Every tool page and article on The AI Bucket carries one of three evidence labels. They tell you how much weight to give what we say.
Evidence labels
- Tested hands-on: we used the tool ourselves on a real task and kept the results.
- Desk research: we checked the vendor’s own documentation, pricing page and public material, but did not run our own test.
- Not yet reviewed: a directory listing only. Treat it as a pointer, not a recommendation.
What a hands-on test records
- The input (for example, the same client recording for every tool in a comparison).
- The task, the settings used and the date of the test.
- The output, how long the work took, and every manual correction it needed.
- Where the tool failed or produced something we would not send to a client.
Prices
Prices change often. When we show a price we also show the billing basis (for example, per user per month, billed annually), the date we checked it and a link to the source. If a price is older than that date, check the vendor before you buy.
No star ratings
We do not give tools star ratings. A short verdict, who a tool is best for, who should avoid it and its limitations are more useful than a number, and they are harder to fake.
Corrections
If you spot a wrong price, a wrong description or a tool that no longer exists, please tell us. We fix confirmed errors and update the “last verified” date.