If you asked ChatGPT about your client's business tomorrow, would it get the details right?
Most agencies assume the bigger risk is invisibility: a business that AI simply doesn't know exists. Our latest research suggests the more common problem is quieter, and arguably more damaging. AI usually does know the business. It's just not always right about it.
The headline number: 94% awareness, barely half accuracy
In our AI Visibility Report, we analysed 10,000 US local businesses using the Insites 360-degree audit platform, asking ChatGPT and Perplexity directly about each one. The awareness numbers looked reassuring at first glance: ChatGPT identified 93.59% of businesses, and Perplexity found 93.15%.
That's the number most people stop at. It's also the wrong place to stop.
When we checked whether the details AI provided actually matched the business's Google Business Profile, the picture changed. Only 55.58% of ChatGPT results, and just 48.05% of Perplexity results, fully matched. Put another way, for roughly half of the businesses AI claims to "know", it's getting something wrong.
Awareness and accuracy are two different questions, and only one of them is being asked by most agencies right now.
What actually goes wrong
We broke the mismatches down by data point. Here's how the errors split, for businesses that AI did find:
| Result | ChatGPT | Perplexity |
|---|---|---|
| All details match | 55.58% | 48.05% |
| Wrong phone number | 30.09% | 38.81% |
| Wrong business name | 6.38% | 5.18% |
| Wrong website | 15.59% | 16.9% |
| No details match (likely hallucination) | 1.68% | 1.04% |
Wrong phone numbers are the single biggest source of error, by some distance. Some of this is likely explained by call-tracking numbers, which are correct from the business's perspective but won't match a "canonical" number, so the real error rate may be a little lower than it looks. But wrong websites and wrong business names are harder to explain away, and in around 1 in 60 cases for ChatGPT, the business AI surfaced didn't match the target business on any detail at all. That's not an outdated field. That's AI confidently describing the wrong business, or a direct competitor, as if it were yours.
Why we used Google Business Profile as the benchmark
To judge "correct", we needed a reference point, and we used Google Business Profile (GBP): name, address, phone number, hours, category and reviews. It's the closest thing local business data has to a canonical, owner-verified source at scale.
It isn't perfect. Plenty of small businesses don't keep their GBP fully up to date, and where that's true, a "mismatch" may reflect a stale GBP listing rather than an AI error. But even accounting for that, a roughly 50/50 split between "matches" and "doesn't match" is a meaningful gap, not measurement noise.
Why this matters more than the awareness stat
Being found and being described accurately are not the same thing, and the second one carries real commercial risk. A wrong phone number sends an enquiry nowhere. A wrong website sends a customer to a competitor. A full hallucination actively recommends the wrong business under your client's name.
This also reframes what an "AI visibility win" looks like. A business that AI has heard of but describes incorrectly isn't visible in any useful sense, it's mis-visible. For an agency reporting on AI visibility, "does ChatGPT know us?" is the wrong question to lead with. "Does ChatGPT know us correctly?" is the one that actually protects the client.
What to check before anything else
If you're auditing a client's AI visibility, accuracy comes before optimisation:
- Ask ChatGPT and Perplexity directly about the business, by name and location, and compare the answer against the current Google Business Profile
- Check specifically for phone number, business name and website mismatches, since these are the three most common failure points
- Where a mismatch shows up, check whether the GBP listing itself is out of date before assuming it's an AI error
- Flag anything that looks like a full hallucination immediately. That's not a visibility gap, it's active misdirection
Awareness is the easy number to report. Accuracy is the one that actually protects a client's business.
This analysis is drawn from Insites' AI Visibility Report, based on a clean audit of 10,000 US local businesses across ChatGPT and Perplexity. Run a free AI visibility check on any business at insites.com/ai-audit.
Andrew is the co-founder and CEO of Insites, the AI visibility platform that agencies, telcos and media companies use to show local businesses how search engines and AI tools like ChatGPT and Gemini find, trust and recommend them. He has spent more than 15 years in digital marketing and UX, and co-hosts The Unusable Podcast on UX design.
