How the checker works — Methodology
Transparency about what this tool does, and what it doesn't · June 2026
1. What this tool is — and isn't
This is a heuristic transparency self-check tool. It fetches the public HTML of the URL you submit and looks for signals that key EU AI Act Article 50 transparency requirements are present. The output is a score between 0 and 100 and a list of specific gaps.
2. The three checks and their legal basis
| Check | What we detect | Legal basis |
|---|---|---|
| Chatbot disclosure | Detectable AI chat widget on the page | Art. 50(1)Providers and operators of AI systems that interact with natural persons must clearly inform them they are communicating with an AI, in a timely manner |
| AI content labelling | Labels, meta tags, or HTML attributes indicating AI-generated text or images | Art. 50(2) / (4)Providers of AI systems that generate synthetic content must mark it as artificially generated; AI-generated text published to inform the public must be labelled |
| Transparency page | A publicly reachable page on your domain describing your AI use (e.g. /ai-transparency) | RecommendedNo single article of the EU AI Act mandates a dedicated page at this exact path — but transparency documentation is encouraged by regulators as evidence of compliance intent |
3. How detection works
The scanner performs the following steps against the static HTML of your page:
Step 1 — Fetch
Your page is retrieved via HTTP GET from our Cloudflare Worker. No cookies or visitor session state from your site are used.
Step 2 — Chatbot detection
The scanner inspects <script src="..."> attributes for known AI chatbot CDN domains — including intercom.io, tidiochat.com,app.crisp.chat, js.driftt.com, livechatinc.com,chatra.io, smartsupp.com, and 10+ more. It also checks inline scripts for JavaScript initialisation patterns such aswindow.intercomSettings, $crisp.push, andtidioChatApi.
Step 3 — AI content detection
The scanner looks for <meta name="generator"> values referencing AI authoring tools, HTML elements with data-ai-generated attributes,aria-label="AI-generated", and similar markup patterns.
Step 4 — Transparency page detection
The scanner checks for <a href="..."> links pointing to paths such as/ai-transparency, /ai-disclosure, /ai-policy, and localised equivalents in multiple languages.
Browser rendering fallback
The default scan runs on static HTML without executing JavaScript. When static analysis is inconclusive, the scanner may activate a headless browser (Puppeteer via Cloudflare Browser Rendering) to render the page and re-run detection on the live DOM. This is a best-effort fallback, not a guarantee of full JavaScript coverage.
4. Limitations — false positives and false negatives
Things we may miss (false negatives)
- A chatbot built in-house or self-hosted without a recognisable CDN script URL
- AI chatbots loaded entirely by dynamic JavaScript with no static
<script>tag - AI content labelled using custom markup the scanner has not seen before
- Transparency pages at non-standard paths
Things we may flag incorrectly (false positives)
- A third-party script on a chatbot CDN that is not itself an AI system
- A
<meta name="generator">tag from a non-AI CMS or tool
5. The 2 August 2026 deadline
EU AI Act Article 50 transparency obligations enter into force on2 August 2026. From that date, providers and operators of AI systems covered by Art. 50 must meet the disclosure requirements described above, subject to any implementing guidance issued by the European AI Office.
This checker is a free tool to help you identify likely gaps before the deadline. It does not replace a formal compliance programme.
← Back to the checker