Article 50 findings
Target: meridian-retail.example (a fictional company; this is a worked example, not a scan of a real business)
This is the exact structure and wording a real report uses, on a fictional company built to show a realistic, honest mix of results, and not every finding is a PASS. Run your own free scan to see your own site's report instead.
2 of 6 checks need attention · Article 50 in force since 2 Aug 2026 · 1 passed · 3 not assessed on this page.
Summary of findings
Every check on this scan, grouped by the Article 50 duty it observes.
| Article 50 duty | Checks | Result | Status |
|---|---|---|---|
Art. 50(1) Disclosure that a visitor is interacting with AI |
2 | 1 Attention1 N/A | Attention |
Art. 50(2) Machine-readable marking of synthetic media |
1 | 1 Attention | Attention |
Art. 50(3) Emotion recognition and biometric categorisation |
1 | 1 N/A | N/A |
Art. 50(4) Labelling of AI-generated text and deep fakes |
1 | 1 N/A | N/A |
Art. 50 Supporting checks |
1 | 1 Pass | Pass |
Turn a scan into a sealed document
Not the web report saved as a PDF: buying it seals the scan — the findings and captured evidence — into one timestamped, hash-verified document ready to hand to counsel, and includes a full-site crawl of up to 100 pages.
- All 9 checks unlocked on the scan instantly — root causes, fix steps, affected pages & assets
- One full-site crawl included — up to 100 pages, graded and sealed as its own report
- Seal your own Art. 50(3)/(4) statements alongside what we observed
- Records kept 365 days · evidence hashes re-checkable by anyone at /verify/
The pack is offered right on your finished report — scan first, buy only if the record is worth keeping.
Findings in detail
Each finding is a technical observation with the rule it maps to. States are detected / not detected / could not verify, not legal conclusions.
Disclosure that a visitor is interacting with AI
Chatbot AI disclosure at first interaction
No AI/automation disclosure was detected in the chat widget's first interaction (vendor: intercom). If this assistant is AI-driven, Article 50(1) requires disclosing that to users.
Chatbot disclosure persists on mobile
Mobile disclosure was not separately assessed (no desktop disclosure to compare, or the mobile panel could not be read).
Machine-readable marking of synthetic media
Machine-readable provenance marking on media
AI-generation indicators were found on 1 of 2 sampled assets, but 1 of those assets lack machine-readable provenance marking (C2PA/IPTC). Under Article 50(2), AI-generated content must carry such marking where technically feasible.
Emotion recognition and biometric categorisation
Emotion recognition / biometric categorisation disclosure
Article 50(3) (emotion recognition and biometric categorisation systems) cannot be verified by an external crawler. Complete the in-product attestation module to cover this duty; attested answers are sealed as evidence like any other artifact.
Labelling of AI-generated text and deep fakes
AI-content label on article-like pages
This page was not detected as article-like, so the Art. 50(4) AI-text-labelling duty was not assessed here.
Supporting checks
AI policy / disclosure page
A dedicated AI policy/disclosure page was found (https://meridian-retail.example/ai-policy). This is supporting context that helps demonstrate transparency.
3 more checks already graded on this scan
Art. 50(1)Chatbot disclosure covers the site's languages
Art. 50(2) (supporting)Provenance marking is structural, not textual-only
Art. 50(4) (supporting)Official EU labelling icons in use
Included with Pro, Business, and the Evidence Pack. Unlocking reveals the stored results for this scan instantly, no re-scan needed.
What unlocking adds: same scan, no re-scan
Every scan already grades 9 checks and stores the full evidence behind them. Pro, Business, and the Evidence Pack reveal the layers below on the report you already have.
Changes since the previous scan of this target (2026-07-13)
Chatbot AI disclosure at first interaction
The same finding as above, now with the root cause of the exact grade path this scan took, and how it moved since your last scan.
Chatbot disclosure covers the site's languages · ADVANCED CHECK
The site serves language(s) the chat disclosure's wording does not cover (de). A disclosure a visitor cannot read does not reliably inform them (Art. 50(1)).
AI-content label on article-like pages
On multi-page scans, the counts open into the concrete pages and assets behind them:
Captured evidence
What our browser saw at capture time. Every file is SHA-256-hashed into the scan's evidence manifest. The thumbnails below are illustrative, shown in place of real screenshots.
Turn this snapshot into standing proof
This sample shows the free diagnosis plus a preview of the unlocked layers above. The paid tiers keep you covered when it matters, and they are live now: the Evidence Pack is a one-time €79, Pro is €20 a month, and both Pro and Business start with a 7-day free trial.
Download the specimen Evidence Pack (PDF) — the actual sealed document a purchase produces, watermarked, so you can read it before deciding.
The deeper read: 9 checks, root causes, change tracking
3 advanced checks, the root cause behind every grade, affected pages & assets, and what changed since your last scan, revealed on scans you've already run. See plans →
Evidence Pack (PDF)
A one-time purchase for one scan: all 9 checks unlocked, sealed as one timestamped, hash-verified PDF, one 100-page crawl included. Details above ↑
Monitoring & drift alerts
Scheduled re-scans that email you before/after evidence the moment a disclosure disappears. See plans →
Full-site crawl & agency white-label
Beyond the homepage: sitemap-driven crawls (Evidence Pack & plans) and white-label reporting on Business. See plans →
Sample report · EU AI Act Article 50 rule-pack eu-ai-act-art50 2026.07.2 · example target meridian-retail.example · DisclosureProof provides informational scanning and evidence-archiving tools, not legal advice or any guarantee of compliance.