Most US car shoppers plan to use AI; under a third of dealers have adjusted. What to fix first

What Cox Automotive found

US car shoppers are ready to use AI, and most dealers have not adjusted the way they show up in it. Cox Automotive's AI in Auto Retail Tracker, published on 11 August 2026, surveyed about 500 dealers and 1,500 in-market shoppers in each of the first two quarters of 2026. The figures below are for Q2 2026 unless noted.¹

Among shoppers, 63% said they will definitely or probably use AI on their next vehicle purchase, and 24% said AI helps them feel more prepared when working with dealerships.¹ Among dealers, 82% use AI today. Yet only 29% have started to adjust to AI-powered search, and another 32% know they need to but have not started.¹

Cox sells software and services to dealers, so it has an interest in the gap it describes. The sample sizes are solid for a trade tracker, though, and the split between shopper intent and dealer readiness is the useful part.

US auto retail, Q2 2026

Shoppers are ahead of dealers

63%of in-market shoppers will definitely or probably use AI next timeCox Automotive ¹

82%of dealers use AI todayCox Automotive ¹

29%of dealers have started to adjust to AI-powered searchCox Automotive ¹

The gap is about where AI is used

Dealers use AI inside the store; shoppers use it before they reach one. In the tracker, dealers' top uses were automating routine and complex tasks (40%), coordinating customer follow-up (40%) and generating content (38%).¹ All three are useful. None of them changes what an AI engine says when a shopper asks which dealer near them has a certain model in stock.

Our view: "we use AI" and "AI recommends us" are separate problems, and most dealers have worked on the first. The 82% and the 29% are not in tension.¹ They describe a sector that adopted AI as a productivity tool and has not yet treated it as a discovery channel.

The discovery channel is real. Similarweb, which sells web analytics, found that about 20% of US ChatGPT prompts in automotive got an answer carrying a citation in May 2026, against 6.8% across all categories.³ Car questions are the kind with specific, checkable options, so engines link out more often, and those links go to someone's inventory page.

What should a dealer fix first?

Start with your Google Business Profile, because Google increasingly answers with its own cards. Profound, which sells AI visibility software, found that google.com's share of citations in Google's AI Mode grew 8.4 times between 15 April and 30 June 2026, driven by Business Profile and Product Knowledge Panel cards.² Its analysis of more than 32 million Business Profile instances found local-intent industries absorbing most of the change.²

For a dealer, "local intent" is most of the business. A shopper asking AI Mode for a dealer with a certified pre-owned SUV nearby may see a Google-hosted card before your site. That card is built from your profile: name, category, hours, phone, website, photos, reviews and, where supported, vehicle listings.

• Check the primary category and the brand names on the profile match what you sell.
• Keep sales and service hours separate and current, including holidays.
• Answer recent reviews, especially unanswered complaints, because review text sits on the card.
• Connect inventory where the profile supports it, so the card shows stock rather than only an address.

Our analysis of which sources AI Mode uses by query type covers why local queries lean on Google's own pages.

Then: inventory and prices AI can read

Make every vehicle detail page readable as text. Engines answer "which dealer has a 2024 hybrid under a certain price" from pages that state the year, model, trim, price, mileage and availability in words, not from photos, image overlays or scripts that load prices after the page renders.

• One page per vehicle, with VIN, stock number, price, mileage, trim, colour and status in the HTML.
• Vehicle structured data with an offer and its price, matching the visible text.
• Remove sold units promptly, or mark them sold. An answer that sends a shopper to a sold car costs you that shopper's trust.
• State fees and what the advertised price includes, in text, next to the price.

The same discipline applies to e-commerce product pages, and our piece on why ChatGPT doesn't recommend your products covers the feed side in more detail.

Then: check the answers every month

Once the profile and inventory pages are fixed, check what the engines say. Write ten to twenty prompts a shopper in your market would ask: model plus "near me", certified pre-owned, trade-in, financing, service. Run them in ChatGPT, Gemini, Google's AI Mode and the others, and record which dealers are named and which pages are cited.

Our view: the 29% who have started¹ will be easy to spot in those answers within a quarter or two. The dealers who wait will find the answers already settled on a competitor's profile and inventory pages, and changing an established answer is slower than claiming an open one.