We use AI tools in producing this site, and we would rather tell you exactly how than let you guess. Readers are entitled to know what a machine touched, what a person checked, and who is accountable when something is wrong.
The short version: AI helps us work. It does not decide what is true, and it does not carry the responsibility. A named person verifies every factual claim and stands behind every published article.
Where we use AI
Research organisation. Gathering documentation, release notes, and support pages on a topic and structuring them so a writer can work through them efficiently. The sources are still read by a person.
Outline exploration. Testing different ways to sequence a guide — which question to answer first, where a caveat belongs — before committing to a structure.
Draft review. A second pass looking for internal contradictions, steps referenced out of order, undefined jargon, missing prerequisites, and claims stated more confidently than the evidence supports. It behaves like a tireless proofreader with no ego, which is genuinely useful.
Transcription. Converting recorded walkthroughs and spoken notes into text a writer can edit.
Image concepts. Exploring visual treatments for illustrations and featured images. Where a published image is AI-generated, it is an illustration rather than evidence, and we do not present it as a photograph of a real screen or product.
Where we do not use AI
These are hard limits, not preferences:
Not as a source. An AI system asserting something is not evidence that it is true. Language models produce fluent, confident, wrong answers regularly. Nothing gets published on an AI's word — every claim traces back to primary documentation, a vendor statement, a standard, or a test we ran.
Not as a substitute for experience. We do not use AI to manufacture the impression that we tested hardware we never touched, used a service we never signed up to, or ran a procedure we never ran. If we could not verify something, the article says so.
Not to imitate anyone. We do not use AI to mimic a named writer's voice, reproduce another publication's article in different words, or generate fake reviews, testimonials, or bylines. Every author here is a real person.
Not to publish unreviewed. No AI-assisted draft reaches the site without a person reading it end to end and taking responsibility for it. There is no automated publishing pipeline that bypasses a human.
Not at scale for its own sake. We do not generate large volumes of near-identical articles to occupy search results. If a piece would not be worth a person's time to write, it is not worth your time to read.
Not with your data. Reader messages, contact-form submissions, personal information, credentials, private documents, and confidential material are not fed into AI systems to produce site content.
How verification actually works
For any factual claim in a guide:
- Primary source first. Official documentation, vendor support pages, changelogs, standards. Where we rely on another publication, we say so and link to it.
- Test where testing is possible. Steps are walked through on a current version and described as they appear. Where we cannot test — hardware we do not own, a region-locked service, an enterprise tier — the article says we could not, rather than guessing.
- Conditions stated. Version, region, device, carrier, and account tier get named wherever they change the answer.
- Second read where stakes are high. Guides touching security, privacy, payments, or official government services get reviewed by another person before publication. The workflow is documented in How We Test and Verify.
Anything that survives all four is publishable. Anything that does not gets cut or explicitly flagged as unverified.
Disclosure
Where AI assistance was material to how an article was produced, the article discloses it. "Material" means it shaped the substance — not that a spell-checker ran.
We think blanket disclaimers on every page are close to meaningless, so we would rather tell you the actual production practice here, in detail, and disclose specifically where it matters.
Why we are this explicit
The web is filling with sites that generate articles at volume and hope nobody notices. Those sites are why "written by AI" has become a warning label.
We are not going to persuade you with a badge. What we can do is describe the process precisely enough that you can check our work — sources are linked, limitations are stated, the author is named, and the correction route is one click away on the Corrections Policy page.
If you find something on this site that reads like it was published without a person checking it, report it through the Contact page. That is a process failure and we want to know.
Accountability
Whatever tools were involved, responsibility for every published article sits with the editor named on the Authors and Reviewers page. There is no version of this where "the AI got it wrong" is our answer to you.