Blog
What the top-3 actually have on their backlink profile.
We pulled backlink data on the top-3 Google local pack businesses in 200 US markets — 25 cities crossed with 8 service verticals. Then, for each of the same 200 markets, we asked ChatGPT the same question: "Name the top 3 [vertical] in [city]."
The two answers disagreed almost completely (internal benchmark, Apr 2026).
ChatGPT names the wrong top-3 three out of four times
Across 200 local markets, ChatGPT's response to "name the top 3" matched Google's actual pack like this:
0 / 3 correct: 146 markets (73%)
1 / 3 correct: 40 markets (20%)
2 / 3 correct: 13 markets (6.5%)
3 / 3 correct: 1 market (0.5%)
Three out of four markets, ChatGPT cannot name a single business that Google considers a top-3 local result. If you have been paying for "AI search optimization" on the theory that AI citations and Google pack position are the same thing — they are not. They are barely correlated.
The vertical matters more than anything else
ChatGPT's hit rate varies enormously by the type of business:
Dental: 0.76 / 3 on average
HVAC: 0.68 / 3
Plumbing: 0.60 / 3
Insurance: 0.32 / 3
Legal: 0.16 / 3
Real Estate: 0.12 / 3
Auto Repair: 0.08 / 3
Restaurants: 0.04 / 3
In restaurants, ChatGPT and Google almost never agree. ChatGPT is citing food critics, editorial lists, and national chains; Google's local pack is weighted toward proximity, reviews, and citation consistency. These are different retrieval systems answering the same prompt.
In dental and HVAC — where the local signals (review count, proximity, profile completeness) carry the most weight and editorial coverage is thin — the two converge more often. Still, the best-performing vertical matches only 25% of the time on all three positions.
Backlinks don't explain the gap either
It would be clean if backlinks predicted AI citation — high-authority domain, more likely to be cited. The data does not support that. Top-3 median referring domains by vertical:
Real Estate: 69 median referring domains (lowest — and AI cites them 0.12 / 3)
Restaurants: 97 (AI cites 0.04 / 3)
Dental: 105 (AI cites 0.76 / 3 — highest)
HVAC: 105 (0.68 / 3)
Auto Repair: 107 (0.08 / 3)
Insurance: 130 (0.32 / 3)
Plumbing: 167 (0.60 / 3)
Legal: 170 (highest backlinks — and AI cites 0.16 / 3)
Dental has fewer median referring domains than Legal, and ChatGPT cites dental top-3 almost 5x more often. Whatever ChatGPT is using to decide who to name, it is not a simple function of backlink authority.
What the top-3 actually have, by vertical
Zooming in on the backlink distribution at each vertical's median market:
Plumbing — top-3 median 167 referring domains. But a few markets skew huge because national chains (Roto-Rooter, Mr. Rooter) hold 7,000-8,000-referring-domain profiles. The independent #2 in Grand Rapids ranks with 61 referring domains. Backlinks are not the entry bar; being Mr. Rooter is.
Insurance — 130 median. The 0.32 AI-citation rate is consistent with the DA-inversion pattern documented separately (insurance-da-inversion). National brands with 6,000+ refs sit at positions 8-15, not the top-3.
Legal — 170 median, highest of the eight verticals. Same structural mix as plumbing: Morgan & Morgan (7,882 refs nationally) stacked next to local firms with 60-120 refs. The mid-range of legal is larger because the local bar has infrastructure — county bar association listings, PA-announced mentions, community sponsorships — that plumbing does not.
Real Estate — 69 median, lowest. Consistent with a vertical where links accumulate mostly via brokerage directories, not editorial mentions. Also the worst AI mention rate among non-restaurant verticals.
Restaurants — 97 median. Middle of the pack on backlinks but last on AI mentions. The gap suggests ChatGPT weighs editorial food coverage much more heavily than the local pack does.
What this implies if you sell local SEO
Stop selling "backlinks for local pack position" as a single service. The top-3 in most verticals have 100-200 referring domains, not thousands. Within that range, adding more does not predict movement. Outside that range — either far below or far above — the relationship is noisier still.
If you sell "AI search optimization," know which verticals it even applies to. In restaurants, auto repair, and real estate, ChatGPT is currently citing sources that have nothing to do with Google's local pack. Optimizing your GBP harder will not fix that. Separately, in dental and HVAC, the two systems converge enough that strong local signals do tend to surface in AI answers.
What to actually spend on, vertical by vertical, based on what distinguishes top-3 from positions 4-10 in the IMPIOUS benchmark: proximity (can't change, but can influence service-area boundaries), review freshness within the last 90 days (see review-gap post), category match, business-name match, and GBP completeness above 85%.
Scan your own market
Our audit pipeline pulls your GBP, the top 20 competitors, and their profile data in about 30 seconds. Three free queries, no signup.
// backlinks + AI-mention data above is a one-time snapshot. the live scan does not call those APIs.
The full dataset
Per-market backlink profiles and AI-mention results live on each benchmark page. Methodology and CSV at /research/state-of-local-search-march-2026. CC BY 4.0 — cite as DuBois, IMPIOUS LLC, 2026.