How does Weglot work?

Weglot intercepts web pages via DNS routing or JavaScript, detects text strings in the DOM, translates them through machine translation engines, and serves the localized pages from its cloud cache. It is fast to launch, but its proxy architecture creates recurring monthly word charges, database lock-in, and translation drift across large catalogs.

For marketing teams seeking quick international reach, Weglot delivers a working multilingual site in an afternoon. But understanding how it functions technically reveals the structural trade-offs between dynamic proxy translation and static, file-based localization.

The mechanics of Weglot: two integration methods

Weglot does not modify your core database or duplicate your CMS pages. Instead, it operates as an edge translation layer between your server and your international visitors.

Depending on your CMS platform, Weglot runs in one of two configurations:

1. The DNS reverse proxy architecture

For platforms like WordPress, Shopify, Webflow, and custom stacks, Weglot’s recommended architecture connects directly at the DNS level.

• Request interception: When an international visitor requests yoursite.com/es/ or es.yoursite.com, DNS records route the HTTP request through Weglot’s edge proxy servers.
• Source page retrieval: Weglot fetches the original source page in English from your web host.
• HTML string parsing: Weglot’s engine parses the HTML tree, extracting visible text, image alt tags, meta titles, Open Graph descriptions, and form placeholders while skipping code scripts and structural markup.
• Hash matching and translation: Each text string is hashed. Weglot checks its cloud database for existing translations. If a hash is new, Weglot sends the string to third-party machine translation APIs (primarily DeepL, Google Translate, or Microsoft Translator).
• DOM reconstruction: The translated strings replace the English content inside the HTML tree, and Weglot injects localized hreflang tags before delivering the fully rendered page to the user’s browser.

Because this happens on the server side, search engine crawlers receive pre-rendered HTML in the target language, enabling proper indexing on localized search engines.

2. The JavaScript snippet architecture

On platforms without direct DNS configuration access, Weglot injects a client-side JavaScript library into the page header.

In this mode, the visitor’s browser downloads the original English page first. The Weglot script executes on the client side, detects the visitor’s browser language, queries the Weglot API, and dynamically swaps English text nodes for translated strings inside the DOM.

While easy to configure, this client-side approach often triggers a visible flash of untranslated content (FUTC) and is less reliable for search engine indexing, as search bots must execute complex JavaScript to read translated copy.

What Weglot handles well

Weglot has become a market standard for early-stage localization because it eliminates engineering hurdles:

• Zero development overhead: You do not need to architect internationalized routing, build language switchers, or redesign database schemas for multilingual fields.
• Automatic content detection: When you publish a new blog post or update a product description in your primary language, Weglot detects the new text nodes automatically on the next page view.
• Integrated visual editor: Translators can click on text blocks directly inside a live visual preview of the website, adjusting text lengths that break CSS layouts.
• Built-in technical SEO: Weglot automatically adds hreflang link attributes, translates URL slugs, and manages subdirectories or subdomains without manual server configuration.

Where the proxy model breaks down at scale

Despite its convenience, running your entire international web presence through an external translation proxy introduces significant operational and financial bottlenecks once your site grows.

1. Compounding subscription costs based on stored words

Weglot does not charge a one-time translation fee. Instead, it charges a recurring monthly SaaS subscription based on the total number of translated words stored on its servers.

If your site contains 25,000 English words across product pages, blog posts, and support documentation, translating into four languages generates 100,000 translated words. That pushes you into higher subscription tiers immediately. If you expand your catalog or add two more languages, your monthly bill jumps permanently, even if those pages receive minimal traffic. You pay every single month to keep existing words alive.

2. Total vendor lock-in

Because Weglot stores your translated content in its external cloud database rather than your CMS repository, your translations do not belong to your codebase.

If you downgrade or cancel your subscription, the proxy shuts down. Your translated subdirectories return 404 errors, your search engine rankings collapse, and you lose all historical translation data unless you have previously performed manual database exports.

3. Context blind spots and phrase-by-phrase drift

Weglot parses web pages into isolated text segments. When it hands those segments to DeepL or Google Translate, the machine translation engine often lacks catalog-wide context.

A word like “Order” might be translated as a verb on one button and as a noun on the adjacent navigation link. Similarly, brand names, product handles, and technical terminology frequently get translated literally because generic machine engines lack customizable workspace instructions. While Weglot offers basic string replacement rules, it cannot apply contextual instructions such as “use formal register throughout” or “adapt measurements to metric units.”

The hybrid model: combining Weglot with AI Glot

For teams that want Weglot’s edge delivery without sacrificing translation quality or paying runaway translation costs, a decoupled workflow provides the solution.

Instead of accepting raw, automated machine output from Weglot’s default APIs, you use Weglot strictly for edge routing while using AI Glot as the intelligent editorial engine.

The three-step round trip

1. Export the raw strings: Export your translation catalogue from Weglot as a CSV or XLIFF file.
2. Refine with prompt-controlled AI: Upload the file to AI Glot. With a single plain-English prompt, specify your brand tone, enforce your workspace glossary, and protect structural IDs and pricing variables. You review a deterministic plan with exact word counts and credit costs before spending anything.
3. Re-import the clean data: Import the refined CSV back into Weglot. Weglot continues serving the localized pages at the edge, but your visitors receive polished, context-accurate prose that matches your brand guidelines.

Comparing localization architectures

The verdict on Weglot

Weglot is an exceptional tool for validating international demand quickly without touching code. It handles DNS routing, page serving, and hreflang headers reliably.

However, treating Weglot as your sole translation engine leads to runaway monthly costs and generic copy. The most sustainable architecture uses Weglot for what it does best, hosting and routing, while decoupling your translation data into structured files that can be processed, verified, and improved with purpose-built AI tools like AI Glot.