are restricted, and each domain needs its own SeaText account. Use a real domain for testing.

What should I do when I find a mistranslation?

First classify it: accuracy, fluency, or tone. Then simplify the source phrase, ask a native speaker for the right wording, edit the rendered translation if possible, and re-test the page.

SeaText connects to Thinkific through a JavaScript snippet you paste into the Site Footer Code field. After saving, you must visit your own site for at least 40 seconds so the script can phone home and register the domain. If that handshake never completes — or completes but later breaks — content stops syncing.

How the SeaText‑Thinkific connection works

The integration is client‑side only. SeaText gives you a unique JavaScript block. You place it in Thinkific's Settings → Code & Analytics → Site Footer Code and save. The script then waits for a real browser visit. When you (or any visitor) load a page and stay for ~40 seconds, the script sends a signal to SeaText's servers, linking the domain to your account. Once linked, SeaText can push translations, variants, and AI‑generated content back into the page via the same script.

Because there is no server‑to‑server API, the connection depends entirely on that browser‑side handshake and on the script continuing to load on every page view. Anything that blocks the script, revokes its token, or changes the page structure can interrupt sync.

Common failure points after initial setup

• Expired or revoked API token. SeaText rotates tokens periodically for security. If the token embedded in your JavaScript snippet expires, the script can no longer authenticate with SeaText's backend.
• Thinkific plan limits. Lower‑tier Thinkific plans may strip or block custom JavaScript in the footer, or they may limit the number of external scripts that load.
• Mismatched content types. SeaText expects to rewrite HTML text nodes. If your Thinkific theme renders course content via iframes, React components, or dynamic AJAX loads after the initial page load, the script may not see the content to translate or optimize.
• Rate‑limit throttling. SeaText's backend enforces per‑domain request limits. A sudden traffic spike (e.g., a launch or ad campaign) can trigger temporary throttling, pausing sync until the window resets.
• Script removal or modification. Theme updates, site migrations, or accidental edits to the Code & Analytics field can delete or corrupt the SeaText snippet.
• Browser‑side blockers. Ad blockers, privacy extensions, or Content Security Policy headers can prevent the SeaText script from loading or making outbound requests.

Diagnostic sequence to pinpoint the issue

1. Verify the snippet is still present. In Thinkific admin, go to Settings → Code & Analytics → Site Footer Code. Confirm the exact SeaText JavaScript block is there and unchanged.
2. Check the browser console. Open your live site, open DevTools (F12), and look at the Console tab. Search for "SeaText" or "SEATEXT". Errors like 401 Unauthorized, 429 Too Many Requests, or blocked by CSP point directly to token expiry, rate limits, or CSP issues.
3. Confirm domain linkage in SeaText. Log into your SeaText dashboard. The connected domain should appear next to the SeaText logo at the top. If it shows "Not connected" or is missing, the handshake failed.
4. Test with a clean browser session. Open an incognito/private window, visit your site, stay 40+ seconds, then refresh the SeaText dashboard. If the domain appears, the script works but something in your regular browser (extension, cache) was blocking it.
5. Review Thinkific plan features. If you're on a Basic or Start plan, verify that custom JavaScript in the footer is allowed. Upgrade or contact Thinkific support if the feature is gated.
6. Inspect page structure. View source on a course lesson page. If the content you expect SeaText to modify lives inside an <iframe> or loads via fetch() after the initial HTML, SeaText cannot reach it.
7. Contact SeaText support with console logs. If steps 1‑7 don't reveal the cause, export the console errors (right‑click → Save as) and send them to SeaText support. They can check server‑side token status and rate‑limit logs.

Key facts

Limitations and when this advice doesn't apply

• This diagnostic covers the standard JavaScript integration only. If you use a custom server‑side proxy or a Thinkific app-store integration (not documented in the source pack), the failure modes differ.
• Thinkific's platform changes (e.g., new CSP headers, script sanitization) can introduce new failure modes not listed here.
• Rate‑limit thresholds are not published; they may change without notice.
• The 40‑second activation window assumes a typical broadband connection. Extremely slow connections may need longer.

Practical scenarios

Scenario A: Launch day traffic spike

You run a paid campaign sending 10,000 visits in an hour. SeaText's per‑domain rate limit kicks in at ~2,000 requests/minute. Sync pauses for 15 minutes, then resumes automatically. No action needed unless the pause exceeds your tolerance.

Scenario B: Theme update removes footer code

You switch Thinkific themes. The new theme overwrites the Code & Analytics field. The SeaText snippet disappears. Re‑paste the snippet, save, then revisit the site for 40 seconds to re‑activate.

Scenario C: Course content loads in an iframe

Your lessons use a third‑party video player embedded via iframe. SeaText cannot rewrite text inside that iframe. You must either move the text out of the iframe or accept that those sections won't be translated/optimized.

FAQ

Why does the SeaText dashboard show my domain as connected but content still isn't syncing?

The handshake succeeded, but the script may be blocked from making subsequent API calls. Check the browser console for 401/429 errors or CSP violations.

Can I use SeaText on Thinkific's free plan?

The free plan allows custom JavaScript in the footer, so the snippet will load. However, free‑plan sites often have lower traffic limits and may not support all SeaText agents (e.g., bot‑refund reporting needs paid ad traffic).

How often does SeaText rotate API tokens?

SeaText does not publish a fixed schedule. Tokens typically rotate every 30–90 days or after security events. When rotation occurs, the JavaScript snippet must be refreshed from the SeaText dashboard and re‑pasted into Thinkific.

What if my Thinkific site uses a Content Security Policy (CSP)?

Add https://cdn.seatext.com (or the domain shown in your snippet) to the script-src and connect-src directives in your CSP header. Thinkific's CSP is managed by Thinkific; you may need to request the change via their support.

Does SeaText sync course completion data back to Thinkific?

No. SeaText reads page content to generate translations, variants, and AI optimizations. It does not write learner progress or completion status to Thinkific's database.

How do I know if rate limiting is the problem?

In the browser console, look for HTTP 429 responses from SeaText endpoints. The response body usually includes a retry-after header indicating seconds until the next allowed request.

Can I run SeaText alongside other optimization scripts (e.g., Hotjar, Google Optimize)?

Yes, multiple scripts can coexist in the footer. Load order matters: place SeaText after analytics but before heavy A/B testing scripts to ensure it sees the final DOM.

When the SeaText script throws errors in your browser console, the problem is almost always one of a handful of configuration or environment issues. The script itself is a lightweight async loader (under 15 KB) that runs before visual paint, so it does not cause layout shifts or PageSpeed penalties when it works correctly. If you see red lines in the console, start by checking the API key, your Content Security Policy, the script's position in your HTML, and whether the browser can reach SeaText's CDN.

How the SeaText Script Loads and Executes

The SeaText snippet is designed to load asynchronously via the async attribute on the script tag. This means the browser fetches and executes it without blocking page rendering. On first load the script writes an anonymous identifier into localStorage so it can recognize returning visitors across sessions. It then contacts the SeaText CDN to fetch the active AI agents and any personalization rules for the current visitor. Because the script runs synchronously in under 15 ms before paint, any network latency or blocking policy shows up immediately as a console error rather than a visible layout glitch.

In a single-page application (React, Vue, Angular) the snippet should be placed in the entry HTML file (typically index.html) inside the body tag, or in the framework's bootstrap file where the app mounts. The documentation recommends verifying the Console and Network tabs after a production build to confirm the script loads without errors. If you inject the snippet via a tag manager or a framework-specific helper, the async attribute may be stripped or the script may execute before the DOM is ready, both of which surface as console warnings.

Invalid or Missing API Key

Every SeaText project receives a unique API key that the script sends with its initial handshake to the CDN. If the key is mistyped, revoked, or omitted entirely, the CDN responds with a 401 or 403 status and the script logs an authentication error. The error message typically reads SeaText: Invalid API key or SeaText: Project not found. Copy the key directly from the SeaText dashboard (Settings → Installation) and ensure no extra whitespace or line breaks are included. If you rotate keys, update the snippet on every environment (staging, preview, production) before the old key is revoked.

Content Security Policy Blocks

A strict CSP is the most common cause of silent script failures. SeaText needs three permissions:

• script-src must allow the SeaText CDN origin (e.g., https://cdn.seatext.com) or the nonce/hash of the inline snippet.
• connect-src must allow the same origin so the script can fetch agent configurations and send analytics events.
• img-src may be required if the bot-protection agent uses a tracking pixel.

Script Loading Order Conflicts in SPAs

In React, Vue, or Angular the SeaText snippet must run before your application mounts, otherwise the virtual DOM may replace the script tag or the script may miss the initial navigation event. The documentation advises placing the snippet in index.html inside the body tag. If you use a framework-specific integration (e.g., react-helmet, vue-meta, Angular's APP_INITIALIZER), ensure the script tag is rendered with the async attribute preserved and that it executes before any route change. A common mistake is adding the snippet via a component lifecycle hook (useEffect, mounted()), which runs after hydration and causes the script to re-initialize on every route change, flooding the console with duplicate initialization warnings.

Network Connectivity and CDN Issues

The script fetches its configuration from SeaText's global CDN. Corporate firewalls, ad blockers, privacy extensions (uBlock Origin, Privacy Badger), or DNS filtering can block the request. In the Network tab look for a request to cdn.seatext.com (or the region-specific endpoint) with a status of 0, net::ERR_BLOCKED_BY_CLIENT, or a DNS failure. If the request succeeds but returns a 5xx error, the issue is on SeaText's side—check the status page or contact support. For environments that must operate offline or behind an air-gapped network, SeaText does not currently offer a self-hosted fallback, so the script will log a connectivity error on every page load.

Local Storage and Cross-Origin Restrictions

The script writes a visitor ID to localStorage under the SeaText domain. If your site runs in an iframe, uses a sandboxed domain, or has a Permissions-Policy header that denies local-storage access, the write fails and the script logs a warning. Similarly, if your SPA serves content from multiple subdomains (e.g., app.example.com and checkout.example.com) and the script is only loaded on one, the ID cannot be shared across origins. The documentation flags this as a cross-origin consideration: ensure the snippet is present on every entry point or configure a shared top-level domain cookie strategy if you need persistent identity across subdomains.

Diagnostic Checklist: Verify Each Layer in Order

1. Open DevTools → Console. Filter for SeaText to isolate relevant lines.
2. Open Network tab. Reload. Confirm a request to the SeaText CDN returns 200 and a JSON payload containing agents and config.
3. Check the script tag in Elements: verify async attribute, correct src, and that the API key parameter matches the dashboard.
4. Inspect CSP headers (Content-Security-Policy and Content-Security-Policy-Report-Only) for script-src and connect-src directives.
5. Test in an incognito window with extensions disabled to rule out ad blockers.
6. If using an SPA, add a console.log('SeaText snippet executed') immediately after the snippet in index.html to confirm execution order.
7. Verify localStorage.getItem('seatext_id') returns a value after load; if null, check Permissions-Policy and iframe sandbox attributes.

Key Facts

Limitations and When This Guidance Does Not Apply

This diagnostic covers the SeaText client-side snippet only. Server-side rendering integrations, custom proxy setups, or enterprise firewall configurations that terminate TLS and inspect payloads may introduce additional failure modes not described here. If you use a tag manager (GTM, Tealium, Segment) to inject the snippet, the tag manager's own CSP, data layer timing, and consent rules add another layer of complexity—consult the tag manager's debug console first. The article also assumes a standard browser environment; headless crawlers, AMP pages, or email clients that strip scripts will not execute SeaText at all, which is expected behavior, not an error.

Frequently Asked Questions

Why do I see "SeaText: Duplicate initialization" in the console?

The snippet runs more than once per page load. In SPAs this usually means the snippet is inside a component that re-renders on route change. Move it to the static index.html or ensure your framework renders it only once during bootstrap.

Can I self-host the SeaText script to avoid CDN blocks?

Not currently. SeaText does not provide a self-hosted bundle; the script must fetch its agent configuration from the CDN on every session.

Does SeaText work if the visitor has JavaScript disabled?

No. The script is a client-side JavaScript module; without JS the personalization, translation, and bot-detection agents cannot run.

How do I know which AI agents are active for a given visitor?

After a successful CDN handshake, the response JSON includes an agents array. Log it in the console or inspect the Network response to see which agents (CRO Optimizer, Bot Refund, Translation, etc.) are enabled for your project.

Will SeaText errors hurt my SEO or Core Web Vitals?

Console errors themselves do not affect rankings. However, if a CSP block prevents the script from loading, you lose the conversion lift and bot protection the agents provide. The script's synchronous execution under 15 ms ensures zero CLS impact when it loads successfully.

What should I do if the CDN returns a 500 error?

Retry after a few minutes. If the error persists, open a support ticket with the request ID from the Network tab (header x-request-id) so SeaText engineering can trace the failure.

Can I run SeaText alongside other translation or personalization tools?

Yes, but avoid multiple tools rewriting the same DOM nodes simultaneously. SeaText's Variants Editor lets you scope which elements it may modify; configure other tools to exclude those selectors to prevent conflicts.

What dynamic landing page personalization actually does

Dynamic landing page personalization rewrites headlines, offers, product blocks, and calls to action on a single URL so the page reflects the visitor's keyword, campaign, referral source, account data, or behavior. Instead of building dozens of static pages, you keep one page and let an AI agent swap copy blocks in milliseconds after the click.

SeaText's Google Ads Landing Page Agent rewrites ad landing pages by campaign intent, while the Visitor Source Rewrite Agent matches pages to Google, Meta, email, and referrals. The AI Personalization Agent adapts site copy to visitor context using CRM data, enrichment, or account lists (S1, S3, S4, S7).

Why the generic page fails

Most websites serve the same landing page to every click. A visitor who searched "enterprise CRM pricing" sees the same headline as someone who clicked a brand awareness ad. The mismatch between ad promise and page content drives bounce rates up and conversion rates down. Industry benchmarks show average landing page conversion rates stuck around 4% when the message does not match the intent (SERP research).

SeaText notes: "Without SEATEXT every keyword lands on the same generic page, so visitors do not see what they searched for and leave" (S5).

How the personalization layer works

Signal detection

The system reads the inbound signal: UTM parameters, referrer, keyword, device, location, or a personalized link token. For account-based campaigns, it can pull company size, budget tier, prior emails, and CRM notes from Clay, LinkedIn, HubSpot, Salesforce, or a CSV upload (S4).

Copy selection and rewrite

Based on the signal, the agent chooses which copy blocks to swap. Headlines, value propositions, proof points, product descriptions, and CTAs are rewritten to match the visitor's context. The same URL renders different copy for each segment without creating new pages (S4, S5).

Testing and rollout

Variants are tested automatically. The AI A/B Testing Agent generates variants, measures lift, and rolls out winners under enterprise review controls (S1, S3, S7). This continuous loop means the page keeps improving without manual test setup.

What changes when you add personalization

Paid traffic efficiency

When each keyword lands on a page that mirrors the search term, click-to-lead rates rise. SeaText reports up to +12% more conversions from Google Ads campaigns and up to +34% conversion rate lift from existing pages (S1).

Account-based and outbound conversion

Personalized links for outbound, ABM, LinkedIn, and retargeting campaigns let the page adapt to the specific account. In a MediaCom-style example, smaller clients saw budget-sensitive messaging while larger accounts saw deeper custom-fit pages built from email and account context, producing more closed leads and up to 65% expected lift (S4).

Organic and AI-search visibility

The AI SEO Content Factory publishes indexed Q&A pages for long-tail traffic, expanding the surface area for personalization entry points (S3, S6).

Main approaches and trade-offs

Choose AI agent rewrite if you run multiple paid campaigns, have an outbound/ABM motion, or serve distinct segments from one core offer. Choose manual variants only for a handful of stable, high-stakes pages where legal or brand review mandates exact wording. Avoid separate pages unless the offer structure fundamentally differs.

Decision framework: do you need dynamic personalization?

1. Count your entry signals. Do you have 10+ active keywords, 3+ traffic sources (paid, email, referral, organic), or an outbound list with account data? If yes, personalization pays off.
2. Measure message mismatch. Compare ad headline to landing page headline for your top 20 keywords. If they differ, you are losing conversions.
3. Check data readiness. Can you pass UTM parameters, referrer data, or a personalized link token? Do you have CRM fields (company size, industry, stage) or enrichment access (Clay, LinkedIn)?
4. Estimate test velocity. How many copy tests can your team run per month manually? If under 5, an autonomous agent will outpace you.
5. Run a pilot. Deploy the agent on one high-traffic page with enterprise review on. Measure lift over 2-4 weeks before expanding.

Practical scenarios

Google Ads campaign with 100+ keywords

Each keyword triggers a headline and offer rewrite. A search for "apartment for rent studio downtown" shows studio-specific copy and a tour CTA; "apartment for rent 2 bedroom" shows family-friendly proof points (S5).

Outbound ABM sequence

Sales sends a unique link per prospect. The page rewrites headlines, proof, product copy, and CTAs using Clay enrichment, CRM notes, and email history. Smaller accounts see budget messaging; enterprise accounts see security and integration depth (S4).

Email nurture to product page

Visitor clicks from a nurture email about a specific feature. The Visitor Source Rewrite Agent detects the email referrer and highlights that feature in the hero, swaps the CTA to a trial for that module (S1, S3).

Retargeting warm visitors

Returning visitor who viewed pricing sees a "Book demo" CTA and social proof from their industry. New visitor sees "Start free pilot" and generic proof (S8).

Limitations and when this advice does not apply

• Low traffic volume. If a page gets under 500 visits/month, statistical significance for variant testing takes too long. Focus on traffic acquisition first.
• Single offer, single audience. If you sell one product to one persona with one traffic source, a well-written static page beats a personalization engine you don't need.
• Regulatory copy lock. Industries where every word requires legal sign-off (pharma, financial advice) may need manual variant approval for each segment, reducing speed advantage.
• No reliable signals. If you cannot pass UTM, referrer, or personalized tokens — and have no CRM/enrichment data — the agent has nothing to personalize on.
• Brand voice rigidity. If leadership rejects any AI-generated copy without line-by-line review, the autonomy benefit disappears.

Key facts from SeaText

Terminology quick reference

• Dynamic landing page: A single URL that serves different copy blocks based on real-time visitor signals.
• Message match: Alignment between the ad/referral promise and the landing page headline and offer.
• Personalized link: A unique URL per prospect that carries account context (token) so the page can adapt on arrival.
• Variant: An alternative version of a copy block (headline, CTA, proof point) tested against the control.
• Enterprise review gate: A workflow where winning variants pause for human approval before site-wide rollout.

FAQ

How fast does the page rewrite after the click?

Milliseconds. The agent reads the signal and swaps copy blocks before the page finishes rendering. No redirect, no flicker.

Do I need to create new pages for each keyword or account?

No. One URL, one template. The agent injects variant copy into designated slots. You manage the template; the agent manages the variants.

Can I edit or reject AI-generated variants?

Yes. You can edit variants, delete them, add your own, and control what percentage of traffic sees experimental copy. Enterprise review gates are default (S9).

What data sources does the personalization agent accept?

UTM parameters, referrer headers, personalized link tokens, Clay.com enrichment, LinkedIn data, HubSpot, Salesforce, CRM exports, and CSV uploads (S4).

Does this hurt SEO or create duplicate content?

No. The canonical URL stays the same. Search engines see the base page; personalization happens client-side for known signals. The AI SEO Content Factory separately publishes indexable Q&A pages for long-tail organic traffic (S3, S6).

What is the minimum traffic to make testing worthwhile?

Roughly 500-1,000 visits per month per page gives enough data for confident variant decisions in 2-4 weeks. Lower traffic? Prioritize acquisition first.

How does bot protection fit in?

The Bot Protection Agent filters invalid clicks before they poison retargeting audiences and prepares refund-ready evidence for Google, Meta, TikTok, and Reddit. This protects your personalization data from bot noise (S1, S3, S7).

What AI A/B testing actually does

Traditional A/B testing requires marketers to hypothesize variants, build them, configure traffic splits, wait for statistical significance, and then manually promote the winner. SeaText's AI A/B Testing Agent compresses that loop: it reads your existing page copy, creates multiple controlled wording variants (headlines, buttons, product descriptions, checkout reassurance), serves them to live visitors, measures which version drives more conversions, and automatically scales the winner — all while you retain approval control and exposure limits.

Why the automated variant loop outperforms manual testing

• Volume of ideas: The agent can generate ten or more variants per element simultaneously, far more than a human team can write and QA.
• Continuous cycle: As soon as a winner emerges, the agent keeps testing new micro-changes, compounding lift over time instead of stopping after a single test.
• No traffic waste: Variants are served only to the fraction of visitors you allocate, and losing variants are retired automatically.
• Preserves brand voice: The system makes small, controlled wording changes — it does not rewrite promises or reposition the offer.

Where it applies first for fastest impact

Start on high-traffic pages where visitors already show buying intent: landing page headlines, hero copy, calls to action, product descriptions, checkout reassurance, and lead forms. On ecommerce stores, the same agent optimizes product names, benefit lines, and descriptions to increase add-to-cart and purchase rates. The agent integrates with the existing CMS or store platform (WordPress, Shopify, WooCommerce, Weebly, and 200+ others) without code changes.

Control and safety guardrails

Marketers approve or reject every AI-generated variant before it goes live, set maximum traffic exposure for experiments, and keep the original copy available as a fallback. Conversion reporting is broken down by page, keyword, and variant so the team can audit exactly what changed and why it won.

What the sources actually cover

The supplied material describes SeaText, an AI marketing platform that deploys autonomous agents for conversion optimization, bot protection, translation, and personalization. It outlines SeaText's commercial model: a free 1-month pilot trial, enterprise demo booking, and agent-based activation (e.g., CRO Optimizer, Bot Refund Agent, Google Ads Landing Page Agent). Pricing pages are referenced repeatedly with "Click here for pricing" links, but no dollar amounts, tiers, or competitor comparisons appear.

Why Copy.ai's cost isn't addressed

None of the eight source pages mention Copy.ai, its feature set, its pricing tiers, or any rationale for its pricing. The sources are exclusively about SeaText's architecture — agents that rewrite landing pages per campaign intent, detect and document bot clicks for ad refunds, translate and optimize across 125 languages, and run continuous A/B tests with enterprise review controls. Without any Copy.ai data in the pack, a direct answer to "why is Copy.ai so expensive" cannot be constructed from these facts.

How SeaText's pricing logic works

• Agent-based activation: Each growth workflow (conversion testing, bot refunds, personalization, translation, SEO content) is a separate agent you turn on.
• Free pilot first: A 1-month free trial lets teams measure lift before committing.
• Enterprise controls: Variants require approval, exposure limits, and rollback — features that affect pricing conversations for larger orgs.
• Outcome-linked agents: The Bot Refund Agent, for example, aims to recover up to 20% of Google/Meta spend, framing cost around recovered budget.

If you're evaluating AI copy tools, the relevant next step is to compare SeaText's agent model and trial against the specific Copy.ai plan you're considering.

Theme customizer settings — site title, tagline, menu labels, and other global strings — are saved as options in the WordPress wp_options table. SeaText AI works by translating the final HTML that a visitor's browser receives. If a customizer string never appears in that rendered output, or if the option isn't registered for translation, SeaText has nothing to translate.

How WordPress Customizer Stores Text

The WordPress Customizer API writes values to wp_options under keys like blogname, blogdescription, theme_mods_divi, and custom option names added by themes or plugins. These are not post content, not taxonomy terms, and not part of the page builder's module data. They are raw option values that the theme reads when building the header, footer, or other global areas.

Why SeaText AI Might Miss Customizer Strings

SeaText AI translates the rendered page. It scans the HTML response, finds text nodes, and replaces them with translations. This works for page builder modules, post content, widget text, and any string that ends up in the final HTML. Customizer strings only get translated when:

• The theme outputs the option value directly in a template (e.g., bloginfo('name') in header.php).
• The option is registered with a translation filter that SeaText hooks into.
• The string appears in a menu item, widget, or other element that SeaText already processes.

If your Divi theme uses get_theme_mod() or get_option() to pull a customizer value and echoes it inside a Divi module, SeaText will catch it because it's in the module's output. If the theme prints the value in a hard-coded template file that SeaText doesn't parse, the string stays in the original language.

The Registration Requirement for Translation

WordPress translation functions (__(), _e(), esc_html__()) only work on strings that pass through them. Customizer values are user-entered data, not static strings in code. For a translation plugin to handle them, the option must be registered in a "string translation" table (like WPML's icl_strings or Polylang's pll_strings). SeaText AI does not maintain its own string registry; it relies on the text being present in the rendered HTML. Therefore, the only reliable way to translate customizer text with SeaText is to ensure the text appears in the page output that SeaText processes.

Diagnostic Sequence: Check These Steps

1. View the page source in the target language. Search for the exact customizer string (site title, tagline, menu label). If it's there in the original language, SeaText didn't see it.
2. Identify where the string is output. Is it in a Divi module (Header, Menu, Text)? In a widget? In the theme's header.php or footer.php?
3. If it's in a Divi module — enable SeaText's Divi integration in the plugin settings. SeaText translates Divi module content automatically.
4. If it's in a widget — ensure the widget area is translated. SeaText processes widget text that appears in the rendered sidebar/footer.
5. If it's in a theme template file — you have two options: move the string into a Divi global module or widget, or wrap the output in a translation function and use a string translation plugin alongside SeaText.
6. Clear caches — server cache, CDN, and browser cache can serve untranslated HTML.

Common Scenarios and Fixes

Site Title and Tagline in Divi Header Module

Divi's Theme Builder header often uses a Site Title module. That module pulls blogname and blogdescription from options and renders them as HTML. SeaText translates the module's output, so the title and tagline translate automatically — provided the Divi integration is active.

Custom Menu Labels Set in Customizer

Menus created in Appearance → Menus store labels in wp_posts (menu items). SeaText translates menu items because they render as standard <a> tags. If you set a custom label via the Customizer's "Menu Locations" or a theme-specific menu setting, check whether that label ends up in the menu HTML. If yes, it translates.

Theme-Specific Customizer Options (e.g., Divi's "Phone Number" in Header)

Divi adds customizer fields like header phone number, header email, and social links. These are saved as theme mods. If you enter them in the Customizer and they appear in the Divi Header module or a Divi Global Header, SeaText translates them. If you use a child theme that outputs get_theme_mod('et_header_phone') directly in header.php, SeaText won't see it unless that template is processed by SeaText (it isn't).

Limitations and When This Advice Doesn't Apply

• Hard-coded strings in theme files — SeaText does not scan PHP templates. Move editable text into Divi modules, widgets, or the Customizer fields that Divi modules use.
• JavaScript-injected text — If a customizer value is passed to a JS variable and rendered client-side, SeaText's server-side translation won't catch it. Use a Divi module instead.
• Multilingual plugins (WPML, Polylang, TranslatePress) — If you run another translation plugin, it may handle customizer strings via string registration. SeaText can coexist, but you only need one translation layer for a given string.
• Admin-only strings — Customizer preview, admin bar, and dashboard text are not translated because SeaText only translates front-end visitor-facing HTML.

Key Facts About SeaText AI Translation

Frequently Asked Questions

Does SeaText translate the WordPress site title and tagline set in Settings → General?

Yes, if your theme outputs them via bloginfo('name') or bloginfo('description') in a template that SeaText processes, or if they appear in a Divi Site Title module. They are stored as options but render as HTML text.

Why does my Divi Theme Builder header translate on some pages but not others?

Check whether the header is a global header (Theme Builder → Global Header) or assigned per page. Global headers translate once and apply everywhere. Per-page headers translate per page. Ensure SeaText's Divi integration is enabled and clear caches.

Can I translate customizer strings without using Divi modules?

Not reliably with SeaText alone. SeaText translates rendered HTML. If the string never reaches the browser as text (e.g., used only in JS, or in a template SeaText doesn't parse), it won't translate. Use a string translation plugin for those cases.

Will SeaText translate my Divi Theme Customizer colors and layout settings?

No. SeaText translates text strings only. Colors, spacing, and layout settings are CSS values, not translatable text.

How do I know if a customizer string is registered for translation?

SeaText doesn't use a registration system. View the page source in the target language. If the string appears in the original language, it wasn't in the HTML SeaText processed. Move the string into a Divi module or widget.

Does SeaText work with Divi's Theme Customizer export/import?

Export/import moves theme mods between sites. SeaText translates the live site's output. After import, visit the site in each language to trigger translation.

What if I need different customizer values per language (e.g., different phone numbers)?

SeaText translates the text, not the data. For truly different values per language, use a multilingual plugin that supports per-language theme mods, or create language-specific Divi global headers.

You installed an AI landing page tool. The copy rewrites itself for each keyword. Yet your cost per acquisition has not moved. The most common reason is that the AI is optimizing a signal that does not reflect real buyers: bot clicks inflate traffic, conversion tracking captures the wrong actions, or the keyword set is too small to generate meaningful variants. Fix the input data first, then let the AI work.

How AI landing page optimization actually works

SeaText's Google Ads Landing Page Agent reads the campaign, keyword, and visitor intent behind each paid click. It then rewrites headlines, offers, product blocks, and calls to action so the page continues the exact promise in the ad. One page becomes a keyword-matched landing page for every search term.

The agent runs three connected steps: detect the keyword that triggered the click, swap the headline and key copy to mirror that keyword, and test variants against each other. Winning variants roll out automatically. Conversion reporting breaks down by page, keyword, and variant so you can see which matches actually convert.

This only helps if the traffic reaching the page is real, the conversions you track match business outcomes, and there are enough visits per variant to reach statistical confidence. Without those conditions, the AI optimizes noise.

Common reasons ROI stays flat after AI implementation

• Bot and invalid click traffic. Non-human clicks consume budget and poison retargeting audiences. The AI sees high traffic but low conversion and may rewrite toward patterns that only bots trigger.
• Wrong conversion actions. If you track page views, scroll depth, or micro-conversions instead of qualified leads or revenue, the AI learns to maximize the wrong metric.
• Insufficient keyword coverage. Activating the agent on five keywords when you bid on five hundred means most visitors still see generic copy. The lift only appears where the agent is active.
• Low data volume per variant. A/B testing needs hundreds of conversions per variant to declare a winner. Low-traffic campaigns never reach that threshold.
• Localization gaps. If you target multiple countries but only serve English pages, visitors bounce. The AI cannot fix a language mismatch it does not know exists.
• Offer or product mismatch. No rewrite can sell a product the searcher does not want. If the keyword intent is informational and your page pushes a demo request, conversion stays low.

Diagnostic sequence: check these five areas in order

1. Audit traffic quality. Pull the invalid click report from Google Ads. Compare SeaText's bot detection evidence (session recordings, IP patterns, behavior signals) against Google's data. If bot share exceeds 10 percent, enable the Bot Protection Agent and submit refund requests before judging page performance.
2. Verify conversion tracking. Open your Google Ads conversion actions. Confirm the primary conversion is a revenue-proximate event: form submission with validated email, phone call over 60 seconds, or purchase. Remove or demote soft conversions like "page view" or "button click."
3. Map agent coverage to keyword volume. Export your search terms report. Filter for terms with 50+ clicks in the last 30 days. Check which of those have the Google Ads Landing Page Agent active. Activate on the top 80 percent of click volume first.
4. Check localization alignment. Segment traffic by country and language. If non-English traffic exceeds 15 percent of spend but the Translation Agent is not active for those languages, enable it. SeaText translates into 125 languages and optimizes localized copy for conversion.
5. Review variant statistical power. For each active keyword, check the variant report. If a variant has fewer than 300 visits and 30 conversions, the test is underpowered. Consolidate similar keywords into ad groups so the AI tests fewer, higher-traffic variants.

Key facts about SeaText's Google Ads agents

When the problem is data volume, not page copy

AI optimization is a statistical process. Each variant needs enough conversions to beat the control with confidence. A campaign spending $500 per month on a $100 CPA generates five conversions. Split across three variants, that is fewer than two conversions per variant per month. No AI can declare a winner there.

Two practical fixes: consolidate campaigns so fewer variants run on more traffic, or accept that the AI will run in "exploration mode" longer. In exploration mode, SeaText rotates variants evenly to gather data. You will not see lift until exploitation begins. Check the variant report for the "phase" indicator if available, or manually calculate conversions per variant.

If you cannot increase budget, consider pausing the AI on low-volume campaigns and using static, well-crafted pages instead. The AI adds overhead (variant generation, testing infrastructure) that only pays off at scale.

When the problem is traffic quality: bots and wrong intent

Google's own invalid click detection catches some bots, but not all. Sophisticated bots mimic human behavior: they scroll, click, and even fill forms. SeaText's Bot Protection Agent analyzes session depth, mouse movement patterns, IP reputation, and timing anomalies. It builds a session evidence log you can submit for refunds.

More importantly, the agent filters bots before they hit your analytics and retargeting pixels. Without this, your lookalike audiences train on bot behavior, and your conversion rate denominator inflates artificially. The AI then optimizes for bot-like patterns.

Run this check: compare SeaText's bot session count against Google's invalid click count for the last 30 days. If SeaText finds 2x or more, enable the Bot Protection Agent and submit the refund report. Then re-evaluate ROI after 14 days of clean traffic.

When the problem is localization gaps

If you run Google Ads in Spain, Mexico, and Argentina but your landing page is English-only, Spanish speakers bounce. The AI landing page agent rewrites English copy for English keywords. It does not translate.

SeaText's Translation Agent handles this separately. It detects each visitor's language, translates the page instantly, and keeps new posts, products, and updates translated in the background. You choose the markets; the agent creates localized versions in up to 125 languages without a separate site for each market.

Performance tracking by language and market lets you see ROI per locale. If Spanish ROI is negative while English is positive, the fix is translation plus localized offer adaptation, not more English headline tests.

When the problem is conversion tracking

Google Ads optimizes toward the conversion action you designate as primary. If that action is "contact form submit" but 80 percent of submissions are spam or unqualified, the AI learns to attract spammers. The landing page copy shifts toward language that attracts form fills, not buyers.

Fix the conversion definition first. Use offline conversion import to feed qualified lead or revenue data back to Google. Set the primary conversion to "qualified lead" or "purchase." Demote form submits to secondary. The AI will then rewrite toward the language that brings qualified buyers.

SeaText's conversion reporting by page, keyword, and variant works best when the underlying conversion data is clean. Garbage in, garbage out applies to AI optimization as much as to bidding algorithms.

Limitations and when this advice does not apply

• Brand new accounts. If you have fewer than 100 conversions total, the AI has no baseline. Focus on getting tracking right and running manual tests first.
• Pure brand campaigns. Branded search already converts high. AI rewrite adds little; the visitor knows your name. Save the agent for non-brand, competitive keywords.
• Single-product, single-keyword funnels. If you bid on one exact-match keyword and sell one product, a static page tuned by hand beats an AI that needs variants to test.
• Regulated industries with copy restrictions. If legal must approve every headline, the AI's automatic rewrites create compliance risk. Use the variant editor for manual approval workflows instead.
• Traffic from non-Google sources. The Google Ads Landing Page Agent reads Google click data (gclid, keyword). It does not optimize for Meta, TikTok, or organic traffic. Use the Visitor Source Rewrite Agent for those channels.

Terminology quick reference

• Variant: A version of the page with specific headline, offer, or CTA changes generated by the AI.
• Exploration vs. exploitation: Exploration rotates variants evenly to gather data. Exploitation shows the winning variant more often.
• gclid: Google Click Identifier. The parameter that lets the agent know which keyword triggered the click.
• Invalid click: Google's term for clicks it deems non-human or fraudulent. Refunds are automatic but incomplete.
• Offline conversion import: Sending CRM-qualified lead or sale data back to Google Ads to improve bidding and AI optimization.

FAQ

How long before I see ROI improvement after fixing the diagnostic issues?

Traffic quality fixes (bot filtering) show immediate cost reduction. Conversion tracking fixes take 7-14 days for Google's bidding to relearn. Keyword coverage expansion shows lift as each new keyword accumulates 300+ visits. Full cycle: 30-60 days for a typical account.

Can I run the AI landing page agent on only my top 10 keywords?

Yes. SeaText lets you choose the page and activate with a small keyword set. Start with the 10-20 keywords driving 80 percent of non-brand spend. Expand once you see variant winners.

Does the AI rewrite the entire page or just headlines?

It rewrites headlines, key copy, offers, product blocks, and CTAs. The page structure, design, and non-text elements stay the same. You can lock specific sections (legal disclaimers, pricing tables) so the AI never touches them.

What if my conversion cycle is 90 days? Can the AI still optimize?

The AI optimizes for the conversion event you track. If that event happens at day 90, the variant test needs 90 days per cohort. Use a leading indicator (qualified demo booked, trial started) as the primary conversion for faster feedback. Import the 90-day revenue as offline conversion for bidding.

Will the AI create duplicate content issues for SEO?

No. The rewrites happen client-side via JavaScript after the page loads. Googlebot sees the original page. The variants are not indexable URLs. This is a paid-traffic optimization, not an SEO strategy.

How does this differ from Google's own responsive search ads or dynamic search ads?

RSA and DSA optimize ad copy and targeting. SeaText optimizes the post-click experience. They work together: RSA gets the click, SeaText makes the landing page match the promise. The agent also reads the actual search term, not just the keyword match type.

What happens if I pause the agent? Do I lose the winning variants?

Winning variants are saved in the variant editor. You can publish the best performer as the static page or keep the agent running. Pausing stops new variant generation and testing but preserves history.

import { defineAsyncComponent, onMounted } from 'vue';

const AsyncHello = defineAsyncComponent(() => import('./Hello.vue'));

export default {
  components: { AsyncHello },
  template: `
    <div v-sea-text>
      <AsyncHello />
    </div>
  `
};

import { ref, onMounted } from 'vue';

export default {
  setup() {
    const current = ref('ComponentA');
    onMounted(() => {
      // After the component switches, trigger a refresh.
      SeaText.refresh();
    });
    return { current };
  },
  template: `
    <div v-sea-text>
      <component :is="current" />
    </div>
  `
};

import { onMounted } from 'vue';

export default {
  name: 'DynamicCard',
  setup() {
    onMounted(() => {
      // Ensure SeaText sees the newly rendered markup.
      SeaText.refresh();
    });
  },
  template: `
    <div>
      <h2>{{ $t('welcome') }}</h2>
      <p>{{ $t('description') }}</p>
    </div>
  `
};