Listen to the podcast episode
Or listen now on Apple Podcasts or Spotify.
Key Takeaways
The ’Google is dying’ narrative fails the traffic test: Conrad pulled live numbers on one of the highest-traffic sites he can see — 47 million sessions over 90 days. ChatGPT referred about 203,000 visits and was trending down; Google referred over 10 million and was growing around 5%. Across almost every account, AI referrals sit under 1-2% of traffic. People are supplementing Google with AI, not replacing it.
Don’t over-invest in AI visibility yet: A channel sending 1-2% of your visitors doesn’t deserve a rebuilt strategy. The sites gaining LLM mentions are the ones doing SEO fundamentals — technical health, schema, content that genuinely answers homeowner and traveler questions. Be aware of your AI visibility and discuss it with your agency, but keep the effort where the traffic actually is.
Prompt tracking is shakier than the dashboards suggest: SparkToro’s research found AI recommendations wildly inconsistent — even Apple only shows up in roughly 70% of headphone-recommendation runs, no brand hits 100%, and results swing 15-20 points between runs. Real users also phrase the same need in endlessly different prompts. Doing it properly means building personas, writing prompt variants, and running each about a hundred times. That’s not a weekly practice for 99.9% of managers — check occasionally, see which prompts surface competitors, and write content that answers the underlying questions.
Test AI visibility in the free tier, where 95% of users live: Only about 5% of ChatGPT users pay for it. If you’re checking how your brand shows up, run the checks in the free model, because that’s where nearly all real-world prompts happen. And with ads coming to ChatGPT at premium prices, expect a paid route into those same conversations soon.
Make your site legible to LLMs with proof, story, and questions: Conrad’s working framework has three pillars, and every one of them was good marketing before AI showed up. People ask LLMs full questions — closer to Ask Jeeves than keyword search — so structure content that answers them directly.
- Proof: explicit trust signals — guests hosted, properties managed, years in business, awards
- Story: an About page worth reading; guests checked it before AI, and now the bots read it too
- Questions: FAQ and Q&A blocks on property pages and blog posts that answer what people actually ask
Point agentic AI at your real constraints, not novelties: Conrad’s daily example: every evening an OpenClaw job pulls tomorrow’s meetings, grabs the last 30 days of email with each attendee, and delivers a three-bullet brief — 20 to 30 minutes of prep done, without opening the inbox and getting pulled into a new fire. Proactive, scheduled work is what standard chat interfaces lack; they do nothing until you show up. Skip the party tricks — an agent reordering your shampoo saves 17 seconds.
Adoption can go from zero to millions in weeks now: OpenClaw went from an unknown open-source project to the fastest software ever to a million GitHub stars, with millions of users within weeks. Nobody’s roadmap saw it coming a month out. Hold your AI predictions loosely and revisit them on a schedule — the hosts now do these check-ins every six months for exactly that reason.
People enjoy shopping for vacation rentals — AI won’t take the booking: Session recordings tell the story: ten-plus minutes, a hundred-plus clicks, most of it inside photo galleries. Browsing homes is legitimately fun, the way a treasure-hunt store is fun — that’s not a step travelers want automated away, any more than voice booking panned out. AI’s real travel use today is upstream research like destination and flight discovery. Invest in your galleries and direct booking experience; that’s where the decision happens.
Put someone in your company in charge of AI: Whether you’re doing a lot or a little, someone should own the tools, the experiments, and the standards — including where AI should and shouldn’t touch hospitality. Use it to make the guest and owner experience better, not just to strip people out. And don’t paste raw ChatGPT copy into your marketing: the flavor is obvious at a glance, and it makes your brand sound like everyone else’s.
What We Cover In This Episode
Conrad and Paul run their early-2026 AI checkpoint: what OpenClaw-style agents are actually good for, live traffic data on the ’Google is dying’ narrative, why prompt tracking wobbles under scrutiny, and the proof-story-questions framework for getting your brand mentioned by LLMs.
