The trust layer
Anyone can put a confident chatbot on a website. The question is who is behind it.
The hardest problem in commercial AI is not fluency, it is accountability. A model that sounds certain and is wrong costs more than one that says nothing, because someone acts on it.
In a network where agents talk to agents, that problem compounds: the receiving side cannot see a website, judge a logo or read a room. It has whatever the other party asserts. Without a verifiable identity underneath, the whole exchange is confident strangers quoting each other.
So verification is not a badge here, it is the load-bearing part. A seal that can be obtained without proof is worse than no seal, because it launders a claim into something that looks checked.
What this looks like in practice
A DNS TXT record, a meta tag, or a file at a well-known path. Whoever controls DNS often cannot edit the CMS and the reverse, so all three are first-class. The seal only appears once the token is actually found.
The twin answers from indexed material and names where each claim came from. Where it has nothing, it is instructed to say so and offer a person rather than fill the gap.
Nothing the twin says is binding unless you explicitly enable binding statements. While off, it gives indicative guidance and defers to your team for written confirmation — the liability guard sits on by default, not off.
Every answer given to the outside world is timestamped, source-attributed and hash-anchored. If a customer says 'your AI told me X', you can check whether it did.
What is true today
This section exists because most pages in this category do not have one. Both columns are maintained against the product's own status list, not written by marketing.
- Domain verification via DNS, meta tag or file
- Source-cited answers with an honest 'I don't know'
- Commitment levels — binding statements require explicit opt-in
- Statement register, exportable as JSON
- Swiss and EU data protection posture with a DPA and named sub-processors
- Commercial-register (Zefix) numbers are recorded but not yet checked automatically against the registry.
- There is no cryptographic company identity or blockchain-anchored ID. Verification proves domain control, which is a narrower and honest claim.
- Sanctions and credit screening are not part of the product.
Check it against your own website
The free Twin-Check reads your site the way an AI assistant would and tells you what it can and cannot answer about you. No account needed.
Questions people actually ask
That whoever set up the twin controls the company's domain. That is a real, checkable fact. It is deliberately not presented as proof of the company's legal existence or solvency — those are different checks we do not perform.
The statement register lets you find exactly what was said, when, and on which sources. Binding commitments are off unless you enabled them, so an incorrect answer is a correction, not a contract.
Reality Twin is a Swiss company, developed in Hinwil near Zurich and governed by Swiss law. The data itself is hosted in the European Union (Ireland), covered by the GDPR and the Swiss FADP, with a Data Processing Addendum available and every sub-processor listed publicly.