Somebody on your team asks whether you can get your docs to rank in ChatGPT. It's a fair question and the honest answer is unsatisfying: nobody can tell you exactly why one page gets pulled and another doesn't.
What you can influence is the second half, and it's the half that decides what your customer actually reads. Once a passage has been retrieved, something has to decide whether to use it, quote it, hedge around it, or quietly drop it.
What gets a passage used
It answers something on its own. A paragraph that completes a thought is usable. A paragraph that sets up the next one isn't, because the next one may not have come along.
It's specific. "Most plans support this" is unquotable — anything repeating it inherits the vagueness and has to hedge. "Available on Team and Enterprise" can be stated plainly, so it gets stated plainly.
It commits. Numbers, names, conditions, dates. Hedged writing produces hedged answers, and hedged answers don't help your customer or make you look like the authority.
It carries its own context. If the passage says "the dashboard", and nothing in it says which product's dashboard, then anything quoting it has to supply that itself, and it might supply the wrong one.
The uncomfortable bit
Most documentation is written to be read from the top, by someone who has already agreed to read the whole page. That reader has patience and context. They'll wait for you to get there.
A retrieved passage gets neither. It arrives alone, with no title above it and no next paragraph below it, and it has about a hundred words to be worth something.
So the writing that gets quoted isn't fancier writing. It's writing where every section was built to survive being read alone — which, conveniently, is also the version a person skimming on their phone gets more out of.
How to check yours
Take an article and ask any AI to answer three customer questions using only that article, flagging every point where it had to guess or hedge. The hedges are your unquotable passages, and they'll usually be the ones you thought were fine.
Our free skeptical-reader prompt does a version of this in about five minutes, and the hedges it finds are usually in the passages you were least worried about.
