Editorial policy

Every number on the site has a named, linked source in the same paragraph, with the date it was read. Two source classes are used and kept apart. Vendor statements come from the vendor's own product, pricing, help and policy pages and are attributed in the sentence. Independent studies are peer-reviewed papers, published benchmarks and lab reports that describe a method and a dataset. The editors read the study itself before quoting a figure; a news report or summary is never the only citation.

The index uses published figures only. When a vendor publishes no value for a field, the card leaves that field out. Nothing is estimated, inferred from a screenshot or carried over from another site.

The index data file is the single record for every figure. A tool's price, limit, language count or study result is identical on its card, in every ranking that uses it and in the page markup. When two studies report different results for one tool, both appear with their sample sizes and dates. The site does not average them or pick one.

Each page shows an "Updated" date that equals the page's last change to a figure or sentence. Prices and plan limits carry the date they were checked. Recheck plans are recorded with a date, never as a standing promise: prices and free tiers on all cards are planned for recheck on 8 December 2026. The index version changes when a card or study is added; version 2026.11 was created on 8 November 2026 with 16 cards and 8 study records.

A vendor's accuracy or false positive statement is quoted as the exact sentence, with the URL and the date read, and attributed to the vendor. It is shown in its own column on ranking pages and never enters a sort alongside study results. A vendor-run report that describes a method and a dataset can be indexed as a study, labelled as a vendor report.

All 22 cards carry the same 14 fields from the same source classes and enter a ranking only where a published value qualifies them. No card is placed, promoted or held back by anything other than its published fields. Links to any vendor's product page are tagged the same way.

A ranking sorts one published field and states how many of the 22 tools qualified. The site publishes no star ratings, no editor score and no "best", "cheapest" or "most accurate" verdict. Where a score is shown, the page carries the caveat that no detector score should be the only evidence in a misconduct, hiring or publishing decision.

An error in a published figure is corrected on the page the day it is confirmed, and the change is written to the corrections log with the date, the page, what changed from what to what, and the source that prompted it. Reports go to editor@aidetectorindex.com; the contact page lists what to include.