Method

How we test

Anyone can post a screenshot. We show the tool, the date, the fake data we used, what went wrong, and a badge that says how much we actually checked.

The three badges

✓ Built & tested by us

We produced the result ourselves and show the real, unedited output.

🔬 Tech tested · app test pending

We proved the capability with an open-source model, but haven't yet tried the consumer app we recommend.

⚠ Not yet tested

Instructions come from the tool's documented features. A scripted test with fake data is queued.

Featured vs in testing

Featured experiences have a full page and a public test record. Anything else sits in “In testing” on the home page until a real test has run. Right now only 1 featured experience is fully built and tested by us; the rest are labelled.

Our loop

  1. Discover
    Official release notes, public demos and what beginners actually ask.
  2. Score
    Surprise, usefulness, ease, cost, how visual, how shareable.
  3. Test
    With invented or public-domain data only: fake bills, fictional CVs, public-domain photos. Never real personal data.
  4. Record
    Exact prompt, inputs, tool and plan, dated raw output, failures and test code go into a public test record.
  5. Publish
    Only real results. Failures become honest notes, not hidden.
  6. Re-check
    AI tools and their free limits change. We re-check them and mark the date (last full check: 15 Oct 2026).

What we never do

  • Show a simulated or staged result as real.
  • Use anybody's real documents or photos without permission.
  • Copy other people's articles, videos or images.

Image credits: “Cat November 2010-1a” by Alvesgaspar, Wikimedia Commons, CC BY-SA 3.0 (background removed; our version is CC BY-SA 3.0). “Migrant Mother” by Dorothea Lange, 1936, public domain (Library of Congress); scratches added artificially by us for testing.