Scowty Research · as of August 6, 2026
The Invisible Dentist Index
We asked ChatGPT and Google Gemini to recommend a dentist in all 67 Florida counties. Half the state’s dental practices never came up once.
What we did
We took every business Google Places lists as a dental practice in each of Florida’s 67 counties — 2,327 practices after filtering out hospitals and other non-dental results. Then we asked ChatGPT and Google Gemini to recommend a dentist in each county, six different ways, 3 times each.
That is 36 separate checks per practice and 2,412 questions in total, all asked on August 6, 2026. We repeated every question because identical questions produce different answers: asked three times, one question named six practices, then six, then eight — eight different practices in total. A practice counts as “never recommended” here only if it failed to appear in any of its 36 checks.
What we found
50.4% of practices were never recommended. 54.8% were never cited as a source. The two assistants agreed closely: ChatGPT recommended 37.9% of practices and Gemini 38.7% — so this is not a quirk of one product.
Reputation on Google did not carry over. Of the 718 practices with 500 or more Google reviews, 295 were never recommended — 41.1% of that group. The most-reviewed practice never to come up has over 2,500 reviews at 4.5 stars.
It depends where you practise
Among counties with at least 30 practices, the share ever recommended ranges from 28.3% to 67.9%.
| County | Practices | Ever recommended |
|---|---|---|
| Nassau | 46 | 28.3% |
| Citrus | 45 | 28.9% |
| Columbia | 48 | 31.3% |
| Monroe | 40 | 35% |
| Palm Beach | 57 | 35.1% |
| Miami-Dade | 60 | 36.7% |
| Hillsborough | 60 | 38.3% |
| Duval | 58 | 41.4% |
| Lee | 58 | 41.4% |
| Volusia | 58 | 41.4% |
| Sumter | 57 | 42.1% |
| Okaloosa | 51 | 43.1% |
| Pinellas | 55 | 43.6% |
| Putnam | 31 | 45.2% |
| Orange | 59 | 47.5% |
| Alachua | 56 | 48.2% |
| Collier | 56 | 48.2% |
| Broward | 58 | 48.3% |
| Lake | 57 | 49.1% |
| Brevard | 57 | 52.6% |
| Sarasota | 55 | 52.7% |
| Clay | 53 | 52.8% |
| Highlands | 41 | 53.7% |
| Hernando | 53 | 54.7% |
| Manatee | 58 | 55.2% |
| Pasco | 55 | 56.4% |
| St. Lucie | 56 | 57.1% |
| Flagler | 54 | 57.4% |
| Polk | 54 | 57.4% |
| Seminole | 45 | 57.8% |
| Escambia | 57 | 57.9% |
| Martin | 59 | 59.3% |
| Charlotte | 52 | 59.6% |
| Marion | 56 | 60.7% |
| Osceola | 56 | 62.5% |
| St. Johns | 55 | 63.6% |
| Indian River | 55 | 65.5% |
| Leon | 59 | 67.8% |
| Bay | 56 | 67.9% |
Counties with fewer than 30 listed practices are excluded from this table because a percentage over a handful of practices is not meaningful. Full results for all counties are in the dataset.
Is your practice on the list?
Search 2,327 Florida practices to see what happened when we asked for a dentist in your county.
Method and limits
- Asked on August 6, 2026. AI answers drift over time. Every figure here describes that date and may already have changed.
- Queried through each provider’s API with web search enabled — OpenAI’s API using the same model generation OpenAI ships in ChatGPT, and Google’s Gemini with Search grounding. This is a close proxy for the consumer apps, not the identical product.
- 100% of the 2,412 questions returned an answer. Failed questions are excluded rather than counted as a non-recommendation.
- Matching is conservative. We credited a practice when its name or its website appeared in the answer. Practices trading under a different name than Google lists may be undercounted, so the true recommendation rate could be slightly higher than reported.
- Sample is what Google Places lists as a dental practice in each county, capped at 60 per county. A practice absent from Google Places is absent here.
- We publish county and state figures, not a ranked list of practices. Individual results are shown only to the practice that looks itself up.
About this research
Scowty helps local businesses get found on Google and in AI search. We ran this study with the same tooling we use for customers. Journalists and researchers can contact us for the underlying data and full method.