On this page
- Assistants Name a Carmaker in Most Answers About Which Car to Buy
- The Named Make's Own Pages Sit Under a Minority of Those Answers
- Reviewers and Rankings Write the Description, With Video and Owners on Google AI Mode
- When a Carmaker's Own Page Gets In, It Settles One Narrow Question
- Head-to-Head Pages Sit Under Most of ChatGPT's New-Car Answers
- Frequently asked questions
If you run a dealer group, import a brand, or build vehicles of your own, you probably assume that online the maker's site and your own pages say what the car is, and the reviews add an opinion.
Most buyers already read other people's pages. In Cox Automotive's 2025 buyer study, 75% of buyers visited third-party sites while shopping, and 25% visited an automaker's website.1
So in September 2026 we put the questions buyers ask about which car to get, new or used, electric or hybrid, to ChatGPT and Google AI Mode, Google's chat-style answers, worded as buyers in the US, the UK, Canada, and Australia would ask them, and read every page each answer listed as its source.
The assistant names a carmaker in most answers about which car to buy and describes it from someone else's page. These are one month's cited pages: they show which pages described the car, not who sold it.
Assistants Name a Carmaker in Most Answers About Which Car to Buy
In our research, 79.44% of ChatGPT's which-car answers named at least one carmaker or one of its models, and 86.92% of Google AI Mode's did. On new-car questions, ChatGPT named one in 98.37% of its answers.
The names are the same few. Toyota appeared in 62.31% of ChatGPT's which-car answers and 70.09% of Google AI Mode's. On ChatGPT, Honda came next at 52.34%, then Kia (39.25%) and Hyundai (38.32%). On Google AI Mode those three were nearly level: Kia at 42.99%, Honda and Hyundai at 42.06% each.
Toyota Tops Both Engines' Car Shortlists
% of which-car answers
View the data
| Label | ChatGPT | Google AI Mode |
|---|---|---|
| Toyota | 62.31% | 70.09% |
| Honda | 52.34% | 42.06% |
| Kia | 39.25% | 42.99% |
| Hyundai | 38.32% | 42.06% |
| Mazda | 28.04% | 19.63% |
| Subaru | 26.48% | 16.82% |
| Ford | 16.82% | 25.23% |
| Chevrolet | 16.20% | 26.17% |
The answers that named no make were mostly where-to-buy answers. Ask for the best way to buy a used car online and you get CarMax, Carvana, CarGurus, and Autotrader. That's why only 52.99% of ChatGPT's answers on used and certified pre-owned cars (inspected and sold with a program's terms) named a make at all.
If your make isn't one of those four, the answer almost never names you alone. Of ChatGPT's which-car answers that named any other make, 94.44% also named Toyota, Honda, Kia, or Hyundai. So a buyer reads about your model in the same paragraph as theirs.
How many buyers read that paragraph is the number people argue about. Four surveys, four answers:
The samples differ: Cars.com partly recruited existing AI users, while Invoca, which sells call tracking to dealers, surveyed buyers for its own report a year after CarGurus. Invoca's figure rose from 41% in its 2025 report.4 Plan on the low end and watch the direction.
The Named Make's Own Pages Sit Under a Minority of Those Answers
In September, on ChatGPT, 43.30% of which-car answers cited a page owned by a make the answer named, against 79.44% that named one. On Google AI Mode it was 7.48%, against 86.92%.
Make by make on ChatGPT, the gap holds. Of the answers that named Toyota, 29.00% cited a Toyota page. Honda: 29.17%. Hyundai: 18.70%.
Kia was named in 39.25% of ChatGPT's which-car answers, and a Kia page sat under 8.73% of the answers that named it. In most answers that name a Kia, every page underneath belongs to someone else.
Named Far More Often Than Their Own Pages Are Cited
% of which-car answers
View the data
| Label | Named in the answer | Named, with its own page cited | Note |
|---|---|---|---|
| Toyota | 62.31% | 18.07% | Of the answers naming Toyota, 29.00% cited a Toyota page. |
| Honda | 52.34% | 15.26% | |
| Kia | 39.25% | 3.43% | Of the answers naming Kia, 8.73% cited a Kia page. |
| Hyundai | 38.32% | 7.17% |
Cars.com's survey hints at why the cited page matters. Of the shoppers it asked who had used AI, the most common step after an answer was to visit a cited dealer or manufacturer website, at 41%.3 Cars.com sells to dealers and ran the survey itself, so read it as one marketplace's view.
Reviewers and Rankings Write the Description, With Video and Owners on Google AI Mode
On ChatGPT, Edmunds sat under 30.53% of which-car answers, Car and Driver under 25.23%, Cars.com 17.45%, Consumer Reports 13.71%, and Kelley Blue Book 10.59%. The motoring press and car-review sites, counted together, appeared in 64.80%.
Edmunds and Car and Driver Sit Under the Most ChatGPT Car Answers
% of which-car answers
View the data
| Label | Value | Note |
|---|---|---|
| Edmunds | 30.53% | |
| Car and Driver | 25.23% | |
| Cars.com | 17.45% | |
| Consumer Reports | 13.71% | |
| Kelley Blue Book | 10.59% | |
| iSeeCars | 10.28% | |
| Autoblog | 10.28% | |
| toyota.com | 9.97% | The most cited carmaker site; Toyota's certified-program and shopping sites are counted apart. |
Google AI Mode reads a different mix. U.S. News sat under 29.91% of its which-car answers and YouTube under 28.97%, followed by Edmunds (20.56%), Reddit (15.89%), and Kelley Blue Book (10.28%).
The videos were creators' rankings and comparisons, the kind titled "9 Best Compact SUVs for 2026." The Reddit threads were people asking each other the same questions: which minivan suits a family, or whether an EV works without a charger at home.
On Google AI Mode, Carmakers' Own Sites Come Last
% of which-car answers
View the data
| Label | Value | Note |
|---|---|---|
| Motoring press and review sites | 58.88% | U.S. News, Edmunds, Kelley Blue Book, Consumer Reports and the rest of the named set. |
| YouTube | 28.97% | |
| Marketplaces and retailers | 26.17% | |
| Dealers' own sites | 17.76% | Guides and explainers on dealers' sites: cargo space, fuel-efficient sedans, certified pre-owned. |
| 15.89% | ||
| Carmakers' own sites | 7.48% |
Google AI Mode also cited dealers' own pages, in 17.76% of its which-car answers: guides on dealer sites about the best cars for cargo space or fuel-efficient sedans for commuting, and explainers on what certified pre-owned means.
For a dealer group, that's the route on this engine, and it runs through pages the store writes. Name a city in the question and the answers cite the stores themselves far more often.
When a Carmaker's Own Page Gets In, It Settles One Narrow Question
Of the carmaker pages ChatGPT cited in which-car answers, 57.01% were model, spec, and trim pages (a trim is a version of a model with its own equipment and price) and 21.98% were certified or approved used programs. Newsroom and pricing pages made up 8.21%, head-to-head comparisons 7.73%, and towing guides and brochures 5.07%.
Carmaker Pages ChatGPT Cited Were Mostly Specs and Program Terms
% of carmaker pages cited
View the data
| Label | Value | Note |
|---|---|---|
| Model, spec and trim pages | 57.01% | Model overviews 37.44% and spec, trim and feature pages 19.57%. |
| Certified or approved used programs | 21.98% | |
| Newsroom and pricing pages | 8.21% | |
| Head-to-head comparisons | 7.73% | |
| Towing guides and brochures | 5.07% |
Most of them hold a fact the maker states with more authority than anyone else: Ford's towing guides for the F-150, Mazda's newsroom notice of price changes on select 2026 models from May 4, Kia's suggested-retail-price page for the K4, Honda's specification page for the CR-V.
The used questions show it most cleanly. In ChatGPT's used and certified answers, almost every carmaker page cited was a certified-program page, and 22.22% of those answers cited one: Toyota Certified, Honda Certified, Mazda's certified pre-owned page, BMW's and Volvo's approved used programs. A program's terms are the maker's to state.
An assistant takes a carmaker's page when the carmaker is the best source for the one fact the buyer asked about.
That's my reading of the pages ChatGPT cited, and the data can't prove it: it shows which pages were cited, not why. Google's own guidance says no special markup or structured data is needed to appear in its AI features,5 so the route isn't a markup project.
Head-to-Head Pages Sit Under Most of ChatGPT's New-Car Answers
One page kind lets a brand speak about its own car in the buyer's frame: the head-to-head against a named rival.
61.79% of ChatGPT's new-car answers cited at least one head-to-head page. Most came from the motoring press (cited in 44.72% of new-car answers), but carmakers' own head-to-heads sat under 17.89%, and dealers' under 4.88%.
Carmakers Wrote Head-to-Heads Under 17.89% of ChatGPT's New-Car Answers
% of new-car answers
View the data
| Label | Value | Note |
|---|---|---|
| Any head-to-head page | 61.79% | |
| Motoring press | 44.72% | |
| Carmakers | 17.89% | Toyota, Chevrolet, GMC, Subaru, Ram and the Honda Information Center, among others. |
| Other sites | 9.76% | |
| Dealers | 4.88% | |
| Marketplaces | 2.44% |
The makers' pages that got cited were plain: Toyota's Corolla against the Honda Civic and Camry against the Accord, Chevrolet's Silverado 1500 against the Ford F-150, Subaru's Outback against the CR-V, the Honda Information Center's Odyssey against the Sienna.
Dealers wrote them too: a Subaru store's Crosstrek against the Corolla Cross, HR-V, and CX-30, and a Toyota store's RAV4 against the CR-V for Chicago winters.
This route has two limits. Google AI Mode cited a head-to-head in only 3.74% of its which-car answers, against 38.01% of ChatGPT's across the same three question sets, so this is a ChatGPT route; on Google AI Mode, the description runs through rankings, video, owners, and dealers' guides.
And which rivals an answer names beside you, and which pages sit under it, change with the question and the engine, and even between two askings: in our August reading of ChatGPT, two askings of the same car question shared 19.08% of their cited pages on average.
So the pairing can't be read off one screenshot. You see it by asking your buyers' questions on both engines on a schedule and keeping every page each answer cites.
The verdict on your car stays on other people's pages. What a carmaker or a dealer can put under these answers is the fact it owns: the model and trim pages, the certified program's terms, and a head-to-head against the rival the answers already name beside it.
Frequently asked questions
Does a cited page mean the assistant recommended that car?
A citation means the answer listed the page as a source, not that it recommended the car. One page can supply a single figure while the answer recommends a rival.
How many car buyers ask an AI assistant?
Between 19% and 63%, depending on the survey. Cox Automotive found 19% of 2025 buyers used AI chat sites or Google's AI summaries;1 Invoca, which sells call tracking to dealers, found in its own survey that 63% of US car buyers surveyed in May 2026 used generative AI to research.4 The samples and dates differ, so plan on the low end.
Why does Google AI Mode cite our own site so rarely when it names our models?
Our September research found that Google AI Mode rarely cited the named make's own pages, but it cannot establish why. The sources included U.S. News rankings, YouTube reviews, Edmunds, Reddit threads, and dealers' guides.
Does a dealer group need its own guides for Google AI Mode?
On Google AI Mode, dealers' own guides sat under more which-car answers than the named carmakers' sites: 17.76% against 7.48%. The guides that got in answered a practical buyer question, such as the best cars for cargo space, fuel-efficient sedans for commuting, or what certified pre-owned means.
Do certified pre-owned program pages matter in used-car answers?
On ChatGPT they were almost the only carmaker pages its used and certified answers cited, and 22.22% of those answers cited one. A program's terms are the maker's to state, so keep that page current and dated.
Do we need special markup or structured data to show up?
Google says no: there's no special schema.org markup needed to appear in AI Overviews or AI Mode.5 The carmaker pages ChatGPT cited held facts the maker states itself: specs, certified-program terms, prices, and head-to-heads.
Which page should our team write first?
Choose your best-selling model and the rival that assistants repeatedly name beside it. Compare the same model year and comparable trims, putting a source and date beside each figure so your manager can approve the page.
What should a head-to-head page include?
Two cars of the same model year and comparable trims, the figures that matter to the buyer's question, and beside each figure its source and the date it was read. Label prices as suggested retail, with their date, and claim nothing beyond measurable attributes and price.
Where do the figures for a head-to-head come from?
Your own spec, trim and price records for your model. In the US, fuel economy comes from the federal records6 and safety ratings from the National Highway Traffic Safety Administration,7 both published as public data. For the rival, a person copies the figures from its own spec page and price list, unless that site's terms allow automated reading.
Can we publish a page that names a competitor's car?
In the US, the Federal Trade Commission, the federal consumer-protection agency, encourages comparative advertising that names competitors, as long as it's clear and not deceptive.8 Compare measurable attributes or price, with sources, and have someone who owns the figures approve it. Outside the US, check your market's rules.
Who should approve a comparison page before it goes live?
Whoever owns the spec sheet: usually a product manager at a maker or importer, or the brand's sales manager at a dealer group. Nothing should publish until they have checked each figure against its dated source.
Is a comparison page worth it if we sell only a few models?
If you sell one make with three models, somebody can write three pages a year by hand. The work grows with the pairings: each model, each rival the answers name beside it, each model year and each country you sell in.
How often should we check which rivals we're paired with?
Check monthly, and again at a model-year changeover or price change. Our earlier automotive research found that repeated answers changed their cited pages, so one screenshot cannot establish a stable pairing.
Do these statistics show which car brands sold more vehicles?
No, the research records the models named and pages cited in answers during one month. It does not establish who sold a vehicle, why an assistant chose a source, or whether a citation caused a purchase.
Sources
- Cox Automotive: 2025 Cox Automotive Car Buyer Journey Study Summary (2026)
- CarGurus: CarGurus Study Reveals How AI and Omnichannel Shopping are Shaping the Car Buying and Selling Experience (2025)
- Cars.com: Cars.com Survey Reveals AI's Growing Influence on Car Shopping (2025)
- Invoca: The Automotive Buyer Experience Report 2026 (2026)
- Google Search Central: AI features and your website (2025)
- FuelEconomy.gov (U.S. Department of Energy and EPA): FuelEconomy.gov Web Services (2026)
- National Highway Traffic Safety Administration: NHTSA Datasets and APIs (2026)
- Federal Trade Commission: Statement of Policy Regarding Comparative Advertising (1979)
