AI is talking about your company. And you answer for it.
We check what ChatGPT, Claude, Gemini and Perplexity say about your company, compare it to your facts, and show exactly where they get it wrong. You get a clear report with evidence — fit for a regulator.
Free audit · no integration · just your domain
Are deposits at this exchange insured?
“Yes, deposits are protected and insured up to €100,000.”
Per your source: Funds are not insured — they are held in segregation.
Illustrative example. Real findings vary by company.
How FactGuard works
Four steps, from your website to defensible evidence.
Map your facts
We read your public pages — fees, terms, licensing — and turn them into one verified source of truth: your version of the facts.
Ask the AIs
Real customer questions are put to multiple AI assistants with live web search, just like a prospect would.
Compare & judge
Each answer is checked against your facts and labelled correct, stale, wrong or critical — with a confidence score.
Keep the evidence
Everything is saved as a timestamped, exportable audit you can act on and show to auditors.
What you get
Start with an audit. Add more when you're ready.
You get a one-time audit and a clear report right away. Add ongoing monitoring and a clean facts layer as you need them.
Core service
AI audit + report
We put real customer questions to four models (ChatGPT, Claude, Gemini, Perplexity) with live web search, compare their answers to your facts, and you get a clear report.
- A verdict, evidence and sources per question
- Across 4 AI models
- Compliance export
Watches it for you
Ongoing monitoring
The audit runs automatically on a schedule and alerts you the moment AI starts getting you wrong.
- Weekly or daily, on a schedule
- Alerts when it worsens
- Trend over time
Raises the odds
Clean facts layer
We prepare your facts in the form AI reads most easily — structured data (Schema.org), a facts page, Wikidata-ready data and llms.txt. The clearer the source, the higher the odds of a correct answer.
- Structured data (Schema.org)
- Wikidata-ready data and llms.txt
- One clear source of truth
Sample report
This is what you get
Every audit produces a clear, timestamped report — a verdict, evidence and sources for each question. Ready to export, fit for a regulator.
Illustrative example for a fictional company.
What's at stake
You answer for what AI says about you
Instead of Google
Customers increasingly ask AI — and trust what it answers.
MiCA, Art. 66
Information, including marketing, must be fair, clear and not misleading (for crypto-asset service providers).
MiFID II, Art. 24
The same duty for investment services. The "misleading" test is objective — intent is irrelevant.
Civil liability
You are liable for inaccurate or misleading information — including what AI says.
Sources: MiCA — Regulation (EU) 2023/1114, Art. 66; MiFID II — Directive 2014/65/EU, Art. 24.
How we differ
Others see only half. We tackle the correctness of what third-party AI says about you.
AI-visibility tools
They measure whether AI mentions you — not whether it tells the truth.
RegTech content checks
They watch your own content — not what third-party AI says about you.
FactGuard
Correctness + evidence of what third-party AI says about you — for regulated finance.
Honestly: large models may not always read your source. That's why the core value is measurement and proof — not a promised outcome.
Plans that grow with you
Start with a one-time audit. Add monitoring as you need it.
Pricing is being finalised — tell us your setup and we'll tailor a plan. Items marked "coming soon" are on the roadmap.
Free
Available nowSee where you stand.
- One-time AI audit (sample)
- Snapshot of key findings
- Clean facts layer (soon)
Pro
Coming soonStay on top of it.
- 1 source / domain
- Weekly monitoring
- 3 AI models + alerts
- Fix suggestions
Team
Coming soonFor growing teams.
- Multiple sources
- Monitoring 2×/week
- 4 AI models + competitor tracking
- API access
Enterprise
Coming soonCompliance at scale.
- Daily / event-based monitoring
- MCP server + SSO
- Regulator-ready evidence
- Priority support
Don't want a subscription? One-time audit
Available nowA full audit across all 4 models (ChatGPT, Claude, Gemini, Perplexity) — a timestamped, exportable report with sources and a verdict for every question. We run it for you and deliver within 1–2 business days.
- All 4 AI models
- Timestamped report with sources and verdicts
- Compliance-ready export
- No commitment — a single invoice
one-time · invoiced
Why you can trust us
Evidence, not promises
We show you — with proof — what AI says about you today, and clean facts raise the odds of a correct answer. We give you evidence you can act on.
Evidence you can stand behind
Timestamped, exportable records of every question, answer and source — built for EU regulated finance.
Disputed cases get a human check
The comparison is done by AI and can err, so low-confidence verdicts are flagged for human review — never buried.
Measurement, not magic
We raise the odds of a correct answer; it's not a switch that forces AI to comply. Anyone who promises that isn't being straight.
Frequently asked questions
Which AI assistants do you check?
ChatGPT, Claude, Gemini and Perplexity, each with live web search so the answers mirror what a real customer would see.
Can you guarantee AI will fix its answers?
No — and we won't pretend otherwise. We measure and document how AI represents you, and a clean facts layer raises the odds of correct answers. Forcing an AI to comply is not something anyone can promise.
How accurate is the comparison?
The answer-vs-fact judgement is done by an AI model and can itself make mistakes. We attach a confidence score and flag uncertain cases for human review.
What do I actually receive?
A clear report: each question, your verified fact, what each AI answered, the verdict and the sources cited — timestamped and exportable.
What is the MCP server and who does it help?
MCP is a standard “socket” that lets an AI query your verified source in real time and get a current, correct answer instead of a guess. It's most useful for a client's internal AI and agents (their own Claude/ChatGPT, support), where you control the connection. Public ChatGPT/Perplexity can't generally be wired to it yet — there, monitoring and the clean facts layer do the work. (Coming soon.)
Where is data hosted?
We are building toward EU-based hosting in line with the needs of regulated-finance customers. Ask us about your specific requirements.
Who is FactGuard for?
Primarily EU-regulated finance — MiCA-licensed crypto (CASPs) and fintech — but any company that cares how AI describes it.
Request your AI audit
Tell us your company and we'll show you how today's AI assistants represent you.
- A free sample audit of your domain
- Full 4-model comparison in the paid audit
- No obligation