Listen to the podcast episode
Or listen now on Apple Podcasts or Spotify.
Key Takeaways
Most GEO advice is snake oil — llms.txt included: Nothing in OpenAI’s, Google’s, or Anthropic’s documentation says their crawlers use llms.txt, yet it gets sold as a GEO best practice. Paul added one a year ago after spotting something similar on ESPN’s site and saw no difference. It’s harmless, but nobody has five years of LLM-optimization experience — the field is too new — so fact-check anyone selling certainty here.
Optimize for two numbers: citations and brand mentions: Strip away the checklists and a GEO audit comes down to how often LLMs cite your site and mention your brand. If the work you’re being sold doesn’t move or even measure those two KPIs, it’s focused on the wrong thing.
Start with guest personas, not tactics: Step zero is deciding what you want to be recommended for — family-friendly, romantic, pet-friendly, large groups. You can’t write test prompts without knowing which guests you serve, and this exercise regularly exposes that a manager can’t actually describe their target guest’s questions. It’s the keyword research of this era.
Baseline your mentions before you change anything: Build 10-20 realistic prompts — a homeowner asking for the best property manager in Destin, a family wanting a cabin within three hours of Atlanta — and run them repeatedly through a prompt-tracking tool like Gumshoe across ChatGPT, Gemini, and Claude. Mentioned 85 times out of 100 runs, you’re in good shape; twice out of 100, you’ve found the work. Answers vary run to run, so never judge from a single response.
Don’t lock the front door on AI crawlers: Wanting ChatGPT to recommend you while blocking the OpenAI crawler is cooking a beautiful meal and locking the restaurant’s front door. Verify the major AI bots can reach your site, and watch for subtler blockers: one client’s authentication wall and a server that throttled bots until they quit crawling.
Put a real human on every piece of content: The generic vacation rental content team byline doesn’t cut it anymore. Blog posts need a named author with a bio explaining why they’re a local expert — lived here ten years, raising kids here — and your About page should carry real business history and team info. Expertise, authority, and trust still matter to LLMs, and since there’s no official fact-checker, you get to tell your own story. Being found is different from being chosen.
Build collection pages for specific guest types: Family-friendly cabins, pet-friendly properties, romantic getaways — these are the pages LLMs lean on when they recommend, mirroring classic SEO’s amenity and travel-type queries. They double as better landing pages for high-intent ad traffic too.
Re-test after 30-90 days and expect lag: Models retrain on their own schedule, much like the old Google penalty era when you fixed everything and waited months for the algorithm to rerun. Hold your prompt set constant, re-run it after a month or a quarter, and compare mention rates. Nobody knows the exact lag yet, so track on a regular cadence instead of reacting to any single check.
Expect influence, not referral traffic: A guest can workshop a trip in ChatGPT, get pointed at Key West, then Google Key West vacation rentals and book — Google gets all the credit for an LLM-born idea. One client planned an entire wine-country trip to Niagara Falls that way without clicking a single ChatGPT link. Make your how-did-you-hear-about-us question multi-select — mentions matter even when clicks never show.
Keep it in proportion — this is 1-3% of your traffic: Get your classic Google SEO to a B-plus before spending real money here, because Google still drives the overwhelming share of bookings. The good news: everything in this audit — real authors, crawlable site, guest-type pages — strengthens traditional SEO anyway. It’s all one equation, and anyone charging separately for an LLM audit and an SEO audit is charging too much.
What We Cover In This Episode
Conrad and Paul share the first draft of their GEO/LLM search audit for vacation rental websites: baselining brand mentions with prompt tracking, opening the door to AI crawlers, humanizing content with real author bios, and building the guest-type pages LLMs recommend from — while calling out the snake oil in the space.
