How reaso.ai works

reaso.ai stress-tests popular claims using adversarial AI analysis and real evidence from peer-reviewed studies. Two independent AI models generate arguments, search for evidence, and deliver verdicts — with full transparency about sources, limitations, and what we still don't know.

The methodology

1

Challenge

AI generates the 3 strongest counter-arguments against the claim

2

Evidence

Real studies searched from PubMed, journals, and institutional sources

3

Defend

AI builds the strongest defense for each challenge, backed by evidence

4

Verdict

An independent AI model weighs all evidence and delivers the final assessment

Step 4

Verdict — an independent AI model evaluates all evidence

Claude Opus 4 (Anthropic) reviews the full argument tree independently and delivers a verdict. The verdict model is separate from the argument generation model to reduce self-evaluation bias.

The verdict scale

Supported

The claim survived all major challenges with strong evidence

Holds with caveats

The claim is broadly correct but needs meaningful qualification

Overstated

The claim contains a kernel of truth but exaggerates what the evidence supports

Not supported

The literal claim is contradicted by strong, independent evidence

AI models used

Claude Sonnet 4 (Anthropic) — generates counter-arguments, builds defenses, searches for evidence, classifies source credibility, and generates FAQ entries.

Claude Opus 4 (Anthropic) — delivers the final verdict independently after reviewing the full argument tree. Separate from the argument generation model to reduce self-confirmation bias.

Serper API (Google Search) — finds real evidence from peer-reviewed journals, institutional reports, and news sources. Each source URL is linked so readers can verify independently.

Who built this

reaso.ai was created by Thiago Alvarez — a software engineer focused on building AI tools that help people think more clearly about evidence. The goal is simple: when someone shares a study or makes a health claim, you should be able to see what the full evidence landscape looks like, not just one side of the argument.

The project is independent and self-funded. It is not affiliated with any health organization, supplement company, or political group. The only agenda is showing what the evidence says.

Limitations

AI can misinterpret study findings, miss relevant research, or produce overconfident conclusions. Evidence searches are limited to what is publicly accessible via web search — paywalled studies may be missed. Verdicts reflect evidence available at the time of analysis and may change as new research emerges. Results may vary between runs. This tool reports findings — it does not provide medical, nutritional, legal, or professional advice.

Contact

Found an error? Have feedback? Reach out at hello@reaso.ai.

Try it yourself