How We Research, Write, and Keep Every Article Honest

A discovery site lives or dies on whether you can trust the page in front of you. This post explains exactly how articles on The Curator are made, what gets rejected, and what we do when we find our own work wanting.

Where a topic comes from

Three sources feed the queue. The first is you: questions readers ask our assistant that no existing page answers are logged and counted, and recurring ones are promoted. The second is coverage: every hub has a shelf of subjects it needs before it is actually useful to a newcomer, and we fill those gaps deliberately instead of chasing whatever is convenient. The third is currency: when a subject changes enough that an existing page misleads, that page is queued for rework rather than left beside a newer one.

How a draft is built

Drafting is AI-assisted and disclosed on every page. The brief is strict: at least twelve sections, each doing a different job. A definition. An origin. How it actually works. A comparison against the things it is most often confused with. A timeline. The misconceptions that refuse to die. Practical implications. Honest limits and open questions. References. Twelve headings restating one paragraph is not twelve sections, and drafts built that way are discarded before they get a URL.

The quality gate

Publication requires clearing a hard floor: 1,400 words of visible body copy and at least two original visual elements, such as a comparison table, a timeline track, a concept map, or a labelled statistics panel. Word count is measured on rendered text, so headings and captions cannot pad it. Figures are generated from the article's own data rather than pulled from a stock library, so the table compares the things the text discusses and the timeline matches the dates on the page. A draft that misses either threshold is held rather than published thin.

Human review

A draft is not an article. Every page carries a named editor accountable for its claims. Review targets what automated drafting gets wrong most often: authoritative-sounding sentences with nothing behind them, dates that have drifted, comparisons that are technically accurate and practically misleading, and definitions that quietly diverge from ordinary usage in the field. Claims that cannot be supported are cut, not softened.

Articles are living, not finished

Every published article joins an evolution queue and is reassessed on a cycle. Still accurate? Has the subject moved? Are there now sibling pages worth linking? Has it become thin against the current library? Pages needing work are rewritten in place, the version number increments, and the updated date changes — which is why each article shows a version badge. A reference page written once and never revisited degrades into misinformation on its own.

What we archive

We audit against today's standard, not the standard in force when a page was written. When our floor rose, a large batch of older pages stopped clearing it. Those were withdrawn from circulation immediately rather than left up quietly, and they are being rewritten to the current standard batch by batch. Anything that still fails after a rewrite stays unpublished. Pulling pages costs traffic in the short term; we do it anyway, because a library is judged by its worst page as much as its best.

What we will not do

We do not publish pages that exist purely to catch a search query. We do not spin a single piece into several near-duplicates aimed at variant phrasings. We do not conceal that automation is part of the process, and we do not dress human editing up as something it is not. We never invent statistics, quotes, credentials, awards, or sources; if a figure cannot be traced to something checkable, it does not go on the page.

Found something that falls short? Tell us on the contact page. Corrections are made at the source, on the page, with the date visible.