How we review AI tools
Reviews are published under the AIUnpacking Team, an organization byline. We do not invent named experts or credentials. This page describes the method we use, including what we can and cannot claim.
What a review is
A review is a structured evaluation of a product: what it is for, what we could verify, where it is weak, and who it is a poor fit for. Scores are editorial judgments on a 0–10 scale, not lab benchmarks unless a page says otherwise.
What we look at
- Product surface: documented features, pricing pages, and official limits.
- Hands-on use when we have access. If we did not run a full test, the page should not imply that we did.
- Primary sources for prices, model names, policies, and legal claims.
- Failure cases and constraints: context limits, data handling, regional availability, and total cost.
Scores and verdicts
The score summarizes usefulness for the stated audience, not a universal ranking. Verdicts sit at the top of the review so a reader can get the answer first. We do not sell scores. Sponsored placements, if present, are labeled on the page and do not purchase a rating.
Dates and updates
Publish dates record when the review first shipped. Updated or verified dates change only after a substantive edit or a re-check of volatile facts. We do not refresh dates for formatting-only changes.
Sources and corrections
Structured source lists should sit on reviews that cite prices, specs, or third-party claims. If we get something wrong, email corrections@aiunpacking.com. Related policy: Editorial policy.