AI Observer reports on AI tools, the agencies that build with them and the people behind both. This page explains the rules we follow, so readers can judge our work and tell us when we get it wrong. It applies to everything we publish, on this site, in our newsletter and on social media.
What we cover and why
We cover what helps readers decide, build or buy. A story earns its place if it tells a working team something useful: what changed, what it costs, who it affects, what can go wrong and what to do next. We do not publish stories only because a company wants attention.
Sources and evidence
Every news story needs at least one primary source. That means a company statement, a filing, a research paper, an official document or a person who spoke to us on the record. We link our sources, usually the first time a fact appears and again in a list at the end of the story.
- When we cannot confirm a claim, we say so in the text.
- We do not make a claim sound more certain than its source. If a study says a tool “may reduce” errors, we do not write that it eliminates them.
- Every number has a source. If a number is only an example, we say so.
- We prefer the original document to coverage of it. When we rely on another outlet’s reporting, we name and link it.
- Benchmark and performance claims made by a company are reported as the company’s claims unless an independent source has checked them.
How we use AI in our own work
Our editorial team uses AI tools to research, draft and illustrate stories. These tools help us find and read sources, write first drafts, check style and produce images.
AI tools make mistakes, so we do not treat their output as a source. Facts in our stories must come from the linked sources, not from what a model says. We do not use AI to invent people, quotes, companies, scores, dates or numbers. If something is a sample, we label it “Sample”.
If a story relies on AI output in a way readers would want to know about, for example a test we ran with a chatbot, we say so in the story.
How stories are checked
Stories are published without a separate sign-off step before they go live. We rely on these safeguards instead:
- Source requirements: a news story without a linked primary source does not publish, and a story with a missing fact is held until the fact is found.
- Automated checks: every draft is checked against our style rules and our checklist before publishing, including sources, headline and summary length, headings and image labels.
- Daily editor review: each weekday, our editor reviews the stories published that day, using AI tools to check them against their sources, and fixes problems.
- Fast corrections: when we find or are told about an error, we correct it on the article with a dated note. See our corrections policy.
Images
Images on our stories are AI-generated illustrations, and each one is labeled “AI-generated image” in its caption. They illustrate a theme. They are not photographs of real events or people, and we do not create fake screenshots or fake product interfaces. If we use a real photo or screenshot, we credit its source.
Story formats
Each story carries one format label so you know what you are reading:
- News: something that happened, with a date and a source.
- Explainer: how something works, or how two things differ.
- Guide: practical advice that is short of a full tutorial.
- Analysis: our reading of what something means, clearly marked as such.
- Opinion: a named author’s view.
Tutorials list the tools used, the level and the time they take, and include a section on what went wrong. We only publish steps that were actually run.
Interviews
Interviews are with real people, who know they are speaking to us for publication. Conversations are on the record unless we agree otherwise before they start. We edit answers for length and clarity, never for meaning, and pull quotes are quoted exactly. We do not let interviewees approve the finished story, though we may check facts and quotes with them.
How agency rankings work
Our agency lists are based on published criteria: services offered, evidence of delivered work, client references and pricing clarity. Each agency’s score, from 0.0 to 10.0, is calculated only from these criteria. Facts we could not check, such as team size or typical price, are marked “provided by the agency”.
Agencies cannot pay to change their position. If a listing includes a paid profile or a partnership, we label it next to the entry. We do not rank agencies that our owners have a stake in.
What we ask every agency
- Who does the work, and who maintains it afterwards?
- What does a first project cost and how long does it take?
- Can we speak to two clients about the result?
Advertising and sponsorship
Advertisers do not decide what we cover, and they do not see or approve stories before they are published. Sponsored content is labeled “Sponsored” at the top and is never presented as news. Paid agency listings are labeled next to the entry. If we ever use affiliate links, the article will say so. More detail is on our advertise with us page.
Conflicts of interest
Writers do not cover companies they work for or hold shares in. We do not accept payment, gifts or free services in exchange for coverage. Our owners may run businesses that work with AI. When one of them is mentioned in a story, we state the connection in that story.
Corrections and updates
If we publish something wrong, we correct it at the top of the article with the date and a short description of the change. When new facts change a story, we show it as updated. Fixing a typo does not count as an update. The full process is on our corrections page. To report an error, write to corrections@aiobserver.news. We aim to reply within two working days.
Feedback
If you think we have broken one of these rules, tell us at editors@aiobserver.news. We review this policy as the site and the tools we use change, and we date each revision below.
Last updated: 30 October 2026.