How we test, and why rankings aren't for sale

Our rules

1. No pay-to-rank. Ever. Money can't buy a top pick, a budget pick, removal from "avoid", or a better score. Vendors can't review or edit a page before it's published.
2. Untested means labelled untested. Every use-case page shows its test status. "Pending hands-on test" means the page relies only on facts we verified from vendor sources. We never invent test results, screenshots or scores.
3. Every price has a source and a date. Prices change constantly, so each price links to where we found it and shows the date we checked. We re-check every page at least every 30 days.
4. Every page names something to avoid. For each job we name a tool or plan that's the wrong choice and explain why.
5. Corrections are public. If we get something wrong, we fix it and note the change on the page.

How we test

Every use case has a standard task, for example "build a to-do app with auth and deploy it" or "turn a 30-minute podcast into five vertical clips". We run the same task on each tool, save the outputs, and score them on output quality, speed, real cost for the task, ease of use and lock-in. Tools we couldn't test (no trial, enterprise-only) are marked as such.

Parts of our research and drafting are done by AI agents: discovering new tools, collecting pricing, and drafting pages. A human reviews top picks and anything with a commercial relationship before it's published.

How we make money

• Affiliate links: some outbound links may earn us a commission if you buy. They're marked rel="sponsored". They never change a ranking, and plenty of our picks have no affiliate program.
• Newsletter sponsorships: always labelled "Sponsored" and kept separate from editorial picks.
• Paid test requests (later): vendors may pay us to test a tool faster. They don't get to choose the result, and we publish it whatever it is.

Contact

Corrections, tips or vendor questions: contact details coming soon.