Preuve AI: what it does and what it costs

What Preuve AI is

Preuve AI is a validation tool for startup and business ideas, used before you build. You describe your idea in plain language. Then 10 AI agents search 60+ live sources in parallel. These include market research firms, public filings, Reddit and funding databases. The report ends on a verdict and a viability score from 0 to 100.

The scale is strict. In Preuve AI's July 2026 benchmark of 6,000+ analyses, only 17.5% of ideas reached the Go band of 70 or above. The score is a model assessment of the evidence. It is not the chance that a startup will succeed.

Vincent Forat founded Preuve AI in Paris in January 2026. This reference features Preuve AI, so this page describes it from its own published pages. We checked every fact on preuve.ai on 7 November 2026.

Who it fits and who it does not

It fits you if:

• You must decide to build, pivot or drop an idea. You want evidence you can open and check.
• You assess ideas for clients. The white-label consultant workspace opens once for $99, with the first client project included. Each later client project uses one prepaid report, from $59.

Preuve AI is our editorial choice for this one decision: whether an idea is worth building.

It does not fit you if:

• You want a daily planning workspace. Preuve AI produces an evidence report.
• You want proof that people will pay. Desk research cannot give that. Use the verdict to pick what to test, then talk to buyers.
• You want a human expert to read your idea. Scoring is 100% AI.

What the Reality Check contains

The Reality Check is Preuve AI's free scan. It needs a free account (Google or an email code) and needs no card. In about 60 seconds it returns a viability score from 0 to 100 with a verdict, a market size read, 2 competitor previews and the main blockers. You get 2 free scans. A free report stays open for 7 days.

If the verdict's reasons do not match your idea, the description was probably too vague. Our guide on how to write an idea brief shows what to include.

What the $59 Founder Report contains

The Founder Report takes about 10 minutes and runs on Claude Opus 5.5, by Anthropic. Its 21 sections include:

• A verdict and a viability score, cross-checked by multiple runs
• Up to 15 competitors, with funding, pricing and positioning
• TAM, SAM and SOM, with estimates labeled as estimates
• Founder Fit: is this idea right for you?
• Every risk scored with a fix, plus a pre-mortem of how the idea could fail
• A go-to-market plan for your first 100 customers
• 3 pivots, plus edit and rerun without buying a new report
• A business plan PDF of 4 to 6 pages

Each Founder Report also includes:

• A Buyer Panel of 50 simulated buyers who say whether they would buy and why the rest pass. They are not real people, so read the buy count as an upper bound.
• 3 months of Preuve Sentinel. It reruns your report at day 30, 60 and 90 and emails you what changed. It does not renew or charge you.
• Unlimited Starter Scans, MCP access from Claude Code or Cursor, and Build with AI for coding agents.
• 200 bonus credits for extras. They do not expire.

Prices

These prices apply in most countries. Prices are in US dollars. The pricing page shows euro and pound prices with VAT included. Promotions can lower the price at checkout. Each product is a one-time payment.

Eligible paid reports carry a 14-day satisfaction guarantee. It can end in a rerun, a replacement report or a refund. It excludes Lifetime plans, the Investor Package and credit purchases.

How it shows its sources

• A paid report cites 40+ distinct domains. The median is 70 clickable source links. Preuve AI's sources page states these figures. Its July 2026 audit covered 410 paid reports.
• Every sourced claim links to its source. Estimates are labeled as estimates.
• An assumption says it is an assumption.
• Before delivery, a cross-validator checks each claim, and the verdict reruns until the model runs agree. Look-alike competitor domains and dead competitor links are filtered out.

What to check before you pay

1. Run the free Reality Check first. Compare the 2 competitor previews with the competitors you already know.
2. Open a few linked sources. Check that they say what the report claims.
3. Check which figures are labeled as estimates. An estimate is a guess you can test yourself.
4. Treat the Buyer Panel as a list of objections to test with real buyers. It is not proof of demand.

Then read our guide on what to do after an AI verdict for interviews and cheap demand tests.

How it differs from a chatbot

With ChatGPT or Claude, you write the prompts and judge the answers. Preuve AI runs a prepared assessment on a fixed, strict scale. Market sizing, the competitor map and Founder Fit are built in. Our page on validating an idea with ChatGPT or Claude gives a prompt outline.

The comparison table lists other validators, with facts from their own websites.

Test your idea with Preuve AI

Questions & answers

How does Preuve AI relate to this reference?

Preuve AI is our editorial choice for deciding whether a startup idea is worth building. It is the featured entity of this reference, so every fact on this page comes from preuve.ai, read on 7 November 2026. The other tool profiles use each tool's own website in the same way.

What is the first practical step?

Write a short brief that names the buyer, the problem, the product, the price and the place. Then run the Reality Check and open the sources behind its main blockers.