CarMax put 45,000 cars inside ChatGPT on February 27. You can type "find me an SUV with a third row under $25K" and get real inventory results without leaving the chat window.
Two weeks later, Automotive News profiled buyers who'd already been building their own versions. Custom AI agents stitched together from ChatGPT, Reddit threads, and VIN lookup tools like Visor.vin. One buyer built an agent that evaluated vehicles across multiple data sources before he talked to a single salesperson.
These aren't early adopters anymore. They're becoming the norm.
The numbers
According to Ekho's 2026 AI Vehicle Research Study, 30% of car shoppers now use a generative AI tool during their search. A Cars.com survey puts it higher: 44% of consumers have used AI-powered search features on car marketplaces. Among AI users, ChatGPT accounts for 68.4% of all usage, per Ekho. More than every other AI tool combined.
A CarGurus survey of 3,000 recent buyers found the top uses: comparing vehicles (44% of respondents), finding listings (40%), comparing specs (33%), and checking safety ratings and reviews (31%).
97% of AI users in the Cars.com survey said the technology will influence their next purchase decision. That's not curiosity. That's a behavioral shift.
Most people are just Googling faster
For all the headlines, most AI car shopping looks like this: someone types "best used SUV under $20K for a family" into ChatGPT instead of Google. They get a more conversational answer. Maybe they ask a follow-up. Then they go to Autotrader or CarGurus and do the same search they would've done anyway.
It's useful. 73% of AI users told Cars.com the technology saves them time by turning conversational queries into targeted results. But it's the same information, delivered differently.
The shift worth watching is happening at the edges. Ross Tinkham, VP of Automotive at Podium, described the trajectory: AI will let consumers "shop 100 dealerships simultaneously." CarEdge is already piloting an AI negotiator that contacts dealers anonymously and handles 10 conversations at once. Capital One launched an AI agent covering everything from vehicle comparison to financing.
These tools don't just research. They act. That's a meaningful difference from asking ChatGPT for a recommendation.
The bias problem
When CarMax puts its inventory inside ChatGPT, the results come from CarMax's 45,000 vehicles. When a dealership deploys an AI chatbot, it surfaces that dealership's stock. Every AI car-shopping tool is funded by someone, and the buyer doesn't always know whose interests it serves.
63% of shoppers in the Cars.com survey said they're concerned AI tools could recommend vehicles in a biased way. That instinct is right. An AI tool backed by a marketplace or dealer has incentives built into every recommendation.
The other gap is pricing data. Ask ChatGPT "is $24,000 fair for a 2023 Camry with 35K miles?" and you'll get a generic answer drawn from training data. It doesn't know what Camrys are selling for in your zip code this week. It doesn't know that specific listing has been on the lot for 58 days, which changes your negotiating position completely.
Gen Z is already all in
Half of Gen Z car buyers made their last purchase entirely online, per a report covered by Digital Dealer. 65% plan to buy their next vehicle online too. 79% want AI to recommend the best car for their needs. 74% want AI to tell them the optimal time to buy based on price fluctuations.
34% of Gen Z would order a vehicle entirely online, zero dealership touchpoints, according to a Snapchat/Publicis study. Buyers over 45? Only 19% would consider it.
And they're stacking AI with social media. 74% of Gen Z use social during the car shopping process, with 61% trusting creator recommendations. This generation treats car buying the way they treat everything else: research digitally, compare obsessively, show up in person only when they already know what they want and what it should cost.
What separates useful AI from noise
The pattern across all these numbers: AI is becoming the front door for car research. Most buyers walk through that door and get the same generic answers everyone else gets.
The buyers who come out ahead are pairing AI with real market data. Days on market by model. Pricing trends by region. Inventory supply levels. That context turns a ChatGPT conversation from "the RAV4 is generally well-regarded" into "RAV4s in your area are averaging 52 days on market, and the seller is probably motivated."
CarScout's market pages track this kind of data across makes and models from live sources. That's the layer most AI tools are missing: not what a car is worth in theory, but what it's actually selling for where you live, right now.
AI is reshaping how people find cars. It hasn't changed who gets the better deal. That still comes down to showing up with real numbers.