How we test

One job, one task, every tool. Then we pick one and tell you who should skip it.

1. Same task for every tool in a job

Each ranking starts with a single, realistic task written down before we touch any tool. For coding agents that means the same repo and the same tickets; for app builders, the same app; for notetakers, the same recorded meetings. Every tool in the job gets exactly that task, so results are comparable.

2. Free tiers and free trials, like you would

We sign up the way a normal person would: free tiers first, then free trials of paid plans. We cancel trials before they bill. If a tool has no free way in, we either pay for a month or mark it Not yet tested. We don't accept special reviewer accounts that ordinary users can't get.

3. What we score

• Result quality: did it actually do the job, and how much fixing did the output need?
• Speed: time from sign-up to a usable result, and time per task.
• Ease: how many steps, nudges and workarounds it took.
• Price and free tier: what the task cost, and how far the free tier gets you.

We keep screenshots and notes from each run and publish the test date on the ranking page.

4. One pick, and who should skip it

Each job gets one Our pick. Every tool, including the pick, gets a plain "Skip it if" line, because the best tool for most people is the wrong tool for some.

5. Dated re-tests

AI tools change monthly. Every ranking shows when it was last updated, and we re-run the task on a schedule and whenever a tool ships a major change. A new release only enters a ranking if it beats the current pick on the same task.

6. Rankings are never sold

No one can pay to be ranked, to move up, or to be called our pick. There is no fee to submit a tool.

If we ever sell sponsored placements, they will always be clearly labelled Sponsored, shown separately, and kept out of the rankings themselves.

7. Affiliate links

In future, some links to tools may be affiliate links, which means we may earn a commission if you buy. We only add them to tools that have already earned their place, and they never affect the order. See the disclosure.

8. Not yet tested

Anything we haven't run through the hands-on task is labelled Not yet tested. Today that is every tool on the site. The label comes off only after the test, and the result is published on the ranking page.

Last updated: 14 Nov 2026. Spot a mistake? Email [email protected].