Tenacious Prompting — Testing AI Disruption on My Own Vacation Rentals
Tenacious Prompting: how I got AI to admit "I can't do that" twice — then prove Airbnb and Booking.com really can be disrupted by Meta Muse.
I own a couple of beach rentals. I wanted to know: could AI actually find a real discount on my own properties that the big platforms don't show?
First answer: it told me straight out — it could only search, and handed back the same big listings everyone already sees.
So I pushed further. I told it to build an agent that took those top results and searched for alternate sites carrying the exact same home, matched by name and description. That worked — it found every listing, direct site included. But it hit a wall again: it said the actual pricing was behind deeper click-throughs it couldn't reach.
So I pushed a third time. Dig deeper, try those click-throughs however you can. That's when it finally pulled real, live pricing across every site — including mine.
Three rounds. Three "I can't do that" moments. Three times I broke the ask down smaller and pushed back. That's Tenacious Prompting — not accepting the first no, or the second. It's often the actual skill in AI work: not knowing the perfect prompt up front, but knowing how to keep guiding it toward what you need.
Here's what it found: I mark my direct rate up 15% to cover what the platforms charge me to bring in that renter. But once it totaled everything — tax, cleaning fee, pool heat — the real savings a guest would see was closer to 8%. Fixed costs don't shrink just because the rent line does.
So yes, the disruption is real. But it took someone who understood the business, and wouldn't quit after the first refusal, to find it.
Even then, I doubt most renters switch over an 8% gap. People book through the big platforms for one reason beyond convenience: protection. They trust a big company has their back if something goes wrong.
That trust is worth something. During COVID, I refunded every guest in full — it hurt, but it was right. Plenty of hosts didn't, because the fine print let them keep the money.
AI didn't just find a pricing gap here. It proved that getting real answers out of it takes the same thing that gets anything done well: not quitting at the first "I can't."