Plain-language policy reports for the agreements people usually skip.
fineprint analyzes Terms of Service and privacy policies with AI, then turns them into clear, scored breakdowns of the clauses most likely to affect consumers. It is part pre-signup research tool, part public database of reports the community has already run.
Paste a policy URL
Start with a Terms of Service, Privacy Policy, subscriber agreement, or billing terms page. fineprint fetches the public document and extracts the readable text.
Evidence is extracted and scored
The analysis finds relevant clauses across the full policy set, verifies each quotation against its source, and applies a consistent deterministic rubric before summarizing the findings.
Results become searchable
Every completed analysis gets a shareable report and joins the public database, so future visitors can compare companies without rerunning the same document.
Use fineprint before creating an account, subscribing to a service, or sharing sensitive data. Reports are designed to surface tradeoffs quickly, not replace legal advice. fineprint is independent: a company's appearance does not imply endorsement, sponsorship, partnership, or any other relationship. Reports are AI-assisted and should be checked against their linked sources. Read the full disclaimer.
Scored breakdowns should be comparable across companies.
Findings should stay grounded in the document, not vague policy commentary.
A thorough run is worth a short wait when the fine print is long.
Thomas Anderson is a recent finance graduate of the College of Charleston. He built fineprint because he didn't like not knowing what he was signing up for. These are agreements almost everyone accepts without reading, on the fair assumption that reading them is not realistic.
It is a solo project, built over roughly four months on Next.js, Supabase, and the OpenAI API. The current analysis engine uses GPT-5.6 Luna to extract evidence from complete policy sets, while application code verifies each quote and calculates impact scores through a versioned rubric. The harder problems were the ones that do not show up in a screenshot: making impact scores comparable across companies, finding relevant clauses beyond a document's opening pages, and ensuring that a real quotation actually supports the scored finding attached to it. Those changes came out of measurement — predictions written down before each evaluation run, saved evidence replayed through revised rules, and results allowed to contradict the original assumptions.
Ready to inspect a document?
Paste a public policy URL or browse reports others have already generated.