Listen to the podcast episode
Or listen now on Apple Podcasts or Spotify.
Key Takeaways
Travelers plan trips with AI, then book the same old way: The pattern Conrad and Paul keep hearing: ’I planned my whole trip with ChatGPT’ — and then that same traveler went back to Google, Airbnb, Vrbo, or a direct website to actually book. LLMs added new planning behavior on top of the funnel; they haven’t replaced the booking step. The pie got bigger, but the booking path didn’t move.
Friction alone will keep bookings out of the LLMs: Booking inside ChatGPT means connecting a plugin, authenticating, signing in — and 90% of people will never do that, with maybe 2-3% of the rest actually using it. Google requires nothing: open a search bar, get results. The likelier future looks like Google Flights — the LLM recommends, and you still click out and book on the property’s own site.
LLMs are eating your blog traffic, not your booking traffic: One large Florida client with 2.1 million recent pageviews: blog content down 14% year over year, booking pages up 18%. Another site pulled 800,000 referrals from Google in 30 days — and 17 from ChatGPT. AI Overviews and in-app planning are absorbing ’things to do’ queries, but booking-intent traffic still flows the same direction. Check your own numbers before you panic.
Voice search already taught you this lesson: In 2017 you were told voice assistants would book vacation rentals. Google Glass and 3D TVs got the same breathless predictions. Some trends resurface a decade later in a new form; some never land. Nobody trusted a voice assistant with a $3,000 booking, and nobody is trusting an LLM with one anytime soon. Don’t rebuild your strategy around a behavior shift the data can’t find.
Swap commodity list content for experience content: Google is now drawing a line between commodity and non-commodity content, and ’top 10 things to do’ is nearly impossible to make unique. What LLMs surface is first-person experience: I went, I did this, here’s what happened. A distant writer can research a list; only someone local knows the pizza place next door is actually bad. Mine the real questions guests and homeowners ask, and turn one genuine piece into five.
Brand mentions and schema markup beat keyword tricks in LLM results: LLMs learn brands that are consistently talked about, discussed, and linked across the web — that’s the durable play, not another keyword-stuffed page. Paul already sees listings with deep vacation rental schema markup perform better organically, with LLM readability as the bonus. It isn’t mandatory today; treat it as non-negotiable before 2027.
Weight your effort 40/40/20 — and LLMs live inside the 20: Conrad’s breakdown: 40% of the battle is the property itself and its amenities, 40% is operations — cleaning, maintenance, communication, the reviews they produce — and the last 20% is promotion. LLM visibility is a sliver of that final 20%. If the property or the operations are weak, no amount of AI-search optimization saves you.
Fix conversion, speed, and email capture before chasing LLM traffic: Go all in on LLM visibility and the realistic prize is a couple hundred visits a month. Meanwhile a property page that takes five to ten seconds to load actively kills the direct bookings you already earn from ads, organic, and email. Obsessing over the 2% channel while the 98% has a leaky funnel is exactly backwards.
Build channels no LLM can intercept: Get every guest onto your email and SMS lists and give them a reason to book direct again — repeat-guest relationships don’t route through anyone’s model. Hotels showed how expensive it is to claw back distribution after surrendering it, and vacation rental managers don’t have a Hilton-scale loyalty program to dig themselves out with. Own the relationship now.
What We Cover In This Episode
Conrad and Paul push back on the doom narrative that ChatGPT and AI search will kill direct bookings. Working from real client data, they show where LLM trip planning actually sits in the funnel, which traffic it genuinely cannibalizes, and where your marketing effort should go instead.
