Terms of use

Short, in English, and about the things that could actually go wrong: a price that moved, a score you want to quote, and a mistake you want fixed.

Last updated 17 August 2026

What you are reading

AI Intelligence is an independent testing site. It publishes the results of tests we ran ourselves: AI tools and models answer identical published task suites, three AI judges from three different labs mark every answer blind, and the score plus the raw evidence is published together. The whole protocol is public.

By using the site you accept the terms on this page. They apply to reading it, quoting it, and using the read-only API. Nothing here is a contract to supply you with anything: the site is free, there is nothing to buy from us, and there is no account to hold.

What a score is, and is not

It is our own measurement. A score is what our published task suite produced on the date shown, marked by the judges named in the receipts. It is not a vendor's claim, not a certification, and not an endorsement by anyone but us.
It is a measurement of a test, not of your job. The suites are built to resemble real work, and they are still finite. A tool that wins our writing suite may lose at yours. Read the tasks — they are published — and judge whether they look like what you need.
It is not advice. Nothing here is legal, medical, financial or professional advice, and the boards covering legal or health questions test how well AI tools answer such questions against public guidance. They are not answers to yours.
It can be superseded. Models change under the same name. A score carries the date of the run that produced it, and a newer run replaces it. Trust the date, not the memory.

Prices and features belong to the vendors

Every price, plan name, free tier and data-handling answer on this site was read from the vendor’s own page and is published with the source link and the date we read it. That is the strongest guarantee anyone can give you, and it is not a guarantee that the fact is still true: a vendor can change a price, rename a plan, withdraw a free tier or rewrite a privacy policy the day after we read it, and they do.

So before you spend money, click the source and check the vendor’s own page. We do not warrant that third-party prices, plans, features or terms are current, and we are not responsible for a purchase you make from a third party or for what happens on their site. If you find a stale figure, tell us — that is a correction and we publish those.

Tool and model names, logos and trademarks belong to their owners. We are not affiliated with, endorsed by, or sponsored by any of the companies we test, and a score should never be presented as their approval of us or ours of them.

The site is provided as it is

We work hard to be accurate and we publish our mistakes when we find them, but the site is provided without warranties of any kind: that it will be available, that it will be free of errors, or that it is fit for a particular purpose. To the extent the law allows, we are not liable for loss arising from a decision you took because of something you read here — and the practical protection against that is the one this whole site is built on: every claim links its evidence, so you can check it before you act on it.

Nothing in these terms limits liability that cannot lawfully be limited.

Quoting us: yes, please

The task suites, the judging rubric, the scores, the receipts, the article text and the design of this site are ours. The permission we give is broad on purpose — a test nobody can cite is a test nobody can check.

Fine, no permission needed

Quoting a score, a verdict, a scoreline or a passage, in an article, a video, a deck, a forum post or an AI-generated answer, with attribution to AI Intelligence and a link to the page it came from. Screenshotting a table. Citing our methodology. Using the read-only API to build something, within reason and with attribution.

Not fine

Republishing our rankings wholesale as your own content. Systematically scraping the site or hammering the API. Presenting a score as a vendor’s endorsement, or as your own testing. Using a score in advertising in a way that implies we endorse a product. Removing a date or a source from a figure you reproduce, which is how a live fact becomes a false one.

Vendors: you are welcome to quote your own result, win or lose. Please quote the score with its date and link the page, so a reader can see what it was measured on.

If we got something wrong

Tell us and we will fix it in public. We correct the claim, mark the old value superseded, and record it in the public changelog — corrections are never silent, and that applies whether the error was ours, a vendor’s or a judge’s.

Email corrections@aiintelligence.com with the page and the sentence, or use the contact page. The full corrections policy says what happens next, and the changelog is where it lands.

What we will not do is remove a verdict because a vendor dislikes it, take a product down on request, or let anyone see a result before it is published. If you believe a score is wrong, show us the evidence and we will re-run the test.

Money, changes and law

How the site is funded is set out in full on how we make money: no vendor pays for placement, position or inclusion, and commission from a labelled paid link never affects a score or the order. That page is part of these terms in the sense that it is the promise we can be held to.

These terms can change. The version you are reading carries its date at the top. Continuing to use the site after a change means accepting the current version; we will not pretend to have obtained your agreement to it any other way.

Law. The site is published from the United Kingdom, and these terms are governed by the law of England and Wales.