83 Automotive AI Search Statistics (2026)

Your dealership’s finance and service pages explain the terms buyers will question at the desk. If an assistant reads a competitor’s explanation instead, your staff must answer expectations they did not set.

For questions about loans, fees, and used-car checks, ChatGPT answers from regulators, the Federal Trade Commission (FTC) and the Consumer Financial Protection Bureau (CFPB), the federal consumer regulators, above all, and cited no dealer’s page. Google AI Mode cited dealers’ own posts in 42.42% of answers that listed a source, next to Reddit and YouTube.

These are one day's answers to questions asked at the country level, and a cited page is a page the answer pointed to, not a click or a sale.

The Chat Window Sits Beside the Car Sites, Mostly for Buyers Under 45

Younger buyers ask an assistant far more often than older ones. In McKinsey's 2026 survey, 28% of buyers under 45 in Europe and the US used AI tools during their last car purchase, against 5% of buyers over 45. EV buyers were at 24%, buyers of gas and diesel cars at 14%.¹

For shoppers still deciding, the chat window now sits beside the car sites. Cox Automotive's tracker of US shoppers planning to buy within a year puts AI tools (36%) nearly level with automotive websites (38%) as a research channel.²

The assistant doesn't replace the brand's own site. In J.D. Power's survey of US new-vehicle shoppers, 94% of undecided shoppers visit a manufacturer's website, and one in three start without a brand in mind.³

In the call-tracking vendor Invoca's own 2026 benchmarks report on its automotive customers, ChatGPT-referred calls had a 60% lead rate. Of those leads, 40% converted on the call, against up to 52% for TV and video.⁶

Invoca says ChatGPT had the highest lead rate of any channel, but call volume remained very low, with no measurable calls from other assistants.⁶

Buyers use the chat window before they call, and the calls it sends are still few. None of those surveys says where the answers come from. That's what the rest of this article measures, and it's the part a company can change.

Which-Car Answers Lean on Reviewers and Marketplaces on Both Engines

These are the questions a buyer asks before picking a car: the best new SUV for a family, a reliable commuter under $25,000, where to buy a certified used car, the plug-in hybrid with the longest electric range.

On ChatGPT, car reviewers and testers were cited in 79.15% of these answers and marketplaces and vehicle-history sites in 63.03%. Google AI Mode leaned the same way, less heavily: 68.32% and 34.16%.

(Our research: answers from ChatGPT and Google AI Mode in the US, the UK, Canada, and Australia on October 20, 2026. We classified cited sites by who runs them; each share covers one engine’s answers to one question type, counting only answers with sources.)

The sites at the top differ by engine. On ChatGPT the four most-cited were Car and Driver (31.28% of these answers), Cars.com (29.86%), Consumer Reports (29.38%) and Edmunds (28.44%). On Google AI Mode, U.S. News's car rankings led (35.64%), then Edmunds (27.23%), YouTube (14.85%), Reddit (13.86%) and Kelley Blue Book (12.38%).

The widest gap between the engines is the automaker. On October 20, ChatGPT cited a carmaker's own site in 41.23% of which-car answers: model pages on toyota.com, honda.com and kia.com, and certified-used program pages. Google AI Mode did in 2.48%.

Dealers appear on both engines, as different kinds of page. ChatGPT's dealer citations (7.11% of these answers) were mostly storefronts: certified-inventory pages of dealers in Phoenix, Portland and Sacramento, and used-car search pages, cited when the buyer asked where to buy a certified or used car.

Google AI Mode's (16.34%) were dealers' blog posts: the best cars under $25,000, certified pre-owned against plain used, the mistakes to avoid when buying used.

Lender and Insurer Questions Go to NerdWallet, the WSJ and U.S. News

The second kind of buying question asks who should do the job: which lender approves bad credit, where to get pre-approved for a car loan, the cheapest car insurance providers, which shop to trust with tires and brakes.

Lenders, insurers and personal-finance publishers were cited in 66.43% of ChatGPT's answers to these and 68.35% of Google AI Mode's. The publishers differ by engine:

NerdWallet ranked first on ChatGPT (37.76%) and second on Google AI Mode (30.94%). Past it, ChatGPT went to the Wall Street Journal's Buyside personal-finance pages (36.36%) and the insurers' own sites, GEICO (32.87%) and Progressive (31.47%).

Google AI Mode went to U.S. News's insurance pages (33.81%) and CNBC Select (20.86%), and barely cited the insurers at all.

Dealers were cited in 4.90% of ChatGPT's answers and 3.60% of Google AI Mode's. When a buyer asks an assistant who should finance the car, the dealer's finance office is almost never among the cited sources, though the buyer will sit across the desk from it.

How-It-Works Answers: Regulators on ChatGPT, Dealers' Posts on Google AI Mode

The third kind of question concerns your F&I office, the finance and insurance office that arranges loans and add-ons at the desk, and your service desk. Buyers ask how leasing works, whether to borrow from a bank or dealer, which fees are negotiable, what negative equity and gap insurance mean (owing more on a car than it is worth, and the cover for that difference), how to check a used car, and how long home charging takes.

On ChatGPT, a government or regulator page was cited in 79.45% of these answers, and the FTC or the CFPB alone in 34.93%.

The pages were the FTC's “Buying a Used Car From a Dealer”⁷ and “Financing or Leasing a Car”⁸, the CFPB's “What things can I negotiate when shopping for a car or auto loan?”,⁹ the recall lookup of the National Highway Traffic Safety Administration (NHTSA), the federal vehicle-safety agency, and the pages on home charging of the Environmental Protection Agency (EPA), the federal environment agency.

Outside the US the same kind of answer cited each market's own government: GOV.UK, the Government of Canada and Ontario, New South Wales, Victoria and Australia's Moneysmart.

Under the dealer-fee question it also cited the price-transparency FAQs that FTC staff published for auto dealers on September 15.¹⁰

ChatGPT's other main sources were Consumer Reports (26.71%) and the motoring club AAA (11.64%).

On ChatGPT, a dealer's page was cited in 0.00% of these answers, and Reddit and YouTube in none.

Google AI Mode answered the same questions from other sources. Government pages were cited in 16.97% of its answers with sources, the FTC or the CFPB in 3.64%. Dealers' own posts were cited in 42.42%, and Reddit or YouTube in 45.45%: Reddit 29.09%, YouTube 28.48%.

Google AI Mode also answered many of these with no source at all: 43.49% of its how-it-works answers listed none, against 5.61% of its which-car answers. The Google shares in this section are of the answers that listed one; counted against all its how-it-works answers, sourced or not, dealers' posts were cited in 23.97%.

I can't tell you why the engines split this way, and the answers don't say. What I can see is the kind of page each one picks.

Ask both engines what negative equity in a car loan is. ChatGPT cited the CFPB's page on trading in a car that isn't paid off, the FTC and a credit bureau.

Google AI Mode cited four dealers' own posts (two Chevrolet stores, a Hyundai store and a Honda store in Canada), a credit union, USAA's education foundation, the FTC and Ontario's dealer regulator.

How-It-Works Answers Talk About Dealers; Only Google AI Mode's Cite Them

Dealers are in these answers either way. ChatGPT's how-it-works answers mention dealers in 32.88% of cases and cite one in none. Its which-car answers mention dealers in 61.14% and cite one in 7.11%.

In ChatGPT's answers the dealer is mostly the party to check: its fees to question, its financing to compare with a bank's or a credit union's quote.

Asked how dealer fees work, one of its answers drafted an email for the buyer to send, asking for “your complete out-the-door price in writing, including the selling price, every dealer fee, taxes, title, registration,” and cited the CFPB, the FTC and New York's DMV.

No dealer's own explanation of its fees was among the sources.

On Google AI Mode, dealers' posts are cited more often than dealers are mentioned: 37.58% of its how-it-works answers with sources mention dealers, and 42.42% cite one.

The cited posts cover the store's own work, and each topic has an owner inside the building:

Finance topics were the biggest group.

One Site List Won't Cover Both Kinds of Question

Each engine leans on a short list for each kind of question. Two or three sites, between them, are cited in half of each group's answers that listed a source:

The lists don't carry over between kinds of question. The sites behind 80% of ChatGPT's cited which-car answers reach 36.99% of its how-it-works answers; on Google AI Mode, 57.58%. Between engines, the ten most-cited sites share three names on car questions (Cars.com, Consumer Reports, Edmunds) and three on how-it-works questions (AAA, Consumer Reports, Edmunds).

So keep a short list per kind of question, per engine, with your own pages marked wherever they appear. A list from one day's answers is a starting point; it stays useful only if the same questions are asked again on both engines and every cited page is kept.

The desk meets whatever those pages told the buyer, so the store's own finance and service pages are the part of each list it can keep current.