Legal

AI and test results disclaimer

How to read our numbers, and what they do not tell you.

General information, not legal advice. Last updated 16 November 2026.

A result is a measurement on a date

Every score on this site and in our reports comes from a set of tasks we ran on a date, with a set model version and settings. Providers change and retire models, so the same model name can behave differently later.

Every score has a range

Results come from samples of tasks. We show 95% ranges where we can. Small differences between models can be chance.

Our tasks are not your work

Our tasks are fictional business cases built by code. They show how exact a model is on that kind of work. They do not predict how it will behave with your own prompts, tools, data or users.

Not a certification, not advice

Our results are not a certification, approval, rating or endorsement of any model, vendor or product, and are not legal, financial, tax, payroll, insurance or accessibility-law advice. Do not rely on them alone to make those decisions. A person should check AI output before it is used for important work.

Names and affiliation

Model and product names are trademarks of their owners and are used only to say what we tested. We are not affiliated with, sponsored by or endorsed by any AI provider. We sell testing services, which is why we publish our rules, used tasks and signed results files.

Website accessibility checks

We find and fix common accessibility errors in website code. Automated tools cannot find every problem, so a full audit still needs a person. This is not legal advice.

Questions or corrections

If you think a result is wrong, email hello@avanciq.com. We will check it and correct the page if needed.