Anyone who has tried to translate HTML knows the sinking feeling when the result comes back: a bold word is now bold in the wrong place, a link wraps half a sentence, or a variable such as {name} has vanished. Markup is the hidden half of every web page, help article and software string, and ordinary translation tools treat it as noise. Wordbeam, an AI translation management system, treats it as part of the text and places the tags for you.
This article explains why tags break, how AI tag placement works, and what a technology team should check before translating formatted content.
Why tags break during translation
A sentence like "Click <b>Save</b> to keep your <a>settings</a>" contains inline tags around specific words. When the sentence is translated, word order changes, and the tags must move with their words. In German the verb may jump to the end, in Japanese the whole structure flips. A translator working by hand has to drag each tag into position, one by one, and every manual step is a chance for error.
The cost of broken markup is rarely visible in the translation itself, but it shows up as a broken page after publishing. Missing closing tags, doubled formatting and dead links are typical results.
How Wordbeam handles AI auto tag placement
With AI auto tag placement, inline tags, placeholders and variables are positioned automatically, so formatting stays intact and you never have to juggle tags by hand. The platform describes four capabilities:
- Automatic tag insertion. AI reads the source markup and places matching inline tags in the target for you.
- Placeholder and variable safety. Variables like {name} and %s are protected, so they cannot be edited or dropped by mistake.
- Formatting preserved. Bold, links and structure survive a round trip through DOCX, IDML, XLIFF and more.
- Built in tag QA. Missing, extra or reordered tags are flagged before export.
Tags in the formats developers actually use
Different formats carry markup in different ways, and a translation system has to understand all of them. The XML based XLIFF format was designed to carry translatable text and its inline codes between tools, which is why it appears in so many localization pipelines.
| Format | Typical markup | Risk when translated by hand |
|---|---|---|
| HTML pages | Bold, links, spans | Misplaced or unclosed tags |
| XLIFF files | Inline codes | Dropped or reordered codes |
| DOCX documents | Styles, hyperlinks | Formatting lost on export |
| IDML layouts | Character styles | Layout shifts |
| App strings | Variables such as %s | Crashes or blank text |
Why this matters for tech content
Technology publishers and software teams produce text full of code terms, product names and interface references. A news article may link to a spec, a guide may bold every button name, and release notes may include inline code. Losing any of that makes the content harder to trust. If your files are documents rather than web pages, this guide to document translation software explains what to look for when layout and formatting must survive.
A safe workflow for formatted content
- Upload the file or connect the source, for example through a native connector for GitHub, GitLab, Contentful or AEM.
- Let the system extract text with its inline tags and variables.
- Pre-translate with an engine such as GPT, Claude, Grok or DeepL and let AI place the tags.
- Review the result in the editor, checking tag flags first.
- Export, and confirm that the file looks like the original.
The tag check before export is the safety net. Wordbeam flags missing, extra or reordered tags, so you find problems in the editor instead of on the live page.
Teams that publish often should also keep a short list of problem files. If one template keeps producing tag warnings, the fix usually belongs in the source markup, such as removing nested formatting, and not in every translation.
Checks every team should run
- Open a sample page in a browser and test every link and button.
- Search the output for stray angle brackets or visible code.
- Confirm that variables appear unchanged in every language.
- Compare the length of key labels with the available space.
Pricing and free trial
You can test AI tag placement on your own files in a 14 day free trial that includes 10,000 words. After that, the Freelancer plan costs $25 per month for 100k words, and the Studio plan costs $95 per month for 300k words.
Frequently asked questions
Can AI really place tags correctly?
It reads the source markup and positions matching tags in the target, and built in tag checks flag anything missing, extra or reordered so a person can fix it before export.
What happens to variables like {name}?
They are protected, so they cannot be edited or dropped by mistake.
Do I need to know HTML to translate a formatted file?
Not in depth. The tool shows tags as small markers and warns you when something is wrong.
Good translation keeps the words and the structure together. Try one real page or document in the free trial and compare the exported file with the original.