The RealReal Bets on AI to Fix Resale's Hardest Problem

There is $200 billion sitting in American closets right now, in the form of luxury goods that aren't being worn. Another $80 billion gets added every year. If you are in the business of moving that inventory, those numbers are either an opportunity or a rebuke, depending on how long you've been watching this market and how many times you've heard the pitch.

For The RealReal, they've been both. Fortune senior writer Phil Wahba went behind the scenes at the company's New Jersey authentication center to see what's changed — and the operational picture he found is worth sitting with, because it gets at something genuinely complicated about what AI can and can't solve in a business built on trust.

A Decade of Losses, Then a Turn

The secondhand clothing market reached $43 billion in the US in 2023, according to the Fortune video. The RealReal was one of the companies positioned to capture that — and still saw its shares crater toward record lows. ThredUp was in the same boat. The market was growing; the companies processing it were bleeding.

The core problem isn't demand. It's unit economics. The RealReal doesn't source from brands. It sources from individuals — which means every item that walks through the door is singular. Different colorways, different conditions, different provenance, different authentication risk. As one supply chain expert describes it in the video: "You have a significant supply side uncertainty, which makes this whole problem much, much more challenging."

That's the structural disadvantage the company has been trying to engineer its way out of. The RealReal started incorporating AI tools across its operations beginning in 2018, per the video — a timeline that tracks with the broader wave of enterprise AI adoption that accelerated through the late 2010s. The results, by 2025, were measurable: four consecutive quarters with positive adjusted EBITDA, the first time in the company's history. GMV rose 22% in the second quarter of 2026, and operating expense leverage improved by roughly 4.7 percentage points, per the video's reporting.

Those numbers matter because they suggest the operational thesis is working — AI reducing the cost per item processed, which is where the margin has always gone to die.

What Athena Actually Does

The system at the center of this is called Athena. The intake process, as Wahba shows it, begins with a merchandise operations specialist scanning tags — designer label, fabric label — and feeding that data into the platform. Athena takes it from there: automating the authentication process, assigning a counterfeiting score to each item, and routing it through the warehouse accordingly.

The company plans for Athena to process half of all new items by end of 2026, up from roughly 20% a year earlier.

The specific example that illustrates the technology's edge involves gemology. The RealReal has built proprietary tools for its specialists — high-magnification cameras that calculate carat weight in one to two seconds. A gemologist describes the shift plainly: "Our humans don't need to so much look at that. They're going to be focusing more on grading the actual diamond, which is where their expertise lies."

This is the stronger version of the AI-in-authentication argument: not that machines replace expert judgment, but that they absorb the measurable, repeatable tasks so that human expertise can be directed at the irreducible ones. Diamond weight is a calculation. Diamond quality is a judgment. Speed up the former; sharpen the latter.

The Counterfeit Problem Isn't Solved — It's Ongoing

Here's where the history matters. In 2018, Chanel sued The RealReal for selling counterfeit products. A class action lawsuit followed, accusing the company of using undertrained staff to hit authentication quotas. The company settled. These aren't ancient history — they're the context in which Athena was built.

When an executive says in the video, "We've invested millions of dollars over the years and continue to do so as fakes have gotten better over the years," the operative phrase isn't the investment — it's continue to do so. What's being described is a permanent cost center with no ceiling, structured like an arms race: counterfeits get more sophisticated, authentication has to keep pace, and the spending never stops. That's not a solved problem with a price tag attached. It's a recurring liability dressed up as a commitment to craft.

The company frames human expertise as a "moat" — the gemologist, the watchmaker who opens the case to check for aftermarket parts. The framing is defensible, but it also quietly concedes something: for all of Athena's capability, the terminal verification on high-value items still lives with a human being. Which means the liability lives there too.

The Verification Problem Isn't Unique to Resale

One expert in the video frames the broader AI risk clearly: "Fundamentally, these are non-deterministic systems, what is known as the verification problem. They do tend to hallucinate... if you make a wrong decision, it can potentially result in millions of dollars of losses. So when you integrate these things, it's super critical to figure out who is doing the verification, where is the human sitting in the loop?"

This is the question every company deploying AI in mission-critical decisions has to answer and most answer incompletely. "Human in the loop" can mean anything from "a specialist reviews every high-value authentication" to "a supervisor can override if something looks wrong." Those are not equivalent levels of oversight, and the distinction matters when the error in question is shipping a fake Chanel bag to a customer who paid for the real thing.

The RealReal's answer — specialists handling final judgment on complex items, AI handling routing and initial scoring — sounds structurally sound. Whether it actually holds at scale, across 50 million items with 200 attributes each, is something the company's claims can't confirm and the historical record gives reason to probe.

The Gen Z Story and Its Limits

The video closes on the demand side: resale as an identity statement for younger shoppers. An executive describes it as authenticity — "in a world where everything's starting to look the same and it's algorithms after algorithms. It's the anti-fast fashion at the end of the day."

It's also exactly what every platform company says before network effects fail to materialize the way the slides promised. Gen Z may well be structurally different from prior cohorts in their relationship to secondhand goods — there's genuine evidence that sustainability and individuality matter to them as purchasing signals. But "the next generation will drive permanent demand shift" is a claim that has flattered a lot of business models that didn't survive contact with those consumers' actual spending behavior.

The more durable demand story is probably simpler: used luxury at authenticated quality is a value proposition that doesn't require a generational thesis. A verified vintage Rolex at 60 cents on the dollar is attractive to a 45-year-old with money as much as a 23-year-old with taste. The addressable market is larger if you don't pin it to one cohort's preferences.

The Market They're Actually Competing For

The global resale market is projected to reach $393 billion by 2030, per the Fortune video. That number is large enough to support multiple winners and still lose money on per-unit processing if the operation isn't disciplined.

Shein, the comparison the video draws, can drop 10,000 new designs a day and ship in weeks at deep discounts. The RealReal cannot and will not compete with that on speed or price. What it's competing on is trust — specifically, the proposition that what you're buying is what it says it is. That's a narrow competitive lane, but it's real, and Athena is built to protect it.

The bet is that verified authenticity at reduced processing cost can produce sustainable margin at scale. The four quarters of positive adjusted EBITDA suggest the math is moving in the right direction. The outstanding question is whether Athena can keep up as the company scales intake volume and as counterfeit technology keeps pace with authentication technology — because so far, in this particular arms race, neither side has found a way to end it.

---

Jin Seo covers business and finance for BuzzRAG.