
Hidden AI Instructions Watch
Evidence-only log of hidden AI-targeted page text, plus a free scanner.
Free1 impressionslaunches Oct 281 comment
Logs public pages with AI-targeted hidden text and provides a free scanner to detect it.
- for
- Developers and AI agents needing to verify hidden instructions on web pages
- pricing
- free
- license
- MIT
Key features
- Evidence-only log — Records what hidden text says, where it is, and how it’s hidden without judging intent.
- Free scanner — Apify Actor scans pages for hidden AI-directed text without following or submitting anything.
- Multiple hiding techniques — Detects CSS tricks, hidden attributes, invisible Unicode, and more.
- Category tagging — Classifies findings as policy notice, manipulation, or harmful with severity levels.
- Render-mode scanning — Scans each URL twice: plain HTTP and headless Chromium with JavaScript.
- SHA-256 summary — Provides a SHA-256 hash of the scanned HTML for each page.
Use cases
- Check a website before an LLM reads it to ensure no hidden manipulation instructions
- Audit public pages for policy notices addressed to AI crawlers
- Detect hidden credential-stealing prompts that could be exploited by agents
Hidden AI Instructions Watch FAQ
How does the scanner work?+
It loads each URL in headless Chromium (optional JavaScript rendering) and checks computed styles of every element for hidden text.
Does the scanner interact with the page?+
No, it only reads pages and never follows links, clicks, or submits forms.
What types of hidden text are detected?+
CSS hiding, hidden attributes, invisible Unicode, comments, meta tags, templates, and other techniques listed in the method section.
Can it scan signed-in or private pages?+
No, it only scans public, signed-out pages; signed-in content is out of scope.
Is there any cost for using the scanner?+
The scanner is free with no per-result charge.
Summarized by DevHunt from github.com · Nov 10, 2026. Details may change; check the official site.


