You're buying cars the way we used to hire

My buddy Jake bought a Tacoma off Facebook Marketplace last month. Looked immaculate in the photos—dealer-quality shots, sunset lighting, the works. He flew to Phoenix, rented a car to drive out to some suburb, and found a truck that had clearly been in a front-end collision. The gaps in the body panels were asymmetric, the paint was three different shades if you looked close, and the frame had been "repaired" with what appeared to be spite and duct tape.

The seller acted shocked. "It looked fine in the photos, man."

Jake's out $800 in travel costs, a day of PTO, and is now more paranoid than my mom on NextDoor.

But here's the thing that's been eating at me since he told me this story: Jake did exactly what everyone told him to do. He checked the photos carefully. He asked questions. He even had the guy send a video walkthrough. He did his "due diligence."

And he still got cooked.

The convenience tax we stopped noticing

For most of human commerce, if you wanted to buy something significant, you went and looked at it. You touched it. You brought someone smarter than you to verify it wasn't garbage.

This wasn't romantic. It was expensive and annoying. It cost time, money, and favors from your mechanic uncle who you otherwise avoid at Thanksgiving.

Then the internet promised us a shortcut: high-resolution photos, detailed descriptions, seller ratings, platform protections. We could "verify" from our couch. The friction collapsed, the market exploded, and everyone convinced themselves this was just... how things worked now.

We treated the convenience as free. It was never free.

What we actually did was defer the verification cost. We swapped expensive upfront inspection for expensive downside risk. Most of the time, we got lucky. Sometimes—like Jake—we ate a four-figure lesson about information asymmetry.

And then the images became generative

I'm not going to pearl-clutch about AI ethics here. That ship sailed, caught fire, and sank somewhere in 2023.

What I am going to point out is that we just gave every seller on earth the ability to generate photos that look like they came from a professional shoot. For free. In seconds.

You know that uncanny valley feeling you get looking at AI-generated faces? Cars don't have that problem. A clean, well-lit Tacoma generated by Midjourney looks exactly like a clean, well-lit Tacoma from a real camera. There's no tell. You cannot distinguish them anymore.

Which means the entire "verify from photos" shortcut we'd been taking? It just became a highway to getting scammed at scale.

The question isn't whether sellers will use AI to make their listings look better. The question is why wouldn't they? There's no enforcement. No penalty. No way for buyers to know. And every incentive to do it because everyone else is.

This already happened once

You know what this reminds me of? Job postings.

For years, companies could post a vague job description, get 500 applications, and sort through them manually because labor was cheap and plentiful. Then LinkedIn, Indeed, and one-click applications made it trivially easy to apply to everything, so companies got flooded, so they built AI resume screeners, so applicants started using AI to spam-apply to more jobs, and now the whole system is just bots talking to bots while actual humans are locked out.

Nobody planned for this equilibrium. It just emerged because every individual actor was responding rationally to incentives.

Vehicle sales are heading the same direction. Sellers use AI to make listings perfect. Buyers can't trust photos anymore, so they... what? Reverse-image search everything? Hire forensic photo analysts? Or just accept that 1 in 10 purchases is going to be a disaster?

The re-localization of trust

Here's the uncomfortable thing I keep coming back to: we're going to end up right back where we started.

You cannot verify condition remotely when images are infinitely manipulable. That's not a technology problem. That's a trust problem. And you don't solve trust problems with more technology that can also be gamed.

You solve them with proximity. With witnesses. With skin in the game.

Which means we're headed back toward a world where, if you're serious about buying something expensive, you need someone physically present to verify it. Not because we want to go backward. Because the forward we built was built on an assumption—that photos represented reality—that no longer holds.

This is already starting. I've seen a couple of services pop up offering "boots-on-the-ground" inspection for remote car buyers. You pay someone local to the vehicle to go look at it, document everything, run diagnostics, and send you a real report from a real human who actually touched the thing.

That used to seem extravagant. Now it seems like the only move that isn't gambling.

What's actually happening here

I think we're watching the cost of verification re-materialize after a decade-long vacation.

For a while, the internet let us pretend we could verify remotely at zero cost. We treated photos as proof. Descriptions as binding. Platform protections as safety nets.

But those were always just conveniences enabled by the fact that most sellers weren't sophisticated enough to deceive at scale. The friction of deception was high enough that honesty was often the easier path.

AI dropped that friction to zero. Now deception is easier than honesty. And when that flips, trust collapses, and you're back to needing witnesses.

The wild part? This is efficient. Having a qualified human physically verify a vehicle before you commit five figures to buying it isn't backwards—it's the only rational response to a world where every image could be synthetic.

We're not regressing. We're re-pricing risk accurately.

Notes

1. The obvious counterargument: "But can't we just build AI to detect AI-generated images?" Sure. And then we'll build AI to defeat that detection. And then better detection. This is an arms race that buyers will always lose because sellers have more incentive to win. The only stable equilibrium is not playing the image-trust game at all.
2. Someone will point out that "most sellers are honest." Maybe! But when you can't tell the difference between honest sellers and sophisticated fraudsters, honest sellers don't benefit from their honesty. The market treats everyone as suspect, which punishes honest actors and rewards the creation of trust-verification infrastructure. Everyone ends up paying for inspections whether they wanted to or not.
3. I know this sounds dramatic for car buying, but the used vehicle market is $800B+ annually in the US alone. That's not niche. That's systemically important commerce that's about to have a trust crisis.
4. There's a version of this where platforms solve it by requiring verified inspections for every listing. Carvana sort of tried this model. But platforms are incentivized to maximize listings, not verify them. They won't police this until they're legally forced to. Which means individual buyers need their own verification layer.
5. The thing that keeps me up: this isn't specific to cars. This is any high-value physical good sold remotely. Real estate, equipment, collectibles—anything where condition matters and photos are your only window. We're headed for a re-localization of trust across entire market categories.

Fred A writes about markets, technology, and the uncomfortable places they intersect. This piece was prompted by too many friends getting burned on remote vehicle purchases and a growing suspicion that we're all living through a trust-infrastructure collapse we haven't named yet.