AI Visibility
Dossiers are published in the forms AI assistants read and the tools they call. Then we monitor what assistants actually say about a company, where they are wrong, and what evidence would correct them.
The machine twin
Every public dossier has JSON-LD (schema.org Organization with sameAs to registries, SEC, IP offices), a Markdown version with the same claims and sources, an entry in llms.txt and llms-full.txt, stable URLs, document and retrieval dates. Crawlers used by assistants are allowed in robots.txt; the sitemap and IndexNow are kept current.
The tool path
The OpenDossier MCP server and OpenAPI let Claude, ChatGPT, Gemini, Perplexity and in-house agents query live dossiers. An answer from a tool call carries the source; that is more reliable than hoping a crawl is fresh.
Monitoring
A fixed question set, run monthly against the major assistants: who the company is, what it does, whether it is trustworthy, what disputes it has. We record the answers, the citations, the errors and the share of voice, and show them to the company that has claimed its profile.
The correction loop
An error is corrected by publishing better evidence to the dossier and its machine twin, and re-running the questions. Nothing is hidden; nothing is seeded under another name; no reviews are generated. Assistant providers' policies on manipulation are tightening and honest structured evidence is the durable strategy.
Questions
Can you make an assistant say something specific?
No. We can make accurate, sourced evidence available in every form assistants use, and measure whether it is picked up.
Is this included somewhere?
In Trust Workspace Plus; also sold alone from US$150 a month per company.
Start
One company, one brief, or claim your own profile.