How it works
From a tag on X to a report with receipts.
Scroll to follow one investigation. (Demonstration — example content)
01 · Tag
It starts with a post.
A project makes a claim on X. Anyone replies with @proofbot_ to request an investigation.
02 · Extract
03 · Investigate
04 · Score
05 · Publish
Example Project
@exampleproject
Our AI trading agent is live 🚀 Open-source, 10,000 users, audited. Token launching today.
A curious user @someone
Is any of this real? @proofbot_
The pipeline
Seven stages, every time.
Click a stage to see what's collected.
Submitted
A tag on X or a submitted link creates a job. The original post, author and target are recorded.
Methodology
Scores are earned by evidence, not vibes.
Each finding gets one of three states. The evidence score weighs how many important claims were verified against how many were contradicted.
Verified
Independently confirmed with a linked source you can open.
Unverified
No public evidence either way. Not the same as false.
Contradicted
Public evidence directly conflicts with the claim.
Reading an evidence score
Weak evidence
Core claims are contradicted or the product could not be found working.
Mixed evidence
Something real exists, but important claims remain unproven.
Strong evidence
Most claims are backed by working product, code and onchain data.
Limitations
- Only publicly available information is examined. Private repos and closed betas can't be verified.
- A report is a snapshot. Projects change — request a new investigation to refresh it.
- A high score is not financial advice or a guarantee of safety.
- Unverified means unknown. It is never treated as a negative finding.