If you installed a disclosure widget, what have you proved?
There is a category of product that installs a script on your site and adds a compliant AI-disclosure notice to your chatbot, often logging each time it is shown. If your own chatbot needs a disclosure and does not have one, that is a direct and sensible purchase. We are not the better fix, and this page is not an argument that you should not use one.
The one thing a fix vendor cannot supply
A disclosure widget and its logs come from the same party whose job it is to make the disclosure appear. When the question is whether the disclosure was genuinely visible to a first-time visitor, on mobile, behind a cookie wall, on a given date, the vendor's own records are the vendor's own records. That is not an accusation of dishonesty. It is the same reason an audit is not performed by the finance team.
The industry term for this is grading your own homework, and it is fine right up until someone with authority asks the question. At that point the useful document is one produced by a party with no stake in the answer.
What actually goes wrong with an installed widget
Widgets are configuration, and configuration drifts. The failure modes we see are mundane:
- The script is present but a consent wall renders above it, so the notice never reaches a first-time EU visitor.
- It is installed on the main site and missing from the subdomain where support actually runs.
- A theme or CSS change hides it in the compact mobile layout.
- A tag-manager change stops it loading on some routes, and nobody notices because nobody is watching from outside.
- It covers the chatbot, while Article 50(2) media marking and 50(4) labels on the same site are untouched by it.
The scope difference
Article 50 has four duties. Disclosure widgets address the first one: telling a user they are interacting with an AI system. Machine-readable marking of AI-generated media, notice for emotion recognition and biometric categorisation, and labelling of AI-generated text and deepfakes are separate duties with separate evidence, and a chatbot banner does nothing for any of them.
There is also a scope difference in the other direction. A widget only works on a site you control and can instrument. If you are counsel reviewing a client's site, an agency auditing a portfolio, or a buyer doing diligence on a vendor, you cannot install anything. You need something that works on a URL.
The two together, honestly
The sequence that makes sense: use whatever fix you like, then verify it from outside and keep watching it. Every fix, ours or anyone's, is a claim until an independent record says it rendered. And a fix that was working in July says nothing about November unless something re-checks.
That is the whole relationship. Fix vendors make the disclosure appear. We record whether it did, and tell you the day it stops.
Side by side
| Disclosure widget | DisclosureProof | |
|---|---|---|
| What it is | A script that adds the notice | An outside check that records what showed |
| Whose record | The vendor supplying the fix | A third party with no stake in the answer |
| Art. 50 duties covered | Chatbot disclosure | All four, graded separately |
| Works on sites you do not control | No: it must be installed | Yes: any public URL |
| Sees what a first-time visitor sees | Renders it, does not observe it | Opens the page like a visitor and captures it |
| Evidence a third party can check | Vendor-held logs | Hashed and sealed; verify the hashes yourself |