AI search can't find your locations

On 17 July, Birdeye published “The Location Blind Spot in AI Search” — a scan of 16,240 individual locations belonging to more than 1,500 multi-location brands across 28 industries, measuring how each one shows up when someone asks ChatGPT for a recommendation nearby. The headline finding is that brand-level and location-level visibility are almost unrelated numbers, and that the corporate dashboard is measuring the wrong one.

This is not a DOOH story. We’re covering it because the venue side of place-based media is made of exactly these businesses — independent and small-chain salons, spas and clinics whose entire demand arrives through local discovery — and because it changes what a network operator can put on the table in a venue pitch.

What the study found

16,240 location-level ChatGPT scans across 1,500+ multi-location brands in 28 industries (Birdeye, 17 Jul 2026 — vendor study, method partially disclosed).
Mean location-level visibility score 29.9 vs brand-level 9.6 — individual locations are roughly 3× more visible than the brand-level score implies (Birdeye — directional).
18.6% of locations never appeared in any ChatGPT response (Birdeye — directional; the press release rounds this to “20% invisible”).
46.7% of brands showed a 50-point-or-larger visibility gap between their best and worst location (Birdeye — directional).
Accuracy: wrong opening hours 97% of the time, wrong website for 57% of locations; fewer than 1 in 100 locations returned a fully accurate profile (Birdeye — directional).
75% of citations came from third-party sources, not brand-owned pages (Birdeye — directional).
The read: the venue’s discoverability is now decided by data it doesn’t control — and a network operator is one of the few partners positioned to fix it (BDOOH analysis).

What the study found

The useful number here is not the invisibility rate. It’s the 97% wrong-hours figure, because it identifies the failure mode precisely: this is not a ranking problem, it’s a data-integrity problem. The model is answering; it’s answering with stale or conflicted facts pulled from directories, aggregators and review sites — which is what the 75%-third-party-citations line means. The brand’s own website is not the primary source of truth about the brand.

The second useful finding is the variance. Nearly half of brands had a 50-point spread between their best and worst location. Averages at the corporate level hide that completely, which is why the study’s framing — a blind spot rather than a deficit — is fair. A chain can look fine and have a third of its sites missing from the answers that matter.

Where the study is weakest is scope. One model (ChatGPT), one snapshot in time, a proprietary 0–100 scoring method, and no cross-check against whether visibility converts into visits. It also arrives attached to a product launch, which is not disqualifying but is worth naming.

What it means for beauty

The venue’s demand problem is the network’s opening line. A salon owner does not want to hear about programmatic supply. They want more chairs filled. Arriving with a diagnosis of their own discoverability — and a fix — is a materially better pitch than a revenue-share slide, and it maps directly onto the objections catalogued in will salons say yes and the mechanics of how to sign salons as venue partners.
It reframes what the screen is for in the venue’s eyes. The screen earns the venue money by carrying third-party ads, but it earns goodwill by carrying the venue’s own services, prices and bookings. That dual role is the argument in how to monetize your salon with screens, and it’s why a house-slot allocation belongs in the venue partnership agreement from day one.
Third-party sources decide the answer — so the network’s own directory counts. If 75% of citations come from outside the brand’s site, then a network that publishes accurate, structured venue pages becomes one of the sources the models read. That’s an underrated asset for an operator, and the same discipline that makes building a media kit that sells work for advertisers.
Attribution still has to close the loop. Visibility in an AI answer is a step earlier in the funnel than anything a screen does, and neither one is proven by the other. Our position on what can and cannot be claimed between a screen, a scan and a booking is unchanged and set out in QR and O2O attribution and QR and O2O attribution for beauty screens.
This is a cold-start lever, not a media metric. Signing the first venues is the hardest part of building a beauty network, and anything that makes the operator useful before the first advertiser exists is worth more than it looks — the sequencing problem in the cold-start problem.

The caveat that keeps us honest

This is a vendor study with a product attached. Birdeye sells AI-search visibility management and launched a location-recommendations feature alongside the release, which is the textbook condition for a study that finds a large, urgent problem. The visibility score is proprietary and not independently audited; the scan covers ChatGPT only, at one point in time, and model answers are non-deterministic — a re-run would not reproduce the same numbers. The 28 industries are not broken out, so there is no beauty-specific figure in this study at all: nothing here says salons are worse or better than the mean, and we’re not implying it. The beauty read-across is our analysis of who venue partners are, grounded in beauty venue base by country, not in Birdeye’s data.

Sources

Birdeye — “The Location Blind Spot in AI Search,” 17 Jul 2026 (press release via EIN Presswire; trade coverage via AiThority, “For Multi-Location Brands, 20% of Their Locations are Invisible on AI Search”): 16,240 location-level ChatGPT scans across 1,500+ multi-location brands in 28 industries; mean location-level score 29.9 vs brand-level 9.6; 18.6% of locations never surfaced; 46.7% of brands with a 50+ point best-to-worst gap; incorrect opening hours in 97% of cases and incorrect website for 57% of locations; fewer than 1 in 100 locations fully accurate; 75% of citations from third-party sources. Vendor study — directional; proprietary scoring, single model, single snapshot, released alongside a related product launch. No beauty vertical is broken out and no beauty-specific claim is asserted; venue-side implications are BDOOH analysis grounded in the linked Research and Guides.