Method

How we check.

Playful on the surface. Rigorous underneath. Here is exactly how every verdict is earned.

The Pist Score

35%ProofDoes it do what the vendor says, on identical inputs?

25%CraftWould a working professional ship the output?

15%ControlCan you steer, edit and iterate, or only re-roll?

10%Cost clarityCan you know what a job will cost before you start?

15%Creator ownershipDo you own what you make, and is your work kept out of training?

Each dimension is scored 0–100. The Pist Score is the weighted sum.

Verdict stamps

Probatum

Latin for “it has been proven.” Did what it claims, on identical inputs, at professional quality.

Overclaimed

It works — just not as advertised. Useful, with caveats you should know before paying.

Pist

Failed our test. The claim didn't survive contact with a real brief.

Pending

Indexed and sourced, but not yet tested. No score, no opinion.

Claim statuses

Vendor claimWhat the vendor says. Recorded, not endorsed.
Third-party reportedReported by a credible outside source we cite.
VerifiedWe tested it ourselves and it held up.
ContradictedOur tests, or solid evidence, say otherwise.
OutdatedWas true once. Kept for the record.

The wrapper check

Plenty of “AI tools” are interfaces over someone else's model. That's fine — but you deserve to know. We label every tool:

Own modelsRuns models it built.
HybridA mix of its own and third-party models.
AggregatorResells other companies' models.

The test kitchen

Every tool gets the same brief from the same fictional brand — identical inputs, assets and constraints. No cherry-picked demos, no vendor-supplied prompts. That's how scores stay comparable.

06 — The independence pledge

We take no vendor money. We use no affiliate links. We never rank untested tools by opinion. Every fact links to its source, with a date.