How we decide

What a verdict means

Pick: worth paying for, for most people who need that kind of tool, at today’s price. Consider: right for some people; the verdict says who. Skip: there’s a better or cheaper way to get the same thing, and the verdict says what it is.

Every tool also gets who it’s wrong for, and the catch, because a recommendation without its downside is an advert.

Where the facts come from

Prices, plan limits and data terms come from each vendor’s own pricing, help and privacy pages. Each tool lists its sources and the date we read them. Where a vendor doesn’t publish something, we say “not published” rather than guess.

Prices are shown the way the vendor shows them to a UK buyer, in pounds where it bills in pounds, with whether VAT is included.

Research-led, for now

Every guide is labelled research-led: built from current, cited sources, not from long-term use. When a verdict comes from hands-on use, the guide will say so.

Written with AI, and the conflict that creates

aitools.direct is researched and drafted with help from AI tools, including Claude, made by Anthropic. Anthropic’s products — Claude and Claude Code — are also among the tools we cover. That is a conflict of interest.

Here is how we handle it. Every verdict rests on facts you can check, linked and dated. Anthropic’s products get exactly the same fields, sources and standard as everyone else’s. And the final call on every verdict is made by the people who publish the site, not by the AI that helped draft it.

How the site makes money

Where a vendor runs a public affiliate programme, a link to it may earn us a commission. Those links are labelled next to the link and never sit inside a verdict.

No vendor can pay to be a Pick, to rank higher, or to change a verdict. No sponsored reviews.

Corrections

Prices change weekly. When we get something wrong, we fix it and log the fix, with a date, on the corrections page.