Option to remove languages
-
Hi,
Could you make an improvement to the widget? I only need it to offer one language — please make it possible to remove the other language options.
I had Claude review one of our product pages from an AI-readability perspective, and it flagged that the widget is injecting a large block of text into the page – around 40 language names and 20+ accessibility toggle labels – that gets read as plain content alongside the actual product description. Here’s what it found:
When I fetched your page, the extracted text didn’t just contain your product content — it also pulled in every language option and accessibility control from the “OneTap” accessibility widget in your footer/toolbar. Concretely, this is what showed up in the raw text stream:
English Deutsch Español Français Italiano Polski Svenska Suomi
Português Română Slovenščina Slovenčina Nederlands Dansk Ελληνικά
Čeština Magyar Lietuvių Latviešu Eesti Hrvatski Gaeilge Български
Norsk Türkçe Bahasa Indonesia Português (Brasil) 日本語 한국어 简体中文
العربية Русский हिन्दी Українська Srpski English (UK) ایران ישראל
Македонија ประเทศไทย Việt NamVision Impaired Mode / Seizure Safe Profile / ADHD Friendly Mode /
Blindness Mode / Epilepsy Safe Mode / Font Size / Line Height /
Readable Font / Cursor / Letter Spacing / Align Text / Font Weight /
Light Contrast / High Contrast / Monochrome / Reading Line /
Reading Mask / Hide Images / Highlight Content / Stop Animations /
Highlight Links…That’s around 40 language names and 20+ toggle labels, all rendered as plain visible text in the page’s HTML (not hidden in a dropdown that only opens on click, at least not from a text-extraction standpoint) — sitting in the same DOM flow as your actual product description.
Why this matters for an AI agent specifically:
Context window cost — if an agent fetches your full page (rather than just structured data or a curated excerpt), this block eats space that could otherwise hold more of your actual content or other pages in a multi-page research task.
Signal dilution — an agent summarizing “what is this page about” has to wade through 60+ irrelevant proper-noun-like tokens (language names, mode names) sitting between sections of real content. It’s not likely to misfire and think “Srpski” is a product feature, but it does add noise the model has to filter, which measurably lowers extraction quality on some models/pipelines.
It repeats site-wide — this isn’t specific to this product page; it’s injected on every page via the accessibility widget, so it compounds across a multi-page crawl.
You must be logged in to reply to this topic.