Know What's True.Prove It.
Trace every claim to its origin. Expose distortions, fabrication, and bias. Fix what's weak — cited or uncited.
CHF 500 one-time · 500 spots · 30-day money back
The Stakes
The Cost of Getting It Wrong
Misinformation spreads in seconds. AI-generated content looks flawless. In a world where anyone can publish anything, verification isn't optional — it's survival.
Even Legal AI Hallucinates
In Stanford's preregistered evaluation, purpose-built legal AI research tools hallucinated 17–33% of the time — general chatbots, 58–88% on legal queries.
Today's AI Still Invents Sources
In a 2025 peer-reviewed study, GPT‑4o fabricated nearly one in five citations in literature reviews — and almost half of the real ones contained errors.
Courts Are Acting
Court decisions worldwide documenting AI-hallucinated content in filings — 717 involving lawyers — as of 5 August 2026. Sanctions have reached six figures, alongside disqualification and license suspension.
The Core Problem
Three Layers of Verification Intelligence
Surface-level tools check if a source exists. We go deeper — analyzing intent, detecting bias, and finding evidence where none was given.
Detect Distortions & Inaccuracies
Claims that overstate, misrepresent, or twist what sources actually say. A paper says "suggested" — the report says "demonstrated." We catch the difference.
Expose Bias & Fabrication
Selective citation, cherry-picked data, fabricated references.
Source Stronger Evidence
Our engine detects uncited claims and automatically searches academic and institutional databases (CrossRef, OpenAlex, Semantic Scholar) for supporting or contradicting evidence.
Our Own Numbers
Measured, not promised.
Before you trust a verification engine, eight questions matter. We answer each with our own measured runs — real dates, nothing rounded up — and publish the full methodology.
Can it be fooled into approving something false?
of deception traps rejected
114 claims engineered specifically to trick a verifier, hidden among genuine ones across five validation runs. It approved none.
N=114 · Internal validation Jun–Jul 2026Does it catch wrong numbers?
of planted errors found
We took real company filings and secretly altered figures. The engine caught every one — and didn't flag a single figure that was actually correct.
N=20 · Real filings · Jun 2026Does it ever reject something true?
true claims condemned
Across every validation run: no genuinely supported claim has ever been marked unsupported. When the engine is uncertain, it says so and shows the evidence.
Internal validation Jul 2026How closely does it match human experts?
agreement with expert grading
Against claims hand-labeled by human experts, on the strictest scoring. The one disagreement: the engine was stricter than the human.
N=34 · Jul 2026See It For Yourself
Read a real verification dossier
A real VeraCita dossier on AI adoption in European SMEs: 7 claims analysed, every verdict backed by evidence you can read.
Founder Circle
The lowest VeraCita rate we will ever offer
500 spots. One price. Your rate locked forever.
One-time payment. Lifetime access. No subscription required.
- 50,000 credits included — 10,000 claim verifications
- 0.01 CHF/credit locked forever (vs CHF 0.03 pay-as-you-go)
- 30-day money back — full refund minus credits used
- 10,000 credits per referral — worth CHF 100; 5 referrals = paid back
- Unlimited seats — single shared credit pool, no per-seat fees
- Full API access included
- Credits spend across every MindTheGap product, current and future
Strictly limited to 500 global spots
Pricing
Simple, Transparent, One Credit System
No hidden fees. No per-document charges. You see the exact cost before you commit.
Founder Circle · Available now
0.01 CHF/credit — locked forever
- 50,000 credits included — never expire
- After exhaustion: top up at the same CHF 0.01/credit
- Unlimited seats · Full API access
Pay-As-You-Go · Available now
0.03 CHF/credit
- Top up from CHF 20 — use as much or as little as you need
- Prepaid top-ups — credits never expire
- Credits shared across all MindTheGap products
Pay-As-You-Go runs in the app, which is built for a desktop screen.
Subscription
—
Credit rate and allowance announced at opening. Opens once the Founder offer closes.
Questions
The things people ask first
Every claim backed by a citation gets verified against its original source. Our AI gives you a confidence score, source quality rating, and a clear verdict: Supported, Partially Supported, Contradicted, Unsupported, or Unverifiable. No more guessing if your citations are accurate.
Yes. Our engine detects uncited claims and automatically searches academic and institutional databases (CrossRef, OpenAlex, Semantic Scholar) for supporting or contradicting evidence. You get a verdict per claim — Supported, Contradicted, or Unverifiable — backed by sources.
Credit cost depends on your document complexity and verification depth. Simple claims cost 1 credit, complex claims cost 5 credits. Source extraction costs 2 credits per source. Chain analysis, source intelligence, and repair have their own credit costs — all shown as an itemized breakdown before you commit. No hidden fees, no wasted spend.
Founder Circle members get a 30-day money-back guarantee, no questions asked: a full refund minus any credits you've used, billed at the Founder rate of CHF 0.01/credit. If VeraCita doesn't meet your expectations within 30 days, contact us.
Swiss-owned company: we store your data exclusively in Switzerland (AWS Zurich). Your document is parsed in your browser. In Server mode, verification runs on our Swiss servers with AI processing in Switzerland and the EU via AWS Bedrock; in Zero-Knowledge mode nothing reaches our servers. Your data is not used for model training. Uploaded files are deleted once processing completes; verification results are retained so you can return to your session — delete them at any time.
You can sign up, buy and manage your account from your phone. Verification itself — uploading a document and reading through the evidence for each claim — is built for a desktop screen, and that is where we recommend doing the work.