Editorial standards & corrections

Principles

This site exists to make a fast-moving, high-stakes field legible to the people who have to make decisions in it. That mission only works if the content is trustworthy. Three commitments govern everything published here:

• Evidence over assertion. If a claim depends on a number, that number comes from testing I can show you.
• Independence. I hold no equity in the vendors I review and accept no payment for favorable coverage.
• Transparency. Methodology, disclosures, and corrections are public - not buried.

Research & testing

Every benchmark follows a documented, reproducible protocol. Tools are evaluated against an identical held-out dataset spanning multiple registers and languages, with a human-written control set used specifically to measure false positives. Tests are re-run and re-published quarterly so figures never silently go stale.

If a piece of content could be written by anyone with a search engine and thirty minutes, it doesn't belong on this site.

Fact-checking

Before publication, each article is checked against primary sources - vendor documentation, peer-reviewed research, and original test data. Claims are linked to their source wherever possible. Every article carries a visible published date and a last fact-checked date, so you always know how current the information is.

Disclosures

Some tool links are affiliate links. When they are, they're explicitly labeled, and they never influence a score or recommendation - rankings come from the benchmark, full stop. Sponsored research, if accepted, is disclosed prominently at the top of the relevant page. Where AI assistance is used in production, it's disclosed (see below).

Use of AI in producing this site

It would be strange for a site about AI authenticity to be coy about its own use of AI. I use AI tools for limited drafting support, transcription, and code - never to generate the analysis, the testing data, or the conclusions, which are mine. All published text is reviewed, edited, and fact-checked by me before it goes live.

Corrections policy

When I get something wrong, I fix it visibly. Corrections aren't edited away silently - they're noted. The policy:

• Substantive errors (anything affecting a conclusion) are corrected promptly, with a dated correction note appended to the article.
• Minor errors (typos, broken links) are fixed without a note.
• Updates from new data are dated and, where they change a finding, flagged.

To report an error, email corrections@lisabraswick.com. I aim to respond within two business days.