Your Banner Speaks Six Languages. Your Consent Records Speak None.

The banner was translated with care. Six languages, reviewed by counsel in two of them, switched automatically by browser language with a manual selector for anyone the detection gets wrong. Then someone looked at the consent export and found a timestamp, a purpose list, and no record at all of which language any of it was agreed in.

This gap is common because translation is treated as a presentation task and consent logging as a separate engineering task, and nobody owns the join between them. The join is where the legal value sits. Consent that cannot be tied to the words shown is consent you will struggle to demonstrate.

Informed means the text they actually read

Consent under GDPR must be informed, and the information is the banner and preference centre text in the form the visitor encountered it. If a translation compresses a purpose description, softens a warning, or mistranslates a vendor category, the English master text does not cure it. The record needs the language code and the text version, so the exact wording can be retrieved for any historical choice.

Versioning translations separately matters for the same reason. Master text updates often ship before translations are reviewed, leaving a window where one language describes the old processing and five describe the new. Without per language version records, you cannot even establish who was shown what during that window, let alone defend it.

Detection errors are consent relevant facts

Automatic language detection gets it wrong for bilingual regions, expatriates, travellers and anyone whose browser language does not match their reading language. Whether the visitor changed the language manually is a meaningful signal: it shows the presented language was not the one they used to decide.

Log the presented language, whether it was changed, and the final language in which the choice was confirmed. That sequence takes three fields and answers questions that otherwise turn into speculation during a complaint. A complainant who says they did not understand the banner is in a very different position if your record shows they switched it to their language and spent forty seconds in the preference centre.

Preference centres drift further than banners

Banners get translated because they are visible. Preference centres, purpose descriptions, vendor lists and policy pages linked from them drift: deeper layers remain in the master language, or lag a version behind, because they are longer, change more often, and are owned by nobody in particular. A visitor who opens the detailed layer, the most diligent visitors you have, meets the least maintained text.

Audit the full path in each language, not just the first layer. Click every control as a visitor who reads only that language would. Any dead end, untranslated purpose description, or English legal page in the middle of the journey is a finding, and it is usually an inexpensive one to fix once someone has actually walked the path.

Make translation part of the release, not an afterthought

Consent text changes should trigger translation and legal review in the same release process, with the banner held, where risk warrants, until the languages you serve are consistent. That slows some releases by days. It prevents the alternative, which is discovering during an inquiry that for six weeks your French banner described processing your English one had already changed.

Keep a change log per language next to the consent records themselves. When a visitor asks what they agreed to, the answer should be retrievable in minutes, in their language, with the version and date attached. If that retrieval takes a project, the record keeping was designed for storage, not for demonstration, and demonstration is the point.

Frequently asked questions

Should consent records include the banner language?

Yes. The record should include the language in which the banner and preference options were presented, and the text version in that language. Consent must be informed, and proving that requires showing the words the visitor actually read, not just the master language version.

What if a translation differs from the master consent text?

The text the visitor read is what informs their choice, so a translation that omits or softens material information is a real compliance problem, not a cosmetic one. Translations of consent text need review to the same standard as the original, and version records per language so discrepancies can be identified by date.

Does automatic language detection affect consent validity?

It can, if detection presents a language the visitor does not read well and no clear selector is offered. Providing an obvious language switcher, and logging the presented language and any change the visitor made, both improve the quality of consent and create evidence that the visitor decided in a language they chose.

Why do preference centres lag behind banner translations?

They are longer, change more frequently, and are often owned by a different team or vendor workflow than the banner. Without an explicit audit of the full consent path in each language, deeper layers stay untranslated or outdated while the visible banner looks complete.