Can You Use AI to Help You Buy a Car?

“Generative AI tools are impressive. They can digest a lot of information, they can be a fun way to get started, but I would continue to think of them as the friend-of-a-friend that you had an interesting but somewhat unhinged conversation with at a party,” says Tracy Anderman, Consumer Reports’ head of fact-checking, who develops Consumer Reports’ own policies around AI usage. “You’re not going to spend tens of thousands of dollars on a car just because an acquaintance says so. You verify their guidance with other experts, then confirm with your own hands-on experience, in this case, with a test drive. ” 

With those limitations in mind, is there a place in the car-buying process for AI tools? To find out, I used three of the most popular AI models, asking each one a simple and popular question in plain language: What’s the most reliable 3-row SUV? Then, I put the answers up against AskCR, Consumer Reports’ AI adviser for members, and my own car-buying expertise. Then, I interviewed experts in machine learning, prompt writing, and fact-checking about AI’s role for car buyers. I also asked them why each model made the mistakes it did, how I could’ve improved my prompts, and what shoppers should know before asking AI to help them make a major purchasing decision.

I found:

• AI tools like ChatGPT, Claude, and Gemini can be helpful assistants for car buyers, but they’re no replacement for critical thinking.
• Popular AI tools often muddle specific vehicle recommendations, even suggesting models that don’t exist or citing made-up accolades.
• To get the most out of AI tools, don’t use them like a search engine. Write specific prompts, and validate the results against trusted primary sources, such as trusted news sites and manufacturer websites.

The AI tools also hallucinated and made serious errors. Gemini made up options that didn’t exist. ChatGPT recommended cars that don’t exist, including a “2026 Cadillac XT6” (the model was discontinued in 2025 for the U.S. market) and a “Lexus TX350h” (only the TX550 comes as a hybrid). All three AI tools produced impressive-looking vehicle comparison charts, but they frequently mixed up 2026, 2025, and even 2024 model year SUVs, which was especially concerning since one of the vehicles had recently been redesigned. ChatGPT sometimes showed images of the wrong vehicles. Claude ranked unreliable vehicles ahead of more reliable ones and also struggled with model years.

The errors show that AI makes mistakes. “AI is a decision-support tool, not a substitute for critical thinking or human judgment,” says Suman Veeramalla, Consumer Reports’ chief data officer. That’s especially true for a major decision like a car purchase. "AI can inform your decision, but it cannot make it for you," he says. "Don’t let AI replace your judgment."

The tools occasionally referenced Reddit threads and forums where owners complained about problems with their vehicles and claimed those issues were common, even when our reliability surveys show they are relatively rare.

Some responses even incorrectly summarized older Consumer Reports articles I had written, even though the data is now long out of date. All three tools took marketing materials from dealerships and automakers as facts, and used advertising-speak such as “rugged,” “most premium,” and “exclusive.”

“AI tools are not always able to distinguish between what’s credible information and what isn’t,” says Anderman. “They can just be flat out wrong. They’ll be really nice to you about it, but they can still be flat out wrong.”

For example, ChatGPT sourced some vehicle specifications from a website that, when I clicked on it, simultaneously advertised itself as a church, a clothing retailer, a cryptocurrency exchange, a pawn shop, and a zombie response team. Anderman recommends using more traditional research methods to check every source, especially for an important purchase like a car. “You want to come out of the system to verify the information,” she says. I found that asking the tools to rely only on primary sources improved the quality of results.

Birss says that generative AI tools are most useful when you give them specific requests. “When the machine has to make assumptions, the less chance there is of getting good information,” he says. So, instead of asking for the “best” three-row SUV, ask for one that fits in a certain price range, that scores well in Insurance Institute for Highway Safety tests, that has the number of cup holders you want, and that gets top reliability ratings from Consumer Reports.

Some of the best uses of AI leverage its ability to generate language. When you know the make, model, and trim of the car that you want to purchase, free AI tools can write scripts that can counter common high-pressure sales techniques and make you sound like a master negotiator. There are even paid services that use generative AI to negotiate with salespeople over email on your behalf, such as CarEdge.

If you don’t know very much about cars, Birss says you can use AI tools to interview you about your wants and needs and help you figure out what you want. That’s what Helena McAleer, Birss’ business partner, did earlier this year. She ended up purchasing a car that wasn’t on her radar initially, but that she is very happy with, and that works well for her needs. “It was a well-thought-out, well-assessed plan that’s going to save me money in the future,” she says.