How AI recommends brands
in the United States
Which brands ChatGPT, Gemini, Google AI Mode and Perplexity recommend across 26 US markets, and which sources they rely on.
- Markets
- 26
- Questions
- 9,789
- Answers
- 117,408
- Citations
- 780,661
01 · Categories
Pick a category
26 measured US markets in 10 categories. Open a category, or jump straight to a market. Leaders on ChatGPT.
- Home Services 7 markets
- Healthcare 4 markets
- Finance & Insurance 2 markets
- Legal Services 2 markets
- Software 2 markets
- Fashion & Apparel 4 markets
- Beauty & Personal Care 2 markets
- Consumer Electronics 1 market
- Sports & Outdoors 1 market
- Pets 1 market
02 · Openings
Where it’s easiest
to break in
The lower the top-5 share, the weaker the leaders, and the easier it is
for a new brand to get named.
03 · Sources
Most-cited domains across all 26 markets
The brands change from market to market, the sources barely do.
| Domain | Citations | Markets In how many of the 26 markets ChatGPT cites the domain |
|---|---|---|
| nih.gov (opens in a new tab) | 10 610 | 20 / 26 |
| aad.org (opens in a new tab) | 6 422 | 7 / 26 |
| irs.gov (opens in a new tab) | 2 589 | 13 / 26 |
| energy.gov (opens in a new tab) | 2 387 | 9 / 26 |
| fda.gov (opens in a new tab) | 2 305 | 14 / 26 |
| ada.org (opens in a new tab) | 2 201 | 1 / 26 |
| epa.gov (opens in a new tab) | 2 167 | 14 / 26 |
| fanniemae.com (opens in a new tab) | 1 654 | 4 / 26 |
| cdc.gov (opens in a new tab) | 1 537 | 18 / 26 |
| ca.gov (opens in a new tab) | 1 480 | 19 / 26 |
| hud.gov (opens in a new tab) | 1 260 | 6 / 26 |
| energystar.gov (opens in a new tab) | 1 216 | 6 / 26 |
| dermnetnz.org (opens in a new tab) | 1 215 | 7 / 26 |
| clevelandclinic.org (opens in a new tab) | 1 200 | 12 / 26 |
| servicetitan.com (opens in a new tab) | 1 195 | 5 / 26 |
| sec.gov (opens in a new tab) | 1 081 | 18 / 26 |
| whowhatwear.com (opens in a new tab) | 1 039 | 7 / 26 |
| consumerfinance.gov (opens in a new tab) | 1 007 | 6 / 26 |
| housecallpro.com (opens in a new tab) | 998 | 4 / 26 |
| dol.gov (opens in a new tab) | 991 | 7 / 26 |
04 · Concentration
How concentrated each engine’s sources are
On Google AI Mode, 84 sites are half of everything it cites.
| Top domains | ChatGPT | Gemini | Google AI Mode | Perplexity |
|---|---|---|---|---|
| 1 | 5.3% | 11.3% | 17.4% | 3.4% |
| 2 | 8.4% | 13.3% | 31.8% | 5.4% |
| 3 | 9.7% | 14.2% | 33.9% | 6.6% |
| 5 | 12% | 15.1% | 36.9% | 8.5% |
| 10 | 16.5% | 16.8% | 39.9% | 12% |
| 20 | 22.1% | 18.9% | 42.5% | 15.5% |
| 30 | 26.2% | 20.5% | 44.3% | 18% |
| 50 | 31.9% | 23.1% | 46.8% | 21.7% |
| 100 | 41.1% | 27.5% | 51.1% | 27.9% |
| 200 | 51% | 33% | 56% | 35.4% |
| 300 | 56.7% | 37% | 59.1% | 40.2% |
| 500 | 63.6% | 42.6% | 63.3% | 46.6% |
| 1000 | 72.8% | 51.3% | 69.5% | 55.9% |
| 2000 | 81% | 61.6% | 76.2% | 65.6% |
| 3000 | 84.8% | 68.3% | 80.4% | 71.4% |
| 5000 | 88.9% | 77.2% | 85.7% | 78.6% |
| 10000 | 94.1% | 89% | 92.9% | 88% |
| 19026 | 100% | |||
| 21969 | 100% | |||
| 22245 | 100% | |||
| 22798 | 100% |
- ChatGPT187of 21,969 domains
- Gemini907of 22,798 domains
- Google AI Mode84of 19,026 domains
- Perplexity649of 22,245 domains
The steeper the curve, the shorter the list you have to be on. Gemini needs 907 domains to reach the same half, so one brand page is a far smaller bet there.
Source: Findrix research. Cumulative share of each engine’s citations by number of domains, United States, 26 markets, measured Sep 2026.
05 · Market context
AI search behaves
differently across countries.
AI engines cite different domains depending on the country.
United States
| # | Domain | Share of citations Share of all ChatGPT citations in this country that go to the domainShare | Citations |
|---|---|---|---|
| 1 | nih.gov (opens in a new tab) | 5.3% | 11K |
| 2 | aad.org (opens in a new tab) | 3.2% | 6.4K |
| 3 | irs.gov (opens in a new tab) | 1.3% | 2.6K |
| 4 | energy.gov (opens in a new tab) | 1.2% | 2.4K |
| 5 | fda.gov (opens in a new tab) | 1.1% | 2.3K |
| 6 | ada.org (opens in a new tab) | 1.1% | 2.2K |
| 7 | epa.gov (opens in a new tab) | 1.1% | 2.2K |
| 8 | fanniemae.com (opens in a new tab) | 0.8% | 1.7K |
| 9 | cdc.gov (opens in a new tab) | 0.8% | 1.5K |
| 10 | ca.gov (opens in a new tab) | 0.7% | 1.5K |
Australia
| # | Domain | Share of citations Share of all ChatGPT citations in this country that go to the domainShare | Citations |
|---|---|---|---|
| 1 | nih.gov (opens in a new tab) | 2.8% | 9.1K |
| 2 | nsw.gov.au (opens in a new tab) | 2.2% | 7.1K |
| 3 | fcfcoa.gov.au (opens in a new tab) | 1.5% | 5.0K |
| 4 | ndis.gov.au (opens in a new tab) | 1.5% | 4.8K |
| 5 | aad.org (opens in a new tab) | 1.4% | 4.6K |
| 6 | ato.gov.au (opens in a new tab) | 1.0% | 3.4K |
| 7 | google.com (opens in a new tab) | 0.7% | 2.4K |
| 8 | health.gov.au (opens in a new tab) | 0.7% | 2.3K |
| 9 | servicesaustralia.gov.au (opens in a new tab) | 0.7% | 2.2K |
| 10 | moneysmart.gov.au (opens in a new tab) | 0.7% | 2.2K |
Other countries
- Australia27 markets measuredOpen Australia
- United KingdomNext country in the measurement planUpcoming
Frequently asked questions
What does this United States AI Search report measure?
Which brands ChatGPT, Gemini, Google AI Mode and Perplexity name when US buyers ask for recommendations, and which websites they cite in those answers. It covers 26 US markets and 9,789 buyer questions — about 377 per market, each asked three times to each engine. Measured Sep 2026.
Which markets are included in the United States dataset?
26 measured markets in 10 categories: Home Services, Healthcare, Finance & Insurance, Legal Services, Software, Fashion & Apparel, Beauty & Personal Care, Consumer Electronics, Sports & Outdoors and Pets.
What does “mention rate” mean?
The share of a market’s questions where the engine names the brand in its answer. Every rate has a 95% range: when the ranges of two brands overlap, their ranks are not really different.
What counts as a citation?
A link or source the engine shows with its answer. We count each cited domain once per answer, so a site linked three times in one answer counts once.
Why can AI mention a brand without citing its website?
Engines usually build an answer from third-party pages — reviews, guides, forums, official bodies — and name the brands those pages recommend. The brand’s own site is often not among the sources at all.
Why do AI engines use different sources?
Each engine has its own search index, retrieval and ranking. In the United States, Gemini spreads its citations widest — half of them go to 907 domains — while Google AI Mode puts half of its citations on just 84.
Why does AI source behavior differ between countries?
Engines prefer local sources: national regulators, local media and local retailers. The same category draws on different domains in the United States and in Australia, so visibility has to be measured country by country.
Where do the questions in this dataset come from?
They are written from how buyers search in each market — comparisons, “best for” questions, problems and price questions — and asked to every engine the same way.
How often is this data updated?
Each page shows the month it was measured; the numbers here are from Sep 2026. When a market is measured again, its page shows the new month and numbers.