The Local AI Visibility Index
Acupuncture across 6 US markets
We measured every Acupuncture practice Google Maps returns in 6 markets — 167 businesses in total — then asked 5 AI answer engines the same 10 patient questions in each one and recorded who got named.
- Practices
- 167
- Markets
- 6
- Questions
- 10 per market
- Engines
- 5
- Measured
- July 2026
85.7%
of the best-reviewed Acupuncture practices across 6 markets were named by no AI answer engine at all.
Finding 01
How often do the best-ranked local businesses show up in AI answers?
The businesses ranked highest on Google Maps still show up in only 13.2% of AI answers. Measured across 167 acupunctures in 6 markets: across the best-ranked 10% by map position, an AI engine named them in 13.2% of the answers about their own category.
Ranking well clearly helps. It is also clearly not the thing that decides it — most of the best-ranked businesses across these 6 markets still never came up. And this is an average of markets that do not agree with each other: rank counted for more in the smaller markets we measured than in the larger ones.
| Where they rank on Google Maps | Named by AI | Share |
|---|---|---|
| The 1st third — best ranked on Maps | 14 of 55 | 25.5% |
| The middle third | 4 of 55 | 7.3% |
| The bottom third | 1 of 57 | 1.8% |
Finding 02
How much of the AI answers do a handful of businesses take?
In the average local market, three businesses take 97.1% of everything AI answer engines say. Measured across 167 acupunctures in 6 markets: the three most-named businesses in each market got 97.1% of all the mentions, and everyone else split what was left.
Finding 03
Are the best-reviewed local businesses named by AI answer engines?
85.7% of the best-reviewed local businesses are never named by an AI answer engine. Measured across 167 acupunctures in 6 markets: among the top half by Google rank, star rating and review count combined, 85.7% show up in Google Maps but were named in none of the AI answers about their own category.
Finding 04
Do Google reviews decide whether AI names a business?
Google reviews do not decide whether AI answer engines name a business. Measured across 167 acupunctures in 6 markets: neither the number of reviews nor the star rating tracked with getting named — both came out close enough to zero to mean nothing (review count -0.037, star rating -0.026, on a scale where 1.0 would be a perfect match).
Finding 05
Do different AI answer engines recommend the same businesses?
No two AI answer engines recommend the same local businesses. Measured across 167 acupunctures in 6 markets: asked the same questions, any two engines agreed on somewhere between 25.0% and 54.8% of the businesses they named — so being named by one is no guarantee of being named by the others.
Across every market combined, each engine named a different share of the businesses measured: Gemini named 26, Perplexity named 22, Google AI Overview named 15, ChatGPT named 12, Claude named 9.
| Engine pair | Agreed on |
|---|---|
| Gemini & Perplexity | 54.8% |
| Gemini & Google AI Overview | 51.9% |
| Claude & Google AI Overview | 50.0% |
| Perplexity & Google AI Overview | 42.3% |
| Claude & Perplexity | 40.9% |
| ChatGPT & Gemini | 35.7% |
| ChatGPT & Google AI Overview | 35.0% |
| ChatGPT & Claude | 31.2% |
| ChatGPT & Perplexity | 25.9% |
| Claude & Gemini | 25.0% |
How much does this vary between markets?
Substantially. The share of practices no engine named ranged from 0% to 93% across the 6 markets measured. Every market’s own figures are published on its own page, computed from that market alone.
| Market | Practices | Never named | Top 3 took |
|---|---|---|---|
| Cape Coral, FL | 72 | 91.7% | 95.2% |
| New Haven, CT | 71 | 93% | 95% |
| La Crosse, WI | 13 | 84.6% | 100% |
| Sioux City, IA | 8 | 50% | 92.7% |
| Laredo, TX | 2 | 50% | 100% |
| Saginaw, MI | 1 | 0% | 100% |
How we measured it
In each of 6 markets we pulled every Acupuncture practice Google Maps returns, then asked 5 AI answer engines the same 10 patient questions and recorded which ones each engine named.
- Sample
- 167 practices across 6 markets: Cape Coral, FL; La Crosse, WI; Laredo, TX; New Haven, CT; Saginaw, MI; Sioux City, IA. A practice with more than one listing is counted once per market. 34 listings that share the websites of 16 organizations are counted as those organizations, because AI engines cite an organization's site rather than any individual listing.
- Market selection
- The 6 markets were picked before any measuring started — deliberately spread across large, medium and small cities and across the country — and every market we opened is reported here, including the ones that came back less interesting.
- Collection
- 10 patient questions per market, held identical across ChatGPT, Claude, Gemini, Perplexity and Google's AI Overview. 50 answers per market, July 2026.
- Visibility
- Share of answers naming or citing a practice, weighted by the reach of the engine that named it.
- What we will and will not publish
- A combined number only goes up when it covers at least 3 markets and 150 practices, with a usable reading for at least half of them. Where markets disagree with each other, we publish the range as well as the average rather than hiding it.
- Scope
- One category, 6 US markets. Practices are never named. Answer engines change over time; these figures describe July 2026.
For anyone who wants the statistics
The findings above are counts, which is how most people want to read them. Here are the same relationships as correlations, across all 167 businesses.
| What we measured | How closely it tracked | What that means |
|---|---|---|
| Google Maps rank | −0.288 | A slight connection |
| Review count | −0.037 | No connection |
| Star rating | −0.026 | No connection |
Reading the middle column: 0.0 means the two things had nothing to do with each other. 1.0 would mean they moved together perfectly. A minus sign means they moved in opposite directions. Anything under about 0.2 either way is noise.