Listen to the podcast episode
Or listen now on Apple Podcasts or Spotify.
Key Takeaways
AI can assist your ads, but it can’t run them: The official Meta ads connector can’t upload videos, and 60% of the hosts’ best-performing ads are videos. You still upload manually, then let AI rebuild the ads around the asset. Where it genuinely shines is work no human does efficiently, like reviewing 7,000 search terms against a landing page, so treat it as the assistant, not the operator.
Trust but verify every change AI pushes to an account: Paul has had budgets not change and campaigns not actually turn off after AI said they did. Take the extra fifteen seconds to confirm the change executed, especially anywhere money is on the table. If you skip the review, be ready to look the client in the eye and explain that you didn’t check the AI’s work.
AI hands you band-aids, so keep digging for root causes: When a broken URL showed up in a Google Ads account, Claude’s suggestion was to point the ad at the homepage until someone figured it out. The real cause was a client edit that silently renamed a WordPress slug, a two-minute fix once a human investigated. Let AI do the tireless checking, like scanning every ad URL daily, and keep a human in the loop for the actual diagnosis.
Never let AI approve its own spend changes: Treat it like pricing software: set the guardrails, don’t go under 220 a night, don’t go over 500, and let it work inside the window. The final authority stays with someone accountable, because the AI is not the one getting on the call to justify why the budget moved. Anything touching money gets human sign-off.
Stop clicking apply on every Google Ads recommendation: Send a brand-new advertiser into the recommendations tab with instructions to apply everything and the account would be a disaster. Conrad just took over an account where the owner applied nearly everything the Google rep suggested, and undoing the damage was the first order of business. Once you’re spending a few thousand a month, every recommendation needs a judgment layer before it goes live.
Bulk platform busywork is still clunky for AI: Computer-use agents editing PMS listings, OTA content, Google Business Profiles, or social schedulers fail at logins and multi-step clicks, and watching them work is often slower than doing it yourself. Before you believe any run-everything-through-AI pitch, open the connector’s permissions list and read what it actually can and can’t do. There might be 40 things it does and 90 it doesn’t.
Keep AI-generated property images out of your marketing: An AI-generated beach house is not your beach house, and running it as an ad or listing photo sets a guest expectation you cannot meet. Best case is a brutal review; worst case is getting banned from the platform. A stock photo at least captures something real, and in this business the property is the product, so ground every image in reality.
AI can’t tell you when your analytics data is wrong: It won’t catch a pixel installed twice, website bookings that don’t match the PMS, or the attribution holes left by iOS privacy changes. Location data is anonymized to broad regions, and Conrad’s phone reports him in Atlanta, six hours from where he actually lives, so don’t build ad targeting lists straight from analytics geography. Pull the source data yourself and let AI assist the interpretation.
Match the AI service level to what your guests pay: An AI check-in beats an overwhelmed clerk at a budget roadside motel, but nobody paying a thousand a night wants a bot in the Four Seasons lobby. The same logic applies to your marketing: DIY-with-AI is a real advantage for a small operator, while a scaling business should demand human judgment layered on top of the tools. The right person armed with the right AI is the superpower.
What We Cover In This Episode
Conrad and Paul catalog what AI still can’t do for vacation rental marketing: uploading Meta ad videos, reliably pushing account changes, root-causing broken URLs, producing honest creative, and validating analytics data. The through line is trust but verify, with a human staying accountable for every decision that matters.
