Shoppers Are Adopting AI Faster Than Dealerships Are Adjusting. Cox Just Put Numbers on It.

The industry finally has a baseline - and it is useful precisely because it is unflattering

On August 11, Cox Automotive published the first edition of a new recurring study, the AI in Auto Retail Tracker - a survey instrument that will measure AI adoption, impact, and attitudes on both sides of the car deal, quarter after quarter. The first read combines Q1 and Q2 2026 results: roughly 1,000 dealership decision-makers (504 in Q1, 483 in Q2) and just over 3,000 in-market shoppers who plan to buy a vehicle within 12 months.

One-off vendor surveys about dealership AI arrive weekly. A recurring tracker from the industry's largest data company is different, because it creates a baseline that future quarters get graded against - and the baseline it just established is a gap, in three places at once. Shoppers are ahead of stores. Adoption is ahead of measurement. And stores working with an outside AI partner are reporting meaningfully better outcomes than stores going it alone.

The myth: With 82% of dealers now using AI, the industry has adjusted to AI-driven car shopping. The data: Only 29% of dealers have started adjusting their strategy to account for AI-powered search behavior - and the share who know they need to adjust but have not started rose from 26% to 32% between quarters. Source: Cox Automotive AI in Auto Retail Tracker, August 2026.

Cox, it should be said, sells AI-powered retail software - it closed its acquisition of the AI marketing platform Fullpath this year - so it has an interest in dealers concluding they need help. That is worth keeping in mind with any vendor-adjacent research. But the sample sizes are real, the survey covers both sides of the counter, and the least flattering findings cut against the industry Cox sells to. It reads more like a measurement exercise than a pitch.

Gap one: the shopper is moving faster than the store

The consumer numbers are the ones a dealer principal should sit with. Sixty-three percent of in-market shoppers say they will definitely or probably use AI on their next vehicle purchase. Twenty-four percent say AI already helps them feel more prepared when working with dealerships. Cox's framing is blunt: AI is changing where car shopping starts, and it is delivering dealers a more informed, more decisive buyer.

This is not happening in an automotive vacuum. The same day Cox published its tracker, Google announced that its Gemini app had crossed 1 billion monthly users - its fastest-growing product ever - and said 63% of those users interact with it by voice. (Google did not detail how it measured either figure, a caveat worth noting.) The precise numbers matter less than the direction: talking to an AI assistant, out loud, is now an ordinary consumer behavior, not an early-adopter one. The person researching a crossover with an AI assistant on Tuesday night is the same person who calls your store on Wednesday.

What does an AI-prepared shopper look like when they reach you? The evidence so far suggests: further down the funnel, with a narrower consideration set, and less patient with friction. They have already compared trims, checked payment math, and read reviews in aggregate. We looked at how AI research is narrowing shopper consideration sets and what it takes to stay visible when AI does the searching. The tracker adds the missing quantification: this is now most shoppers, and the store-side response has barely begun.

Gap two: adoption is ahead of measurement

The dealer-side numbers look strong until you ask what the AI is producing. Eighty-two percent of dealers report using AI today. The most common uses are automating routine or complex tasks (40%), coordinating customer follow-up (40%), and generating content and creative (38%). Sixty-nine percent expect AI to drive sales and revenue growth.

Then the funnel narrows sharply. About one in three dealers either are not measuring AI's impact at all or have no clarity on how they are measuring it. And only 22% of dealers using AI report actually seeing sales and revenue growth from it so far.

That 69-versus-22 spread is the most important pair of numbers in the report. It does not say AI fails to produce results. It says most stores have no instrument that would tell them either way. A tool that answers chats and sends follow-ups produces activity by default and outcomes only when someone defines them - which is the same conclusion the industry's execution-gap research reached from the vendor side in July, and it is why we keep arguing that dealership AI should be judged on outcomes, not conversations handled.

The practical translation: if your store adopted AI in the last two years and cannot state what it changed - in answered calls, logged leads, booked appointments, or gross - you are in the industry's largest cohort. The tracker just made that cohort visible.

Gap three: partnered stores report better outcomes than DIY

The third finding will get quoted in every AI vendor's deck this fall, so it is worth stating carefully. Dealers working with an external AI partner were more likely to say they are using AI optimally (66%, versus 46% of those without a partner), more likely to express high confidence in AI outputs (30% versus 8%), and more likely to report sales and revenue growth from AI (36% versus 21%).

Source: Cox Automotive AI in Auto Retail Tracker, combined Q1+Q2 2026 base. Figures are dealer self-reports.

Two honest caveats. These are self-reported outcomes, not audited financials. And there is a selection effect lurking: stores disciplined enough to engage a specialist partner are often disciplined about process and measurement generally, and that discipline - not the partner alone - drives results. Even so, the deltas are large and consistent across three separate questions, and the confidence gap (30% versus 8%) is hard to explain away. Running AI as a side project of an already-stretched staff appears to produce tools that are turned on but not trusted, and not measured.

What this means for the phone room specifically

Pull the three gaps together and they converge on the store's first point of live contact - which, for high-intent shoppers, is still overwhelmingly the phone.

An AI-prepared buyer makes fewer, later, higher-stakes calls. Someone who has already narrowed the list to two vehicles and knows the payment range is not making six exploratory calls; they are making one or two decisive ones. Every mishandled call in that world costs more than it did when shoppers cast wider nets - and the industry already loses an alarming share of these. The benchmark data on what missed calls cost and how fast leads go cold was concerning before the caller showed up pre-researched.

The measurement gap has a phone-shaped answer too, because phones are the easiest place in the store to instrument honestly. If the tracker's one-in-three number describes you, start where the metrics are unambiguous: answer rate, after-hours miss rate, appointments booked and kept, share of calls logged to the CRM with complete records. Those numbers exist whether or not you measure them. A store that instruments its phones for ninety days will know more about its AI's real contribution than most of the 82% currently know about their entire stack.

The measured takeaway

The right response to Cox's first tracker read is not to buy more AI. Adoption is plainly not the constraint - 82% of the industry has done that part. The constraint is alignment and instrumentation: aligning with how shoppers now start (AI-assisted, better informed, more decisive) and instrumenting whatever AI you run so that 69% expectation and 22% observed results can be reconciled in your own store rather than taken on faith.

A reasonable 90-day plan looks like this: pick the shopper-facing workflow where the gap bites first - for most stores, phone coverage and first response. Establish the baseline before changing anything. Deploy or tune one tool against that baseline, with CRM logging non-negotiable. And if nobody in the building owns AI measurement, take the partner finding seriously rather than personally - the data suggests the stores getting results are the ones that stopped treating AI as a side project.

The tracker will publish again next quarter. The useful question for any GM is which side of its numbers your store will be on by then.

If the first workflow you want instrumented is the phone - every call answered, every appointment booked, every lead logged, with numbers you can audit - that is what Carbuki builds for U.S. dealerships, and pressure-testing the tracker's findings against your own store's baseline is a conversation we are glad to have.

Sources

• Cox Automotive, "New Cox Automotive AI in Auto Retail Tracker Finds Growing Gap Between Dealers and AI-Powered Car Shoppers," August 11, 2026: https://www.coxautoinc.com/press-releases/new-cox-automotive-ai-in-auto-retail-tracker/
• Cox Automotive, AI in Auto Retail Tracker (study page): https://www.coxautoinc.com/retail/ai-in-auto-retail-tracker/
• Auto Remarketing, "Cox Automotive dealer survey: 82% use AI but 'still early in the journey'," August 2026: https://www.autoremarketing.com/ar/retail/cox-automotive-dealer-survey-82-use-ai-but-still-early-in-the-journey/
• Car Dealership Guy News, "Cox study finds 63% of car shoppers plan to use AI, but dealers are falling behind," August 11, 2026: https://news.dealershipguy.com/p/cox-study-finds-63-of-car-shoppers-plan-to-use-ai-but-dealers-are-falling-behind-2026-08-11
• Google, "Gemini app hits 1 billion monthly active users," August 11, 2026: https://blog.google/innovation-and-ai/products/gemini-app/one-billion-monthly-users/
• Forbes, "Gemini Becomes Google's Fastest-Growing Product Ever After Hitting 1 Billion Monthly Users," August 11, 2026: https://www.forbes.com/sites/antoniopequenoiv/2026/08/11/gemini-becomes-googles-fastest-growing-product-ever-after-hitting-1-billion-monthly-users/