The difference between a mention and a sale.
Most of what decides whether a recommendation becomes a booking happens after the mention, not during it. Here's what sits in that gap.
4 min read
Checking what an AI says about you only tests the mention, never the gap after it
Say a countryside spa hotel gets recommended by an AI to someone planning a birthday weekend. The mention goes well: accurate, specific, genuinely well-matched. What happens in the next sixty seconds decides whether that becomes a booking or nothing at all, and almost none of it is about how good the mention was.
The AI, or the person reading its recommendation, needs to know whether the actual dates are free. Then the actual price for those dates, which might differ from a generic rate quoted on a page. Then whether the room type mentioned is the one that's really available, not a similar one. Then a way to hold or confirm the booking, there and then, rather than being sent off to call the hotel during opening hours. Every one of those is a separate small check, and a single one failing is enough to lose the booking, however good the original mention was.
A hotel checking what an AI says about it sees the mention, reads it, and if it's accurate, moves on satisfied, with no way to know the gap even exists. That check only ever tests the mention itself, not what happens in the sixty seconds after it. The only way to find the gap is to try to push a real booking through the whole path, dates, price, room, confirmation, and see exactly where it breaks.
What tends to break, in order
Dates rarely break: availability calendars are usually current, because a hotel loses money fast if they aren't. Price breaks more often, when a generic rate is quoted somewhere an AI can read but the real rate for those specific dates lives only inside a separate booking system. Room type breaks when a description mentions a suite that's one of three variants, and nothing distinguishes which one's really free. Confirmation breaks most often of all: even when everything else lines up, there's frequently no way to complete a hold or a booking without a phone call, which quietly ends the process for anyone who was hoping it wouldn't need one.
Why fixing this rarely means new work
None of the four checks, dates, price, room type, confirmation, are things a hotel doesn't already handle for a person calling the front desk. Staff already know today's real rate, already know which suite variant is free, already take a card and confirm a hold over the phone in under a minute. The gap isn't a missing capability. It's that none of it is reachable by anything other than a person on the other end of a phone call, at the specific hours someone's there to answer it, which an AI acting on someone's behalf at any hour simply can't work around.
Closing each of those gaps, one at a time, dates, price, room type, confirmation, connected to the booking system a hotel already runs rather than a new one, is what turns a good mention into a completed booking. That's the specific, unglamorous work Selfe exists to do: not making the mention better, but making sure nothing breaks in the sixty seconds after it. For what has to connect before those sixty seconds can go right, see how we think about booking as its own job.
How do we find out where our own gap is?
Try to push a real booking through end to end, dates, price, room, confirmation, the way an AI would, and see exactly where it stops.
Isn't a phone number enough as a fallback?
It's a fallback for a person. An AI acting on someone's behalf generally can't make a phone call, so a phone-only confirmation step is effectively a dead end for that path.
Being found is not the same as being bought from.
This is the distinction the rest of this hub, and most of this site, keeps coming back to. Being found by an AI is real progress. It is not, on its own, a booking, a sale, or a reason to stop.
Can an agent buy from you?
Being chat-visible and being buyable are different tests. Most shops pass the first without ever finding out whether they'd pass the second.
Agentic commerce for portfolios: one setup, every property.
A group running six cottages doesn't need to solve this problem six times. Getting one property right the hard way, and then repeating that work five more times, is usually the wrong approach entirely.