Epic Insights
Edition 35 · Sep 1, 2026
|
|
Four AIs, One Business, Four Different Answers |
|
|
Last week we asked Google's Gemini and Google's AI Overviews the same question about the same chiropractic clinic, in the same hour. Gemini named the clinic in every single answer. AI Overviews didn't mention it once. Same company, same question, same afternoon. |
|
|
The Big Story
We Ran the Same Question Past Four AIs. They Disagreed.Here's what we actually ran. One chiropractic clinic, four AI systems, same afternoon, ten questions a real patient might type. We use Local Falcon for this, which checks how often a business gets named in the answer across a grid of points around the office. The results don't agree with each other. Not a little. On the phrase "chiropractor [city]" Gemini named the clinic in every single answer. Google's AI Overviews named it in none. Ask what an adjustment costs and AI Overviews had them in about two thirds of answers while Google's own AI Mode had them in under one percent. On "chiropractor near me" ChatGPT never mentioned them at all, and AI Mode had them in almost half. So there's no such thing as your AI ranking. There are four of them, they move independently, and a vendor who shows you one number is showing you a quarter of the picture. But one line in that table did something the others didn't. "Find a chiropractor near [landmark] open today" scored 97, 99, 96 and 96 across all four systems. The long, specific, awkwardly-worded question everybody agreed on. The short generic one is where the clinic kept disappearing. Which is the least surprising finding in the world and the most useful. The pages that answer a real question, in the words somebody would actually say out loud, are the ones every system can use. That was true before any of this existed. It's just now measurable in four places at once. |
|
|
|
|
Your move: Pick one question a customer would actually ask out loud, in their words not yours. Ask it in ChatGPT, in Gemini, and in a Google search that returns an AI answer. Write down who gets named each time. |
|
|
Quick Hits
The rest of the week, in bite-sized pieces |
|
|
|
|
|
|
|
|
|
|
Ask Epic
"Somebody pitched us on getting optimized for ChatGPT. Is that a real thing we should be buying?"Real thing, yes. Something you should buy as a separate premium service, almost never. We've been building for AI search for about two years now, well before it turned into a sales pitch, and here's the uncomfortable part for the people selling it: most of the work is the same work. Clear pages that answer real questions. Accurate hours and services. Current reviews. A profile that says what you actually do. You already own that, and paying a premium to have it redone isn't a strategy. What's genuinely different is measurement. Until recently nobody could tell you whether an AI was naming your business or your competitor's, which is exactly why we run the scans in this edition. That's the part most vendors can't do. So when somebody pitches you, ask them to show you a number, for a specific platform, for a specific question your customers ask. If they can't measure it, they can't improve it, and you're paying for a guess. |
|
|
What Epic's doing for you this week
This week we're running AI visibility scans for managed clients across ChatGPT, Gemini and Google's two AI surfaces, because checking one and calling it done clearly doesn't work. Those results land in your Client Center. We're also setting holiday hours and services on Google Business Profiles ahead of Monday, and finishing security patching on the WordPress sites we maintain.
|
|
|
|
|
Thanks for reading! Got questions? Ideas? I'm all ears. Share your thoughts with me and help shape future editions of Epic Insights. — Kristy Freeman |
|
|
