Our review methodology
Every score on CompareBest is the result of structured hands-on testing, not PR briefings or spec-sheet comparisons. Here is exactly how we do it.
The five scoring dimensions
Each AI tool is scored out of 100 across five equally-weighted dimensions. The final score is an aggregate of all five.
Ease of use (20%)
Time to first useful output, clarity of the interface, onboarding quality, and how well the tool handles first-time users. Tested by evaluators with varying technical backgrounds.
Features (20%)
Breadth and depth of capabilities relative to stated category. A coding assistant is scored on code quality, debugging, IDE integration, and language support. An image generator is scored on output quality, style range, and editing controls.
Value for money (20%)
Output quality per dollar at each pricing tier. A tool can score highly here with a generous free tier. We penalise tools where paid plans do not meaningfully exceed free alternatives.
API quality (20%)
Documentation quality, SDK availability, latency, uptime history, rate limits, and developer experience. Tools without public APIs score a maximum of 50 in this dimension.
Support quality (20%)
Response time and quality of official support, quality of documentation, community health, and availability of status pages. Tested with genuine support queries submitted under standard user accounts.
Pricing verification
Pricing data is checked monthly against each company's official pricing page. We record the URL and date of each check. Where pricing changed during a review period, we show both prices and note the effective date of the change.
We do not use third-party pricing APIs or databases, as we have found these to be unreliable. All pricing on this site is verified first-hand.
Conflict-of-interest policy
CompareBest uses affiliate links to generate revenue. These are links to company websites that pay us a commission if you sign up. All affiliate links are labelled with "sponsored" or similar language in close proximity.
Affiliate relationships do not affect editorial rankings or scores. A company cannot improve its score by offering higher commissions. Companies that pay us nothing score identically to those who pay us something, by the same criteria.
Vendors may not review or approve content before publication. We do not accept payment for positive coverage.
Review update policy
Each review page shows when it was last updated. We commit to reviewing all listed tools at minimum quarterly. When a material change occurs (pricing increase, new model launch, significant feature removal), we aim to update within 14 days of the public announcement.
If you notice outdated information, please contact us at editorial [at] compare-best.com. Verified corrections are applied within 48 hours.