What changes the moment you go from mentioned to shortlisted.

What narrows a long list down to a short one

Say a countryside hotel gets mentioned when an AI agent is asked about weekend breaks near a particular town. Being mentioned took relatively little: existing, being roughly in the right place, offering roughly the right kind of stay. Making the actual shortlist, the two or three options an AI puts directly in front of the person asking, is a narrower, harder test, and most of what decides it never shows up in the original, broader search.

The gap between the two is worth understanding precisely, because a hotel that's regularly mentioned but never shortlisted has a specific, fixable problem, not a vague visibility one. Mentioned means the AI found enough to include it in a long list of plausible options. Shortlisted means the AI found enough to be confident recommending it specifically, over the other plausible options, for this particular request.

Fit to the specific request, not the general category. A hotel that's a fine general match for "weekend break near this town" but hasn't said anything about whether it suits a quiet anniversary trip specifically, versus a lively group weekend, gives an AI nothing to narrow on when the actual request is more specific than the category. The hotels that make shortlists tend to be the ones whose facts answer the specific version of the question, not just the general one.

Evidence density matters more here than at the mention stage

Being mentioned might only need one or two solid facts, a location, a price range. Being shortlisted tends to need several reinforcing ones: the specific room type suited to the occasion, a confirmed detail about what makes it good for exactly this kind of trip, something that distinguishes it from the other three hotels in the same town with similar general facts. An AI narrowing a list is looking for reasons to prefer one option specifically, and a hotel with only general facts gives it nothing to prefer.

The same hotel can be shortlisted for one trip and not another

Say a countryside hotel has said, specifically, that its garden rooms sit away from the bar and stay quiet after nine. Asked for somewhere quiet for an anniversary, that one fact is enough to shortlist it directly. Asked for somewhere for a lively stag weekend the same month, that identical fact, unchanged, now works against it: an AI narrowing on "lively" has nothing in the hotel's own facts to prefer it for, even though nothing about the place has changed. Being shortlisted was never a fixed property of the hotel. It's a property of how well its specific facts answer the specific request in front of it, which shifts from one request to the next even when the hotel doesn't.

Why chasing more mentions doesn't fix this

A hotel trying to improve its odds by broadening its own description, appealing to more kinds of trips at once, usually makes the shortlisting problem worse, not better, because vaguer facts fit more categories loosely rather than fitting any one category precisely. The fix runs the opposite direction: sharper, more specific facts about exactly what kind of trip the hotel suits, even if that means being explicitly less suited to trips it doesn't. That's the layer Selfe exists to build: turning general facts into the specific, evidenced ones that let an AI narrow confidently, so being mentioned stops being the end of the road and starts being the entry point to the list that gets shown.

How do we know if we're stuck at mentioned rather than shortlisted?

Ask an AI a specific version of a request in your category and see whether you're named directly or only appear if asked for a longer list.

Won't being more specific rule out some customers?

It rules out customers it was never going to convert well anyway, and makes the business far easier to recommend confidently to the ones it's right for.