How we review

Most AI reviews tell you whether a tool's output looks good. We focus on what you agree to when you pay: the documents nobody reads.

What we read

For every tool, we read the vendor's current public pages: the pricing page, the terms of service, the refund or billing policy, the privacy policy and the help-centre articles about limits, credits and data. Each review lists these sources with the date we read them.

We do not claim hands-on testing in these reviews. When we cite user complaints or vendor claims, we say so and write them as claims.

The five questions

The sixth grade: what users say

Documents tell you what you agree to; users tell you what actually happens. For each tool we read its public ratings on review platforms (Trustpilot, G2, Capterra, and the App Store or Google Play when there is an app), plus recent reviews and discussions, and summarize the complaints and praise that keep coming back.

The user grade comes from a fixed rule, not from our impression. We take the plain average of the platforms with at least 20 reviews, each platform counting once so that a huge app-store count cannot drown the others: A from 4.5 out of 5, B from 4.0, C from 3.5, D from 2.5, E below. With fewer than 100 reviews in total, the grade cannot go above C, and with a single platform it cannot go above B. If no rating could be read, the grade is C and the review says so. Platform figures are shown with the date we read them.

Review platforms have known biases: unhappy customers write more, and some vendors invite satisfied ones. Averaging several platforms limits both effects; it does not remove them.

Grades

Each question gets a grade from the buyer's point of view:

• A clear terms that protect you (for example, a no-questions refund window, or no training on your content by default);
• B fair, with minor catches;
• C standard industry terms, or documents too vague to grade higher;
• D terms that mostly favour the vendor;
• E terms that can cost you money or rights with little warning.

When a document is silent on a question, we say so and the grade cannot go above C. The Zone score is the average of the six grades (the five fine-print grades and the user grade; A = 4 points, E = 0), out of 10. It measures how fair the deal is and how users rate the experience; the verdict covers the rest.

Updates

Prices and terms change often. Each review shows the date its documents were read. When we re-read them, the date changes and the grades are revised.

Money

Some links are affiliate links: we may earn a commission if you buy, at no extra cost to you. Vendors cannot pay for a grade, and a tool without an affiliate programme is graded the same way. See our affiliate disclosure.