1. What we read
Each raw score is built from evidence in the actual Terms of Service and Privacy Policy text. Rank and percentile are shown only when evidence coverage and score-trust gates are high enough.
Methodology
How WhatTerms scores a service.
One rubric, applied the same way to every service. Scores change when the policy text changes — not because of who the service is.
How to read a WhatTerms score
Inspectable in seconds: what evidence was read, which facts moved the evidence score, and whether the consumer-grade utility gate has passed.
Policy text evidenceProfile-weighted scoringSupporting trust signals
2. What profiles change
Profiles do not change the facts. They change how much each category matters in the final score so different tradeoff preferences are easier to compare.
3. What trust signals mean
External security signals support the picture, but they are not treated as proof that a company is broadly secure or fair. Public help-center claims and independently observed signals are labeled separately.
4. How to use the scale
Raw scores are source-backed evidence values mapped to a consumer grade. Public bands, ranks, and percentiles are shown once the consumer-utility gate confirms the score spread is useful — which it now does.
When a consumer band is shown, any verified severe risk — like forced arbitration, selling your data, or no self-service deletion — also appears as a specific red-flag chip next to the band, so a high score never hides a serious catch.
- Excellent85–100
- Strong70–84
- Fair50–69
- Risky25–49
- Critical0–24
What each category measures
Hover or focus any term to see a plain-English definition.
- Data practicesData practicesHow the service collects, shares, sells, and retains your data — and what controls it gives you over each. The biggest single driver of the score.
- DeletionDeletionWhether you can fully delete your account and data, how long that takes, and whether the company keeps anything after deletion.
- Legal rightsLegal rightsWhat recourse you have when something goes wrong — class-action waivers, arbitration clauses, jurisdiction lock-ins, and whether you can sue.
- Content ownershipContent ownershipWho owns what you upload — your photos, posts, comments — and what licenses the company takes over your content even after you leave.
- CancellationCancellationHow easy it is to cancel a subscription or close your account. Auto-renew clauses, free-trial traps, and dark patterns all hurt this score.
- Account terminationAccount terminationWhether the company can terminate your account at will, with or without notice, and whether you get any meaningful appeal.
- Unilateral changesUnilateral changesWhether the company can change the agreement on its own terms — usually with vague notice — and what your options are if you object.
- Security trustSecurity trustIndependently observable security signals: bug-bounty programs, security.txt files, two-factor support, breach history, and similar.
How peer ranking works
A service can lead its realistic peer set without meeting the absolute ideal. Signal is compared against private communications peers; Bandcamp is compared against creator marketplaces. Peer rank explains market context when publication gates are normal, while the band still reflects the consumer-protection standard.
This is not legal advice
WhatTerms helps you make more informed decisions about the services you use. It is not a substitute for legal advice, compliance certification, or a security audit.
The rubric at a glance
We read each service's published Terms of Service and Privacy Policy, extract criteria (e.g. "does not sell personal data", "mandatory arbitration with no opt-out", "at-any-time changes"), and score eight categories on a 0–100 scale:
- Data practices (sharing, selling, advertising)
- Deletion & retention
- Legal rights (arbitration, class actions, jury trials)
- Content ownership & AI training
- Cancellation & subscriptions
- Account termination
- Unilateral changes
- Security & trust signals
The overall score is a profile-weighted blend of the eight category scores. Profiles (Balanced / Privacy-first / Legal-rights-first / Convenience-first) change the weights but never the facts.
Two surfaces, one methodology
The directory uses the deterministic rubric described above. The /scan tool uses a narrower 5-bucket LLM analysis for services that aren't in the directory yet. The rubric is the canonical WhatTerms score. See the decision record for why they're separate today.
Not legal advice
WhatTerms helps you make more informed decisions about the services you use. It is not a substitute for legal advice, compliance certification, or a security audit. Always read the source policy for any decision with real stakes.