How we test

Every verdict comes from running the same real tasks through each AI and scoring the results on one rubric.

Three rules for every verdict

We run an identical set of real tasks through each AI, on its free and paid plans.

Accuracy, quality, speed and price, weighted for the job. The rubric is on every page.

Models change fast. Every verdict shows the date it was last tested.

A verdict marked draft is an early pick that has not been through the tests yet. Draft pages are not submitted to search engines and show no scores.

A price appears only after we check the vendor’s pricing page; the page shows the date and links the source.

Each AI gets a 0–10 score on the task’s rubric. A score of “Not scored” means the AI has not run the tests yet.

Some links may earn us a commission. Those links are marked as sponsored, and a commission never changes a verdict.