Copart is a colossal, international car salvage business, and right now, business is booming. The company reported that revenue, gross profit, and net income are all up for both the quarter and the year. On a recent earnings call, analysts and Copart CEO Jeff Liaw shared insights on how the auto sale, salvage, and auction outfit is leveraging AI.
In Q4 of 2025, I think every large-scale business is messing with large language models to some degree. For an operation like Copart’s, which deals with a huge volume of transactions, many of which are similar, the deployment of AI especially makes sense. At this stage, it sounds like, rather than using tools to physically inspect cars like we’ve seen in the rental car world, Copart is just using it to do things like price vehicles and decide whether or not a car is totaled based on data.
In practice, that means processing the cost of estimated repairs versus vehicle value far more rapidly than such equations could have been done before.
What’s interesting about that, to me, is that this means AI could and will end up influencing insurance rates and car values around the world in general. In fact, to a degree, it will affect the global distribution of used cars.
See, while the severity of car accidents has reportedly been declining, the rate at which cars are totaled out has been going up. Copart’s people attribute that to the increasing complexity of modern cars—they become uneconomical to repair more quickly than their predecessors. Case in point: A 2024 F-150 taillight lists for over $1,000. “…total loss frequency has generally increased at a rate far exceeding the decline in accident frequency itself. In fact, for the quarter, total loss frequency has continued its long-term upward trend, consistent with, again, the entire history of our company. In the U.S., total loss frequency for the second calendar quarter of 2025 was 22.2%, up from 21.5% in the same quarter in 2024,” said Liaw.
Cars that are totaled, largely, get sent overseas, where less-expensive labor and a fresh title make their restoration more financially logical. “International members account for approximately 40% of all vehicles sold at Copart’s U.S. auctions, comprising almost half of auction proceeds because international buyers generally purchase vehicles that are more valuable than those acquired by domestic buyers,” Liaw remarked at another point in the call.
Wholesale car auctions like Copart have a unique relationship with sales data and the repair viability of cars, because they kind of simultaneously react to, and set, those values. For example, Copart knows how much a 2005 Honda Civic LX with 150,000 miles is worth because it has facilitated the sale of thousands of them. Buyers base what a good deal is on the auction’s average transaction prices, and that same number dictates at what point a car is worth repairing versus totalling. It’s kind of a self-feeding loop.
While none of this is mind-blowing or new, I find it intriguing to think about AI supercharging the speed and scale at which all this is processed, and by doing so, influencing what all our cars are worth. Since AI will be able to evaluate cars with more data than anyone ever before, it will become the ultimate arbiter of vehicle value, guiding what gets totaled and what gets repaired. And, ultimately, what it costs to insure certain models.
Algorithms and averages were already doing that before LLMs became buzzwords, but now everything will be happening more quickly.
Here’s the extent of what Mr. Liaw had to say about Copart’s usage of AI, prompted by a question from CJS Securities/Baird analysts Bob Labick and Chris Bottiglieri:
“…we equip many of our sellers with tools to allow them to make instantaneous total loss decisions informed by literally millions of similar vehicles we’ve sold over the years. Those decision support tools are very much empowered by current generation large language model technologies. Beyond that, certainly in the obvious arenas, such as customer support and also in agent support here at Copart.”
“Even the folks who still are interacting day-to-day with members and buyers are equipped with better information with LLM behind the scenes and on and on. I think we are still, like many companies, in the early stages, but have many different arenas in which we have that technology deployed today. We also have it at the auction level. As I think about the products and the vehicles that we’re recommending to our buyers, we are search results, et cetera. All of these different domains are informed by AI as it stands now. I have no doubt that as the tools themselves improve and as our deployments become still more sophisticated, that it’ll enhance business as it is. It’ll make us radically more efficient in delivering the service as we deliver it today. No doubt it’ll unlock future opportunities as well.”
“I think it’s fair to say it is both allowing us to do what we do more efficiently, compressing cycle times for our clients, compressing cycle times for us. You heard Leah describe that phenomenon when it comes to inventory. One of the reasons that inventory contracts is that our Title Express offering for which we are providing this service of procuring the original titles on behalf of our insurance clients and doing so for more and more clients with each passing quarter. That is, we’re yielding better cycle times there in part because when we do the work, it’s enabled by our tech stack, our LLM deployments in ways that are more efficient than the insurance carriers before they transfer that responsibility over to us. Great question. No doubt that answer will evolve over the quarters and years to come.”
Liaw also remarked that the company could run a whole presentation on its AI use specifically, so I guess we can stay tuned for that one if you’re interested in the role of LLMs in the auto salvage business.
Got a tip? Drop us a line at tips@thedrive.com.
Automotive journalist since 2013, Andrew primarily coordinates features, sponsored content, and multi-departmental initiatives at The Drive.
