100 AI Agents Mystery-Shopped 100 Dealers. The Real Story Is What Happens Next

CarEdge's live AI mystery shop was not a stunt. It was a preview of the next customer.

By Dealer AI Guy. Published June 7, 2026. Based on Zach Shefska's public recap, AutoIndustry.ai Summit coverage, ASOTU coverage, CarEdge Dealer Transparency Index materials, and FTC pricing-transparency enforcement activity.

TL;DR

CarEdge's Zach Shefska sent AI agents to mystery-shop 100 dealerships, live from the stage at the AutoIndustry.ai Summit.
Within 24 hours, 92 stores replied and 62 gave some price information. Only 33 delivered the full out-the-door breakdown the agents asked for.
AI shoppers turn your internet lead into an auditable data trail. Every dodge, fee, and mismatched quote gets logged, compared, and remembered.
CarEdge is feeding those verified quotes into a Dealer Transparency Index that scores stores on fees, add-ons, and markup.
The winners will be stores that answer fast with clean, itemized numbers that match the listing. Template-and-stall now costs deals.

At the first AutoIndustry.ai Summit, Zach Shefska of CarEdge did something that should make every dealer operator stop and audit their internet process.

He did not show a slide about AI.

He sent AI agents to mystery-shop 100 dealerships while he was on stage in a room full of dealers.

The agents contacted stores by email and SMS. They asked for out-the-door pricing. They tracked response behavior. They captured structured data. They negotiated add-ons. They created receipts.

According to Shefska's public recap, within 24 hours:

92 of the 100 dealers replied.
62 provided some price information.
Only 33 provided the full requested out-the-door price breakdown.
Average response time was about one hour.

That is a strong response rate. It also exposes the new problem.

The question is no longer, "Did the dealership answer the lead?"

The question is, "Did the dealership answer in a way that a machine can verify, compare, score, and remember?"

That is a much bigger shift.

What Actually Changed

For 25 years, digital retail was mostly about the shopper finding information.

Search inventory. Compare prices. Read reviews. Submit a lead. Wait for a human. Negotiate through friction.

AI agents change the sequence.

The shopper no longer has to personally chase the dealership for clarity. The agent can do it.

It can contact ten stores. Ask the same question every time. Save every reply. Parse every fee. Compare every out-the-door quote. Identify which quote matches the advertised price and which one quietly grew after contact.

That turns the internet lead from a sales conversation into an auditable transaction trail.

CarEdge's Dealer Transparency Index is the clearest early example. CarEdge says its ratings are based on verified out-the-door quotes collected through its AI-powered buying agents. The methodology scores dealers on fee transparency, add-on behavior, dealer markup, and data quality.

That matters because the agent is not just helping one buyer. It is building a memory layer over the market.

Why This Is Significant For Dealers

Most dealerships still treat the internet lead as a human conversation with a messy handoff.

An AI shopper treats it like a data extraction task.

That mismatch is where the pain will show up.

If the VDP says one price, the autoresponder says another, the salesperson replies with a payment instead of an out-the-door number, and F&I introduces mandatory products later, the AI agent will not experience that as "the normal process."

It will log it as inconsistency.

It will ask again.

It will compare the store against competitors.

It may negotiate the add-ons down.

It may route the shopper away from the store entirely.

And increasingly, it may feed that result into a public or private score.

This is the new dealer visibility problem. It is not only whether Google sees your website. It is whether buyer-side agents can trust your answers.

The Dealer Perspective

From the dealer side, this feels uncomfortable because the process was not built for machines.

A human shopper may tolerate ambiguity for a while. They may accept a vague answer if the salesperson is friendly. They may come in because the car is close, the store has the right color, or the payment sounds plausible.

An AI agent has no emotional sunk cost.

It has a job to do.

It wants a written quote, itemized fees, add-on status, rebate assumptions, taxes, title, registration, doc fee, availability, and expiration terms. It wants the quote to match the listing. It wants the dealership's answer to survive comparison.

For a dealer, that means several old habits become more expensive:

Advertising a number that does not match the real selling path.
Hiding doc fees until the buyer asks.
Bundling add-ons without making them optional and explicit.
Sending "when can you come in?" instead of the price requested.
Using templates that dodge direct questions.
Letting CRM, website, DMS, F&I, and desk process tell different versions of the same deal.

The upside is just as important.

Transparent dealers get a new way to win.

If your store can answer quickly, quote cleanly, disclose fees, explain products, and honor the written number, AI agents will reward you. They will not get tired. They will not forget. They will not be impressed by vague charm. They will keep picking the cleaner transaction.

That is bad news for stores that rely on friction.

It is good news for stores that already operate cleanly and have not had a good way to prove it.

The AI Agent Perspective

Now look at the same event from the agent's side.

The agent is not "shopping" the way a person shops.

It is reducing uncertainty.

It begins with a target vehicle and a buyer objective. It reads the listing. It checks the advertised price. It contacts the store. It asks for the full out-the-door price for the buyer's ZIP code. It requests add-on status. It parses the reply. It compares the quote to the listing. It asks follow-up questions. It logs the result.

The agent will prefer the dealership that gives it clean data.

That means the best dealership response is not necessarily the longest, warmest, or most salesy response. It is the response with the least ambiguity.

From the agent's perspective, a great dealer reply looks like this:

The vehicle is confirmed available.
The selling price is clear.
Taxes and government fees are separated from dealer fees.
Doc fee is disclosed.
Add-ons are listed by name, price, and whether they are optional.
Rebates and financing assumptions are explicit.
The out-the-door price is written in a way that can be saved and compared.
The quote expiration is clear.
The next step is offered without withholding the requested information.

That is the language AI agents can act on.

The Regulatory Backdrop

This is not happening in a vacuum.

In March 2026, the FTC warned 97 auto dealership groups about deceptive pricing. The agency specifically called out advertised prices that do not reflect mandatory fees, discounts not available to all consumers, financing-conditioned prices, required add-ons not reflected in the advertised price, and unavailable vehicles.

The important point for dealers is not that every AI mystery shop is a regulatory event.

The point is that the same categories regulators care about are the categories agents can now document at scale.

An AI agent can preserve the advertised price, the written quote, the timestamp, the add-ons, the doc fee, the response delay, and the follow-up thread. That turns a pricing complaint from a story into a file.

That is why this moment matters.

Consumer AI agents are not just convenience tools. They are evidence machines.

Prediction: The Next 12 Months

Within one year, AI mystery shopping will become normal.

Not exotic. Not conference-stage theater. Normal.

Consumers will ask AI assistants to gather out-the-door prices from five or ten dealers before they ever speak to a salesperson. Car-buying services will run agent fleets across dealer groups. OEMs, agencies, and consultants will use AI shops to audit stores.

The first wave of dealer adaptation will be tactical:

Better response templates for AI-assisted shoppers.
Faster written out-the-door quotes.
Clearer doc fee and add-on disclosure.
More internal mystery shops.
More pressure on website vendors to make pricing assumptions explicit.
More attention to whether advertised price, CRM quote, and final buyer's order match.

Some dealers will try to block or ignore AI shoppers.

That will be a mistake.

The agent does not need every store to cooperate. It only needs enough stores to produce a ranked recommendation.

Prediction: The Next Two Years

Within two years, the market will split between agent-ready stores and agent-hostile stores.

Agent-ready stores will build a deliberate response path:

AI lead classification in the CRM.
Quote templates designed for machine readability.
Clear escalation rules for negotiation.
Manager-approved add-on language.
Inventory and pricing feeds that line up with what the store is willing to quote.
Compliance reviews of written digital responses.

On the other side, agent-hostile stores will keep trying to force every buyer into the showroom before answering basic questions.

That may still work with some human shoppers.

It will work less often with agents.

The bigger change will be dealer-side AI. Stores will deploy their own agents to respond to buyer-side agents, verify inventory, generate compliant quote packets, and hand off only high-intent shoppers to humans.

That sounds strange until you realize the current alternative is a salesperson manually answering the same pricing question hundreds of times under time pressure, with inconsistent information, inside a CRM workflow that was designed before agentic shopping existed.

The next two years will be the rise of agent-to-agent retail.

The human will still matter. But the human will increasingly enter after the deal structure is already cleaner.

Prediction: The Next Five Years

Within five years, AI agents will be part of the default car-buying workflow.

The buyer may never say, "I am using an agent."

They will just expect their assistant to:

Find the right inventory.
Eliminate stores with poor transparency.
Request and compare out-the-door quotes.
Check incentives and financing assumptions.
Flag suspicious fees.
Negotiate unwanted add-ons.
Schedule the test drive or delivery.
Preserve the paperwork trail.

At that point, dealer competition changes.

The stores that win will not simply be the stores with the lowest advertised number. They will be the stores with the most trusted transaction path.

Trust will become machine-readable.

That means clean data, consistent pricing, fast quote response, transparent add-on handling, strong reviews, accurate inventory, and clear local authority will all compound.

The worst position will be a dealership that is invisible to AI search and untrusted by AI shoppers.

That store will not just lose leads. It will be filtered out before the customer ever knows its name.

What Dealers Should Do Now

The right response is not panic.

It is instrumentation.

Dealers should treat this as a readiness audit across sales, marketing, compliance, CRM, website, and F&I.

Start with these moves:

Mystery-shop yourself with AI. Ask an agent to contact your store like a buyer and request an out-the-door quote. Save every reply. Compare it against your advertised price.
Standardize the written OTD response. Build a quote format that includes selling price, taxes, title, registration, doc fee, add-ons, rebate assumptions, financing assumptions, and expiration terms.
Clean up VDP-to-quote consistency. If the website price and the first written quote do not match, fix the workflow before someone else scores it publicly.
Make add-ons explicit. If a product is optional, say so. If it is mandatory, make sure the advertised price and legal review support that claim.
Train the BDC and sales team for AI-assisted buyers. The shopper may be human, but the first negotiation may be handled by software. Dodging the question creates a record.
Track quote completeness as a KPI. Response time is not enough. A fast vague answer is still a bad answer. Measure whether the team gave the requested OTD breakdown.
Make the website agent-readable. AI shoppers need accurate inventory, crawlable pages, clear pricing language, schema, strong local signals, and technical performance. A broken website does not just hurt SEO. It hurts agent trust.

The Dealer AI Guy Take

This is the second internet moment for car dealers.

The first internet moment moved inventory and pricing online.

This one moves the shopper's labor online.

That is the part many dealers will underestimate.

AI agents do not only help consumers "research." They perform the annoying work consumers used to avoid: calling, emailing, asking again, saving the reply, comparing the math, and pushing back on the add-ons.

For consumers, that is leverage.

For good dealers, it is a chance to prove they are good before the customer ever walks in.

For sloppy dealers, it is a spotlight.

The event at AutoIndustry.ai Summit showed the industry something simple and uncomfortable: the next shopper may not browse your website like a person, fill out your form like a person, or negotiate like a person.

It will still represent a person.

And it will remember everything.

Sources And Context

Free Dealer AI Tools

If you want to check whether your store is ready for AI-assisted shoppers, start here: