How We Research, Write, and Keep Every Article Honest

Writing about film, video, and the people who make it is easy to do badly. Plot summaries dressed as analysis, release dates copied from three other sites, and opinions delivered with no evidence are the default. This post sets out how meVid.online works instead, including where our process fails and what we do about it.

Where a topic comes from

Subjects enter the queue three ways. Reader demand is first: questions our assistant cannot answer from an existing page are logged, counted, and promoted when they recur. Coverage is second: each hub has a set of subjects it needs before it is genuinely useful to someone new to the area, and those gaps are filled on purpose rather than opportunistically. Currency is third: when a subject changes materially, the existing page is queued for rework rather than a second page being spun up beside it.

How a draft is built

Drafting is AI-assisted, disclosed on every page, and tightly constrained. The brief requires at least twelve distinct sections with genuinely different jobs: a definition, an origin, how the thing actually works, a comparison against neighbouring concepts, a timeline, the misconceptions that persist, practical implications, honest limits, and references. Twelve headings over one idea restated is not twelve sections, and drafts that do this are discarded before they reach a URL.

The quality gate

Nothing publishes without clearing a hard floor: 1,400 words of visible body copy and a minimum of two original visual elements. The word count is measured on rendered text, so headings, captions, and chrome do not pad it. The visuals are drawn from the article's own data rather than sourced from a stock library, which means a comparison table actually compares the things the paragraph above it discusses and a timeline reflects the dates in the text. Drafts that fall short are held, not shipped thin.

Human review

An AI-assisted draft is a starting point. Every page has a named editor who is accountable for what it claims. Review concentrates on the failure modes automation is worst at: confident sentences with nothing behind them, dates that have drifted, comparisons that are literally true and practically misleading, and definitions that quietly depart from how a field actually uses a term. Unsupportable claims are removed rather than hedged.

Articles are living, not finished

Once published, an article joins an evolution queue and is periodically reassessed. Is it still accurate, has the subject moved, are there now related pages worth linking, has it become thin relative to the rest of the library? Pages needing work are rewritten in place, the version increments, and the update date changes. This is why every article carries a version badge. A reference page written once and never revisited slowly becomes misinformation.

What we archive

We audit the library against today's standard rather than the standard when a page was written. When the floor rose, a substantial batch of older pages no longer met it. Those were withdrawn from circulation the same day rather than quietly left up, and they are being rewritten to the current standard in batches. Anything that still fails a rewrite stays unpublished. This costs traffic in the short term. It is still the right call, because a library is judged by its weakest page.

What we will not do

We do not publish pages whose only purpose is to intercept a search query. We do not spin one piece into several near-identical variants. We do not hide that automation is involved, and we do not pretend a machine wrote something a person shaped. We do not fabricate statistics, quotes, credentials, awards, or sources; a number that cannot be traced to something checkable does not appear on the page.

If a page falls short of any of this, tell us via the contact page. Corrections are made at the source, on the page, with the date shown.