There's a guy in Massachusetts who built an AI agent to negotiate a car purchase for him. He gave it a fake email, a fake phone number, and instructions to play eight dealerships against each other over three days. It worked. He got a good deal on a Hyundai Palisade without ever talking to a salesperson.
The remarkable part isn't that the AI worked. It's that the fake identity was the whole point.
We've reached a moment where buying a car, one of the most normal things an adult does, requires a tradecraft operation to do safely. And if you're wondering how we got here, the answer isn't one villain. It's three, stacked on top of each other.
Layer one: The dealer
Car dealerships have been running the same psychological playbook since the 1950s. You Buy a Palisade every five years. The salesperson across from you closes twenty deals a month. They have access to auction data, financing markup tools, and decades of pressure tactics refined to exploit exactly the kind of person who just wants to drive home in something reliable.
The information asymmetry was always the point. Confusion is a feature. Exhaustion is a closing technique. Ever spend hours in a dealership trying to make a purchase? Then you know what I mean.
Layer two: The protectors
So naturally, an industry emerged to fix this. Services like CarEdge entered with a compelling pitch: let AI negotiate for you. We'll shield your information. We'll contact the dealers. We'll handle the back and forth. You stay anonymous until you're ready to close.
Here's what that actually looks like in practice.
You land on the site. You start filling out what you're looking for: make, model, price range. You're three steps in, invested, mentally already shopping. Then the wall goes up. Name. Last name. Email. Phone number. Hit continue to proceed.
And just below the friendly copy that says "We won't share your info without your permission. You're in control the whole way" (in the fine print) they're telling you that by entering your phone number you're consenting to automated telemarketing calls from CarEdge, dealers, and partners.
The service that promises to protect you from dealer spam is, in step one, harvesting your real contact information and pre-clearing itself to robo-call you.
This isn't ironic. It's structural. CarEdge's business model isn't protecting consumers. It's generating qualified leads. The AI negotiation tool is the funnel. You are the product moving through it.
Layer three: The escape hatch that isn't
So some people go a layer deeper. Open source agents like MoltBot. DIY setups. Fake names, burner emails, alias phone numbers, the whole operation, just to browse for a car without getting harvested.
This is where we are. This is the logical endpoint of the enshittification of trust.
Every layer exists because the previous layer failed the consumer. Dealers got predatory, so protective services emerged. Protective services got extractive, so DIY tools emerged. And now the baseline requirement for safely participating in a normal consumer transaction is a ghost identity and three days of agentic automation.
The real problem isn't the technology
AI agents are genuinely useful here. The Massachusetts guy proved it. An agent that doesn't get emotionally manipulated, doesn't get fatigued, and can simultaneously work eight dealerships at once is a real equalizer.
But an AI agent works for whoever deployed it. MoltBot worked for that guy because he deployed it. CarEdge's AI works for CarEdge. That distinction is everything, and almost nobody is saying it out loud.
When a service tells you "our AI is your advocate", ask who's paying for the infrastructure. Ask what the business model is. Ask what happens to your data if you never complete a purchase. The AI is only as aligned as the incentives behind it.
Trust has always been the commodity in these transactions. Dealers sold the illusion of a fair deal. CarEdge sells the illusion of protection. The enshittification of trust isn't a bug in the system.
It's the product.