The publication / editorial policy
Evidence before enthusiasm.
The rules behind every AI Tech Bench guide, recommendation, and correction.
Research before writing
We begin with a reader task and current sources. Official documentation, source repositories, release notes, and original evidence take priority over recycled summaries. A source must support the claim beside it; a keyword match is not enough.
Version-sensitive advice is checked again before publication. We keep source titles, publishers, URLs, access dates, and the claims they support in the editorial record.
No borrowed authority
We do not fabricate tests, benchmarks, author qualifications, clients, or experience. If a capability has only been researched, we describe it as source-verified. “Tested” means the test occurred and its evidence was saved.
AI assistance can support research, drafting, and editing. The named author remains responsible for verification, original examples, clear explanations, and the final article.
Publication gates
A full guide needs useful body content, a real byline, sources, accurate metadata, meaningful image descriptions, and a quality review. Draft plans never appear as published guides. Empty topic pages are excluded from indexing.
Our internal 100-point review covers verification, usefulness, sources, visuals, examples, accuracy, limitations, readability, SEO, and transparency. An article must score at least 85 to proceed, and a score does not replace owner review.
Commercial relationships and corrections
Commercial relationships must be disclosed close to the relevant recommendation. Commissions must not determine the conclusion. Sponsored links receive an appropriate sponsored relationship.
Send correction requests to contact@aitechbench.online, using the email options on the contact page. Include the guide URL, the claim and a primary source where possible. Changes that affect the advice receive a substantive update date; readers should be able to distinguish a correction from a redesigned page.