CarEdge says its AI car-buying agent has now run more than 159,000 negotiations with dealers. Agents like it build their comparisons from whatever a store has published online, so we wanted to see what one finds on a real dealer's website.
We ran this week's AutoIndustry.ai prompt on a Ford store we know well, shopping a new F-150 Lariat. We removed the store's name, location, VIN and phone numbers, and rounded every price.
The takeaway: Before an agent compares your price with anyone else's, it checks whether your price holds up from one page to the next.
What the agent reported back
The vehicle: A new F-150 Lariat 4x4 SuperCrew, with the VIN, stock number and a window sticker link right on the listing.
Price, fees and conditions:
MSRP of about $80,000, with a dealer discount and four manufacturer incentives itemized on the page
Advertised dealer price of about $72,000
Dealer and electronic filing fees of about $1,200, disclosed on the page, for a listed total of about $73,500
Fine print says some incentives may require manufacturer financing or a qualifying trade, without saying which ones: NOT FOUND
Fine print says added accessories may not be reflected in the price, with no add-on list shown: NOT FOUND
Estimated out-the-door price: About $78,000 after state sales tax and county surtax. Title and registration costs: NOT FOUND
Conflicts between sources:
The search results page listed this same VIN about $3,700 lower than the vehicle's own page.
One incentive built into the price expired the day of the test, and another expires within two weeks.
The page advertised a low promotional APR next to rebates the fine print says may be offered in lieu of special financing.
Recommendation: Contact this store early. It discloses fees upfront, which most stores don't. The first question to ask is which of the two prices is real.
The agent's note to the dealership:
Show one price for each vehicle on every page.
Label which incentives every buyer gets and which require financing or a trade.
List any dealer-installed add-ons and their prices on the vehicle page.
Why a consistent price comes before a low one
This store did something most don't. It put its fees on the page and showed exactly how it got from MSRP to its price. A shopper reading that page would likely trust it.
An agent reads every page, though. When it finds two prices for one truck, it has to stop and ask which one is real, and a store with one price across the site gets compared first. Incentives that expire mid-month create the same problem, because the price an agent quoted on Monday can be gone by Tuesday.
Run the same test on your own store
Paste this into ChatGPT, Claude or Gemini with web search turned on, then run it again on your closest competitor.
Act as an AI car-buying agent working for a shopper. Your client wants a [YEAR MAKE MODEL TRIM] and is considering [DEALERSHIP NAME] in [CITY, STATE].
Using only what you can find online right now (the dealership's website, Google Business Profile, third-party listings and reviews):
1. Find one in-stock vehicle that matches. Give the listing URL and VIN if shown.
2. Find the advertised price and every fee, add-on or condition attached to it. Note any incentive that expires within 14 days. Mark anything you can't confirm as "NOT FOUND."
3. Compare the price on the dealership's search results page with the price on the vehicle's own page. Flag any difference.
4. Estimate the out-the-door price for a buyer in ZIP [ZIP], line by line. State every assumption you make about tax, title and registration.
5. Find the best phone number and the fastest way to get a written quote.
6. List anything that conflicts between sources, such as prices, hours, phone numbers or availability.
7. Tell your client whether you'd contact this dealership first, and why.
Then switch roles. Write a short note to the dealership naming the three things that would have made your job easier.We ran our test with an AI assistant's search and page-reading tools, so your results will be worded differently and may catch more or less. Treat every "NOT FOUND" as something to check, since the AI can miss things too.
If an agent compared your search results page with your vehicle pages today, would it find one price or two?
