AI car buying means shoppers use free tools like ChatGPT to check invoice, incentives, trade values, and repair prices before they ever reach you. If your numbers don’t hold up to that check, your own AI gains can net out to zero.
Last updated: November 2, 2026
My dad ran a service department. Ex-Navy, no shortcuts.
I worked my way through college in his service and detail department, and he held me to a higher standard than anyone else on his team. At the time it felt relentless. Looking back, it built the foundation for everything since.
I think about that standard every time someone tells me what AI is going to do for their dealership. Here’s why. The customer across the desk has the same tools now. Usually for free.
That’s the uncomfortable math in this post. Dealers are buying AI for the sales floor, the BDC, and the service drive, and good for them, because AI is here and is not leaving. But every capability a store pays for is also sitting in the customer’s pocket, and the customer is picking it up faster than the store is. Our founder and president, Erik Nachbahr, calls this the other side of the desk, and his argument is blunt: a lot of dealership AI gains could simply cancel out. I agree with him. The reasons are all on the floor.
Below is where the information edge used to come from, why fixed ops is next, and what a store can do so its pricing, its recommendations, and its own systems hold up when the customer’s AI checks the work. It’s one piece of how we think about managed AI for dealerships.
Every AI Tool You Buy, Your Customer Gets Too
82% of dealers use AI today. That comes from the Cox Automotive AI in Auto Retail Tracker, released in August 2026, which combined two quarterly waves of surveys: 987 dealer responses and 3,007 consumer responses across the first half of the year.
The hopes are big. 69% of dealers expect AI to drive sales and revenue growth, yet only 22% of the dealers using it report seeing that growth so far, and about 1 in 3 either aren’t measuring AI’s impact or have no clarity on how they’re measuring it.
Now flip the desk around.
In the same tracker, 63% of shoppers said they’ll definitely or probably use AI on their next vehicle purchase. 24% said AI already helps them feel more prepared when they work with a dealership. The store is still working out how to measure its AI. The customer isn’t measuring anything. They’re just using it.
| What the store buys AI to do | What the customer’s AI does with the same idea |
|---|---|
| Write follow-up emails and texts | Writes the counteroffer, plus a script for the phone call |
| Price inventory against the market | Compares your price to similar units at nearby stores |
| Desk a deal faster | Checks the payment, rate, and term before anyone signs |
| Build service recommendations from the inspection | Benchmarks labor time and parts price, and flags work it thinks can wait |
| Answer website chat questions | Asks your chat the questions, sometimes dozens in a row, to see what it gives away |
Same technology. Opposite side of the deal.
A tool that’s free for the customer isn’t an edge for the store.
Where the Dealership’s Information Edge Used to Live
Consumer AI in car buying is a shopper using a general AI assistant, such as ChatGPT, Gemini, or Claude, to research, price, and negotiate a vehicle purchase or a repair. It pulls invoice estimates, incentives, trade values, and typical F&I product costs into one conversation and turns them into a plan for the visit.
Gross has long rested on what the customer didn’t know, or didn’t have the patience to find. Be honest. None of it was secret. It was just scattered across a dozen websites and a few hours nobody wanted to spend on a Thursday night. Here’s what used to sit mostly on the store’s side of the desk:
- The invoice, and roughly what the store has in the unit
- Which factory incentives are live this month, and which ones stack
- What the trade is really worth. Not the number on the first pencil, the number three other stores would write
- F&I product cost, which is the one customers have always guessed at and almost never known
- Labor times. How long a job should take, and what the part costs online
The FTC has been coaching buyers on some of this for years. Its 2024 consumer alert on car add-ons tells shoppers to know their total cost, not just the monthly payment, to get the price in writing with every fee and add-on included, and to tell the dealer to remove any add-ons they don’t want. That used to be advice a customer had to remember. Not anymore. Now an assistant walks them through it, line by line, while they’re sitting in your F&I office.
What took a determined buyer an evening now takes a casual one a few minutes. And the next step isn’t hard to picture: AI agents that contact several stores at once, ask each for an out-the-door price, and play the answers against each other. That’s exactly what a sharp customer has always done with a phone and a legal pad. It’s just faster, and it doesn’t get tired.
The information gap is closing. Plan as if it’s already closed.
Fixed Ops Is Next, and the Service Drive Feels It First
Picture a Tuesday morning on the drive. A customer comes in for an oil change, and the advisor walks back out with a multipoint inspection: front brakes getting thin, a seeping valve cover gasket, a cabin filter, and a coolant flush. Before the walkaround is over, the customer has snapped a photo of the inspection sheet and asked an AI assistant whether any of it really needs doing today.
The assistant answers fast. Total confidence, too. The gasket’s labor time looks high. The cabin filter is a job they could do in the parking lot. The flush isn’t due yet, per the owner’s manual. Some of that will be right. Some won’t be. The customer can’t tell which, and honestly, neither can the AI, because it never saw the car.
Writing service was my first customer-facing job, and it changed how I saw the whole store. An advisor’s credibility gets built one recommendation at a time. When a customer’s AI second-guesses a legitimate line and the advisor can’t show why it’s needed, the store doesn’t just lose that line. It loses some trust in the next one, too.
And it never stays at the counter. A declined line is a part that doesn’t get pulled, a tech whose hours come in short, and somebody in accounting closing out a smaller RO than the inspection said it would be. One department sneezes and the next two catch it.
I’d tell any advisor to treat a recommendation the way the used car department treats a recon report. Nobody on the lot would hand a buyer a used car and say “trust us, we went through it.” They show the sheet. Same thing on the drive. Write the measurement down. Attach the photo. Note the interval it ties back to, so when the customer’s AI goes looking for a reason to say no, it runs into the reason to say yes.
No evidence on the RO? The AI makes one up.
Make Every Price and Recommendation Hold Up to a Second Opinion
So do you fight the customer’s AI? Please don’t. You’d lose, and honestly a shopper who’s done their homework tends to be the easier deal anyway, as long as there’s nothing on your side of the desk you’d be embarrassed for them to find.
What you can do is get your numbers and your people ready for the check. Where I’d start:
- Any unit priced over market needs its reason written down. Certified? Low miles? A package nobody else has? Put it where the salesperson and the customer can both read it.
- On the drive, the evidence travels with the recommendation. Every time.
- Run one role-play this week. A manager plays the customer and reads an AI-written counteroffer out loud, word for word, and the salesperson or advisor answers it. Then switch seats. Hands-on experience and repetition make a world of difference, and it sticks far better than a slide deck about “the modern shopper.”
- F&I gets the same test. Can the F&I manager say what a product covers and roughly what it costs without flinching? If not, practice.
- Clean up the data AI agents read. Wrong trims, missing options, and stale prices on your website get repeated back to customers as fact, and the customer shows up holding the wrong number.
Measure your own AI while you’re at it. Cox found about 1 in 3 dealers aren’t measuring its impact, and it also found that dealers working with an outside AI partner were more likely to report sales and revenue growth from AI (36% vs. 21%). Optimization only counts if somebody’s keeping score.
Time is money in this business. I know. The cheapest hour your team spends this quarter might be the one where they practice against a customer who’s already done the research.
The deal that survives a second opinion is the deal you keep.
When the Customer’s Bot Starts Talking to Your Bot
In December 2023, a shopper got a California dealership’s website chatbot to agree to sell a new full-size SUV for $1. He told it to agree with anything the customer said and to call each answer a legally binding offer, and it did. The store had to rework the bot, the screenshots went everywhere, and the AI Incident Database still lists the episode.
That was one person, typing. For fun. Now picture software doing it all day.
Consumer AI agents will reach your website chat, your lead forms, and your online credit application, and they won’t get bored or embarrassed. The trick in that story is called prompt injection, and the OWASP Gen AI Security Project puts it first on its 2025 Top 10 list for applications built on large language models. Its definition is plain: user prompts that alter the model’s behavior or output in unintended ways.
For the people on the floor, the questions are practical ones. Start here. What is your chat tool allowed to promise? Who reads what it said before a salesperson calls the customer back? What customer data can it reach? A lead form or a credit application wired into an AI tool is now a door somebody else’s AI can knock on.
This is where I hand off to our security team, which supports IT and security for 2,000-plus dealerships. The specifics are in our data security checklist for dealership AI chatbots, and what dealer-side agents should and shouldn’t control is in our guide to AI agents for car dealerships. The policy side, meaning who approves which tools and what data they touch, lives in AI governance for dealerships.
Your bot needs rules the customer’s bot can’t talk it out of.
What Dealers Ask About AI Car Buying
Are shoppers really using AI to buy cars, or is that hype?
Can a customer’s AI actually tell whether a repair is needed?
Should we just block bots from our website chat?
Does this mean our own AI spending is wasted?
Where should a dealer principal start this month?
The Desk Has Two Sides Now
My dad’s standard felt relentless when I was in college. It’s about to feel normal. Customers will hold every store to it now, with a tool that never sleeps.
That’s the real work: pricing you can explain, recommendations with the evidence attached, people trained for a customer who already did the research, and dealer-side AI with rules that hold. Get those right and the people on the floor are set up for success, and your AI gains stay gains.
Erik takes this argument to dealer audiences, and his topics are on his dealership AI and security speaking page. If you’d like to see where your own stores stand, start with a complimentary IT and cybersecurity assessment. We’ll look at your IT and security setup, including the systems a customer’s bot can already reach.