Editorial policy

How we choose what to write about

We start from questions readers actually ask, not from keywords. The person who runs this site reads the incoming reader email and the search trends, and picks the topics. A topic is greenlit only when we believe (a) there is real confusion in the public conversation, (b) there is established research that clarifies it, and (c) we can say something useful that isn't already saturating the first page of search results.

How an article gets written

1. Source assembly.Before a single sentence of the article is drafted, a short bibliography is assembled of the books, papers and clinical frameworks the piece will rest on. That bibliography lives in the article's working file and survives to publication as the "Sources" section.
2. Draft. The article is drafted by AI in English against that bibliography, under the rules on this page. We say so plainly rather than implying a newsroom that does not exist; the detail is in our AI disclosure. Claims that go beyond common sense must be traceable to a source on the list.
3. The source gate. A claim that cannot be traced to a named, linkable source does not get published, however confident the draft sounded. That gate exists because the failure mode of AI-drafted text is not bad writing — it is plausible, invented numbers. Where a finding is contested in the literature, the article has to say so rather than pick a side in silence.
4. Publish.The piece goes live with the byline of the desk, publication date, the date it was last written or updated against its sources, and the "Sources" section visible at the bottom. That date is not a claim that anyone re-read the page that morning.

What we treat as established vs. suggestive

Relationship science is not physics. Some findings (the predictive power of contempt in Gottman's longitudinal work, the broad stability of attachment patterns across adulthood, the neurobiology of involuntary signals like pupil dilation) are well-replicated and widely accepted. Others (specific claims about texting frequency, single-study findings on attraction micro-signals) are interesting but only suggestive. We treat the first kind as fact and the second kind as "some research suggests" — and we'd rather lose a reader than oversell a study.

What we will not do

• Accept paid placements from dating apps, coaches, or course-sellers. Our only revenue is display advertising.
• Invent statistics. If we cite a number, it traces back to a source you can find on our sources page or in the article itself.
• Pretend clinical authority we don't have. We are writers and researchers, not therapists. If a reader's situation calls for professional support, we say so.
• Use sensationalised headlines that the body of the article doesn't deliver on. The headline must be a fair description of what the reader will actually find inside.

Updates and revisions

Articles are reviewed at least once a year and updated when new research changes our reading of a topic, when a reader points out an error, or when we simply find a clearer way to say something. Material changes are logged on our corrections page.

Use of AI tools

The prose here is drafted by AI. What a person decides is what this site covers, the rules on this page, and where the line sits on a contested finding. What holds a claim up is not our word for it: every research figure links to the study it came from, and you can open it. For the full picture, see our AI disclosure.

A correction to this page

An earlier version of this page said a human writer drafted each article, that a second editor reviewed it, and that every factual claim was verified by a human editor. None of that was accurate — the drafting is AI, as the AI disclosure has said all along — and we corrected it in September 2026. A page setting out the rules is the last place that should carry a claim that does not hold.