Urgent audit: what ChatGPT and AI Overviews are saying about your locations (and how to fix it)

A client calls me on a Tuesday morning. 89 stores, and 60% of AI answers are wrong.

An optical retail chain with 89 locations. The marketing director reaches out: « Stéphane, we’re losing 150 calls a day. I don’t get it. »

I run the audit. In 12 minutes, I have 4 screens open. ChatGPT, Google AI Overview, Ask Maps, Perplexity.

I search: « optician open on Sunday with eye exam in Lille ».

Google returns one store. The right one. But the hours shown are 2 p.m.–6 p.m. In reality, the store opens at 10 a.m.

ChatGPT mentions a different location. The phone number displayed is from the old headquarters, closed since 2022.

I audit all 89 listings. 53 of them broadcast at least one error in AI responses. Incomplete address, outdated phone number, phantom closing day.

The problem wasn’t classic local SEO. It was what the AI had aggregated without anyone verifying it.

We audited. We cleaned. We put a monthly monitoring protocol in place.

+37% inbound calls in 9 weeks. No extra ad spend.

The 4 prompts of the urgent AI audit: what your brand must verify now

Annie Jackson and Jason Wertham from GatherUp created a four-question audit. I’ve been applying it for 3 months across 15 retail networks.

Copy these prompts and test them on 3 of your locations.

1. « [Brand name] [city] [service] » in Google to surface an AI Overview.
2. « What’s the best [service] in [city]? » in ChatGPT (or Perplexity).
3. « [Brand name] [city] phone » in Ask Maps or via Google voice assistant.
4. « [Brand name] [city] customer reviews » to check how AI summarizes your ratings.

What I observe on the ground: 8 out of 10 brands discover at least one wrong number in the first 10 results. And 4 out of 10 find that a lower-rated competitor appears as the top AI recommendation because its name matches the query better.

« A car wash rated 3.3 stars won the AI response for ‘touchless SUV wash in Norfolk, VA’ because Google prioritized query match over rating. » — Annie Jackson, Director of Revenue Operations, GatherUp

The audit takes 18 minutes. One uncorrected error? Thousands of dollars in customers who go elsewhere.

Why a 3.3-star rating can win the AI response

The logic of local AI isn’t yours. It wasn’t even Google Maps’ two years ago.

AI assembles:

• Query match (exact keywords in name, category, reviews)
• Freshness and volume of contributions (reviews, photos, posts, Q&A)
• Web signals (mentions in articles, forums, social networks)

The average rating becomes secondary. What AI captures first is the content’s ability to answer the specific question precisely.

I replicated the test for a network of 34 hair salons. On the query « natural dye hair coloring open Saturday in Bordeaux, » Google AI Overview mentioned a salon rated 3.7 with detailed descriptions of natural dye coloring. The 4.8-star salon 200 meters away didn’t appear. Its content only mentioned « hair care ». No mention of « natural ».

In the AI economy, content precision trumps reputation.

Keep the reviews, but add strong textual signals: rich service descriptions, Q&A that mirrors exact customer vocabulary, Google posts documenting each specialty.

From 300 million listings to a single answer: how AI decides

Google manages over 300 million business listings and 500 million review contributors. ChatGPT, Perplexity, and Claude scrape the web, listings, and aggregators.

Result: AI doesn’t cite one source. It fuses them.

I’ve seen ChatGPT quote hours from an unupdated Facebook listing from 2021. I’ve seen Gemini invent a phone number by cross-referencing two orphaned pages from an old website.

Response quality rests on three pillars:

1. Structured data accuracy: Schema.org LocalBusiness, sameAs, openingHours. Any inconsistency between your site and Google listing triggers an error.
2. Consistency across aggregators: Yelp, PagesJaunes, Apple Maps, Bing Places. One divergent number across them can corrupt the entire aggregation chain.
3. Fresh mentions: local news, blog posts, forum citations. A 3-month-old update outweighs a listing updated 2 years ago.

Auditing means mapping these 3 pillars, not just checking Google My Business.

L’article détaille trois actions clés pour passer d’une correction d’urgence à une défense continue. Ce diagramme résume le processus à suivre après votre audit AI.

Fixing isn’t enough. You must defend.

The GatherUp framework rests on three verbs: Build, Manage, Defend.

Build = give AI reliable, complete data. Long descriptions with attributes your customers search vocally. Q&A that anticipates 20 phrasings. One page per service per location, indexed, dated.

Manage = monitor what AI says. Re-run the prompts monthly. Track variations. Spot misinformation spikes.

Defend = react fast when wrong info sticks. Flag the error on Google listing. Update the faulty aggregator. Publish fresh content that contradicts the old data.

For the optical chain mentioned above, we defended 53 listings. We republished 89 Google posts in 15 days. We added 340 Q&A across the entire network.

Result: 100% of verified AI responses showed accurate data within 60 days. Local organic traffic jumped +42%. And the call-to-booking conversion rate increased +22%.

+37% calls with zero ads. Another network, another result.

Second client: 46 real estate agencies. The AI audit triggered by a similar call: « Stéphane, our leads are down 18% in four months. »

In 20 minutes, I find that 32 of 46 agencies have a closing time offset in AI Overviews. AI says « closes at 6 p.m. », actual close is 7 p.m.

Google AI Overview also mentioned a competitor rated 2.9 stars because its local tagline contained « free estimate within 24 hours »—exactly the search phrase agents were using.

We aligned structured data. We created per-agency FAQ. We submitted corrections to 7 aggregators.

Three months later, +37% inbound calls. Zero extra ad spend. Just AI responses finally telling the truth.

These numbers, I see them repeat. Fixing AI errors produces an immediate effect on calls and visits. Not in 6 months. Often within 30 days.

And what’s Google saying about your brand?

One more number: 48% of your potential customers have already asked ChatGPT about a local business (GatherUp, fall 2025). This isn’t a trend anymore. It’s an invisible layer of traffic.

You have a Google listing. A website. Reviews. But have you verified what the AI says when someone asks:

• « What’s the best [your profession] in [your city]? »
• « Is [your brand] open on Sunday? »
• « What’s the phone number for [your brand]? »

If the answer contains an error, you don’t lose a click. You lose a customer who’ll never know they got it wrong.

My job is building systems that run without me. An AI monitoring cockpit for your 10, 50, 100 locations. Alerts. Automated fixes.

But the first step is the audit. In 18 minutes. With these 4 prompts.

Stéphane Jambu

SEO & AI Engineer

I build growth systems / AI / Neuroscience | 650+ clients · 80 LinkedIn testimonials · 30 years of expertise · 15 years of systems running without me.

Follow on LinkedIn