How We Score AI Tools

Every tool on this site gets a trust score out of 100, built from five dimensions. The rubric is versioned — when weights change, we re-score everything and say so.

The five dimensions

• Company & Identity (20 points) — Is there a registered company? Are the founders public? How old is the domain? Is there verifiable funding or press? Can you actually contact someone?
• Data & Privacy (25 points) — Does it train on your data by default? Can you delete your data? GDPR/CCPA commitments, security certifications, and any lawsuits or breaches on record.
• Billing & Subscription (20 points) — Transparent pricing, trial-to-paid behaviour, how hard cancellation really is, whether refunds are honoured, and verified billing complaints.
• Product Reality (20 points) — Does the output match the marketing? Is the free tier real? Does support respond?
• Community Reputation (15 points) — Trustpilot and app store ratings, Reddit sentiment, and repeated complaint patterns.

Grades

• A (80–100) — strong trust signals across the board
• B (60–79) — generally trustworthy with caveats
• C (40–59) — meaningful concerns; read the review before paying
• D (20–39) — serious concerns
• F (0–19) — avoid

The evidence rule

A score component does not exist on this site unless it is attached to at least one piece of documented evidence: a policy excerpt, a public source URL, a screenshot, or our own test log. If we have no evidence, the section says "Not yet tested" — we never fill gaps with guesses.

Automated signals vs hands-on tests

Company identity, privacy posture and community reputation are largely collected by automated crawlers (company registries, WHOIS, policy pages, review platforms) and refreshed on a schedule. Billing and product reality are tested by humans with real accounts — we sign up, pay, and cancel.

What we exclude

We do not list AI companion/NSFW services, trading or investment bots, or tools that claim to diagnose or prescribe. These categories carry risks our rubric is not designed to measure.

AI assistance disclosure

We use AI to assemble review text strictly from our own structured data — every generated sentence maps back to an evidence record, and the mapping is stored for audit. AI never invents claims, scores, or test results.