One job, written down first. Every issue starts with a specific task, for example "transcribe and summarise a 30-minute meeting" or "build a one-page landing page and publish it free". The job decides what matters, not the tools' feature lists.

Three tools, chosen because people actually use them for that job. We look at what readers, Reddit threads and searches ask about, then pick three well-known options.

Published limits and prices, from the company itself. We compare each tool's free-plan limits, paid prices and key features using its own pricing pages, help centre and docs. Every claim in an issue links to the page it came from.

One winner, with "use it if" and "skip it if". The winner is the tool that fits the job best for most people on the evidence we have. We also name a runner-up and a best pick for a specific group (for example students, or people willing to pay) when there's a clear one.

We say when we haven't tested something hands-on. So far, issues compare published features and free-plan limits. We haven't run hands-on tests, so we don't make claims about accuracy, writing quality or design quality. If we do run a hands-on test in future, the issue will say so and describe the test.

We don't use reviews from other sites as evidence. If we ever quote third-party data, it's labelled as third-party.

We don't give legal, financial or medical advice. Where a job raises a legal point (for example recording a meeting), we flag it and tell you to check the rules that apply to you.

Prices are the company's US dollar list prices, converted to pounds at the European Central Bank reference rate on the day we check (via the Frankfurter service), rounded to the penny. Each issue states the rate it used.

Tax (for example UK VAT) may be added at checkout. Introductory or first-time-customer offers are labelled as such when the company says so.

Every issue says when its facts were checked, and the one-line verdict near the top carries its own "checked" date.

We re-check every price and limit on the company's own pages shortly before an issue is sent.

AI tools change their plans often. When we spot a change in a published issue, we update the web version, add a "re-checked" date, and, if the change affects the verdict, say so at the top of the post.

If you spot something out of date or wrong, email us (below) with a link to the source. We aim to check it within a week.

No company pays to be included, and no company sees or approves what we write before it's published.

Affiliate links: Tool Picker may earn a commission if you sign up through some links, at no extra cost to you. Every affiliate link is marked (ad), and each issue says whether it contains any. As of 29 Oct 2026 we haven't joined any affiliate programmes, so no issue so far contains an affiliate link.

Commissions never decide the winner. If a tool with an affiliate programme loses, it loses. If we join a programme for a tool we've written about, the next issue that links to it will say so.

Sponsors: if an issue is ever sponsored, the sponsored section will be clearly labelled as an ad, and the sponsor will have no say over the comparison.