Editorial standards

This is how we decide what to publish, what counts as a source, how we check it, and how we keep pages up to date. We publish it because we do not name our writers and the topic touches health. We cannot ask you to trust our qualifications. We can show our rules and keep to them. If you find us breaking one, tell us. See corrections.

What we publish

Information for adults about heavy and compulsive use of AI companions and chatbots: guides to specific apps, explanations of research and reporting, pages for parents and partners, a free self-check, and a news archive. We are an information site. We do not give medical advice, diagnose, or treat. See the Medical Disclaimer.

Who writes and edits

The pages are written and edited by an unnamed editorial team (the editors). We do not list names or qualifications because none are claimed. No page is signed by a clinician. We have no medical review board. See editorial review.

How we source

What counts as a source, in order of weight:

1. Primary research and official documents: peer-reviewed studies, regulator and government pages, court filings, professional bodies (for example AACAP and the American Psychiatric Association), platform policies and terms.
2. Reputable news reporting, used for what happened, not for what it means.
3. Statements by the companies involved, labeled as their statements.
4. Forums and social media, only as illustration of what people say, and labeled that way. Never as evidence of how common something is.

Rules:

• Factual and numeric claims link to a source in the sentence or right after it. A page’s main sources are listed in its text. See sources for the sources we rely on most.
• If we cannot find a source for a claim, we remove the claim. We do not publish statistics we cannot trace.
• For lawsuits, studies and named people, we link to the primary document (the complaint, the paper, the regulator’s page) where one exists.
• We describe allegations in court filings as allegations, unless a court has decided.
• We say where evidence is thin, mixed or contested.
• Some guides include short illustrative stories. They are written by our team, not quotes from readers, and each one is labelled as a composite. Real reader quotes appear only with the person’s permission, and we say when we cannot confirm a quote.
• We do not publish methods of self-harm.

How we check facts

Before a page goes live:

1. Claims are matched to a source, and the source is opened and read, not just its summary.
2. Numbers are re-checked, and quotes are checked against the original text.
3. Facts about apps (prices, age rules, features, policy links) are checked against the app’s own pages on the date of review.
4. Crisis numbers on the page are checked against the official source.
5. For pages we have rewritten, we keep a private log of the sources checked.
6. The editors read the page before it goes live.

Health and mental health claims

• We say what research found, not what to do about your health.
• We do not say or imply that a page, a quiz or a guide diagnoses, treats or cures anything.
• We do not use words like “proven”, “clinically proven” or “evidence-based” for our own tools. Where a claim is about an outside study, we name the study and its limits.
• We put crisis resources where results or distress are discussed.
• Our own assessments are described in full, with their limits, on the methodology page.

Independence, advertising and affiliate links

• We earn money from sales of our own guides (on Gumroad). When we recommend our own guide, we say so.
• Some pages and emails include affiliate links. Where we may earn a commission if you sign up or buy, we say so next to the link.
• Commissions and payments never decide what a page says, how an app is described, or how it is ranked. Rankings and comparisons are not for sale.
• We do not run display ads or accept sponsored articles.
• If our interests ever conflict with the accuracy of a page, accuracy wins and we say so.

Comparison pages

Pages that compare apps give the criteria used and the date they were checked. A rating is our judgment against those criteria. It is not a clinical assessment of any product. If an app changes or a company tells us we are wrong, we recheck and correct.

How pages are updated and reviewed

• Evergreen pages show a “Reviewed” line only after we have re-opened the sources and re-checked the facts on the page.
• We aim to review platform guides and parent pages about every 6 months, comparison pages about every 3 months or sooner when an app changes materially, and the methodology and assessment pages once a year or whenever scoring changes.
• A typo fix or a design change is not a review and does not change the review date.
• News posts keep their publication date and say “facts as of” where things may have changed.
• Changes to the scoring are listed in the change log on the methodology page.

How AI tools are used

We use AI tools to help draft, edit and tidy text, and to catch repeated or awkward phrasing. We do not use AI as a source for facts, and we do not publish anything an AI tool wrote without a person checking it against the sources. A person reads and approves each page and is responsible for it. We also do not use AI tools to invent people, quotes, studies or reviews.

We think that is worth saying plainly, because this site is about how people relate to AI.

What we will not publish

• Claims we cannot source.
• Names, photographs or details of private people, or identifying details of a person in crisis, without a clear public-interest reason and a primary source.
• Instructions for self-harm.
• Fake testimonials, reviews or credentials.
• Pages that present our assessment as a diagnosis.

Corrections and complaints

If we got something wrong, tell us at info@theaiaddictioncenter.com. We will check, fix it, and list confirmed corrections on the corrections page.

Related

• About
• How our assessments work
• Sources
• Corrections