The short version
Open ChatGPT, Claude, or Perplexity and ask it the same five questions a customer would ask before hiring your competitor: "Is [competitor] open on Sundays?" "What does [competitor] charge for a basic tune-up?" "Where is [competitor] located?" "Does [competitor] take walk-ins?" "Is [competitor] better than [you]?" Write down what the AI says, then check it against the competitor's actual website and Google Business Profile.
You're not looking for flattery or trash talk. You're looking for gaps between what's true and what the AI believes — because those gaps are where a customer gets sent to voicemail, shows up to a closed door, or picks someone else based on a wrong price. Every gap you find in a competitor's AI presence is either a customer they're about to lose, or a customer you could pick up instead.
Why this works better than a manual site review
A normal competitive audit means reading their website, checking their reviews, maybe calling to ask about pricing. That tells you what they've published. It doesn't tell you what AI assistants are actually repeating back to people who ask.
Those are two different things. A dentist's office might have accurate hours on their homepage and a completely different set of hours on an outdated directory listing that ChatGPT happens to trust more. A plumber might have raised their rates in March, but if their price page still isn't indexed cleanly, the AI is quoting last year's number. You can't see any of that by reading their site — you have to ask the AI directly, the same way a prospective customer would.
What to actually test
Run each competitor through the same short list:
- Hours and location. Ask for both, then verify against their Google Business Profile.
- Pricing or starting rates. Ask "how much does [competitor] charge for X." If the AI gives a number, check where it might have come from — often a stale directory or a review from years ago.
- Services offered. Ask if they do the specific thing you specialize in. If the AI says no or hedges, that's a service the competitor isn't getting credit for.
- Comparison questions. Ask "[competitor] vs [you] for [service]" and read how each is described. Note which one gets named first, which gets more detail, and which gets a caveat like "call to confirm."
Do this in at least two assistants — ChatGPT and Claude tend to draw on different sources, and Perplexity will show you citations directly, which is the fastest way to find out where a wrong answer is coming from. For more on why the same question gets different answers across tools, see why AI gets your business info wrong.
Reading the results
A local bakery owner we'll call Dana ran this check on the two other bakeries in her town. One had listed its wedding-cake minimum order on its website but never on its Google Business Profile — ChatGPT told a test prompt the bakery "doesn't appear to offer wedding cakes." That's a real customer, mid-search, getting steered away from a business that actually wants that work. The other bakery had a phone number that had changed eighteen months earlier; the AI was still repeating the old one, sourced from a directory neither bakery owner knew existed.
Neither of those is a strategy failure. They're plumbing failures — information that exists somewhere but isn't reaching the sources AI trusts. That's useful to know about a competitor, but it's more useful to check whether it's happening to you too.
Turning it into your own advantage
Once you've found where a competitor is weak, don't copy their gap — close it before they do, and make sure your own presence doesn't have the same hole. The three usual suspects are:
- Directory drift — old hours, phone numbers, or addresses sitting on a listing you forgot about, still getting cited months or years later.
- Missing structured data — no schema markup telling AI systems your services, hours, and pricing in a format they can parse cleanly. See LocalBusiness schema for local businesses for what that actually looks like in practice.
- Thin or contradictory review language — reviews are a bigger input to AI recommendations than most owners realize, and conflicting details across reviews can confuse the summary an AI gives.
The fastest way to see where you stand next to that gap is to run your own site through a free scan — it grades the same kind of accuracy issues you just went hunting for in a competitor, and shows you exactly what's wrong before a customer finds out the hard way. If the fixes go beyond what you have time for, the paid plan handles the correction work directly.
Do this quarterly, not once. Directories update, AI models refresh their sources, and a gap you closed in January can reopen by summer. The businesses that stay ahead aren't the ones with the best answer today — they're the ones who keep checking.