I Used AI to Buy a Car—Here’s Where It Works and Where It Doesn’t

Why I Turned to AI Before Stepping Onto a Car Lot

The last time I bought a new car was Thanksgiving 2010. That car came with a CD player, one USB port, and a very thick owner’s manual. Fifteen years later, enough parts were wearing out—or breaking—that repairs were starting to cost more than the car’s Kelley Blue Book value.

Car buying had clearly changed since then, and not just because dashboards now look like tablets. Pricing, inventory, and dealer communication had largely moved online, and comparing options felt more complicated than it used to be. So when it came time to start looking for a replacement, I turned to artificial intelligence (AI).

I wasn’t expecting AI to negotiate a deal for me. But with so many listings, features, fees, and decisions involved, I wanted to see how much AI could help.

How AI Helped Me Compare Car Prices and Dealers Early On

I used ChatGPT to compare cars at different dealerships in my area and to get ideas for what to ask dealers. Because I had such a good experience with my current car, I started my search looking at models that were most similar to what I already owned. The Volvo XC70 is no longer made, so I focused on the XC60 and the V60. I also explored a few other makes and models, from Audi to Hyundai to Subaru, but ultimately decided that a certified pre-owned (CPO) Volvo would be my preference if one was available.

After doing some web searches, I asked ChatGPT to help find cars at dealerships within a few hours of where I live in the Washington D.C. area. I specified the years and models, then narrowed results further by adding details such as mileage, trim line and features, and exterior colors. My own searches and ChatGPT’s suggestions yielded fairly similar results, though ChatGPT would occasionally surface a car that was older or more expensive than what I asked for.

Once I found a handful of vehicles that met my criteria, I asked ChatGPT for suggestions on what to ask when contacting dealers. It generated a list of questions, including:

• Can you share the CPO inspection checklist and a Carfax or AutoCheck report?
• When does the CPO warranty start, and what are the time and mileage limits?
• Are there any dealer add-ons included or required?
• What is the out-the-door price, including all taxes and fees?

I used that list alongside what I’d learned from past car-buying experiences and guidance from other sources—such as Consumer Reports, Consumers’ Checkbook, the Virginia Department of Motor Vehicles, the Federal Trade Commission (FTC), and CarEdge.com—when contacting dealers. I gathered prices from several dealerships and then leveraged them via email, phone, and text to get the best offer.

What AI Got Right—and Wrong—Once I Was at the Dealership

ChatGPT gave me talking points for when I went to the dealership in person. Because I had already received final pricing and used that information when choosing which dealer to buy from, I didn’t need much of the in-person negotiation advice it offered. Some of the suggestions included lines like, “Based on similar certified listings I’ve seen, I was expecting an out-the-door price closer to [$X]. Is there flexibility there?” and “If I finance through you, does that improve the out-the-door price?”

Where the guidance was more helpful was in the financing office. The advice there was straightforward: read everything carefully, ask questions as needed, skip add-ons or extended warranties I didn’t think I’d need, and avoid focusing only on low monthly payments that might depend on a longer loan term.

Other recommendations didn’t make as much sense in my situation. For example, ChatGPT suggested what to say if I felt pressured by a salesperson, including, “I don’t make same-day decisions without reviewing everything. If the deal works tomorrow, it works today, too.”

It also suggested asking about sourcing other vehicles through the dealer’s network. That can be useful when buying a new car, but it’s far less relevant when you’re shopping for a certified pre-owned car or other used vehicle. It also continued to recommend asking for pricing with all fees included even after I clarified that I had received out-the-door pricing before even getting to the dealership.

What I Learned About Using AI in the Car-Buying Process

AI was most useful as a research assistant at the start of the car-buying process, when I was researching models, comparing listings, and narrowing down options. It was reasonably accurate in helping me sort vehicles by price, mileage, features, and location, and its suggested questions to ask dealers were helpful as I compared out-the-door prices.

ChatGPT was less helpful in situations that required context and nuance. It sometimes suggested cars that didn’t fully match my criteria and repeatedly recommended negotiation steps I had already completed. Some advice for the in-person stage—like asking about sourcing additional vehicles—didn’t apply to buying a certified pre-owned car. 

Overall, ChatGPT worked best when I paired it with other research tools and my own judgment. Because I’d been through the car-buying process before, I could recognize when its advice was useful and when it wasn’t—though your mileage may vary depending on where you live, what you’re shopping for, and where you are in the process. In the future, I’d consider newer AI tools designed specifically for car buying—such as one from CarEdge that can contact dealers directly—to see whether more hands-on assistance would improve the experience.

While using AI didn’t guarantee a better price, it did help me stay organized, keep track of details, and feel less stressed.