How does The RealReal use AI for authentication, pricing and intake?
The RealReal pairs human authenticators with image tools, prices most items algorithmically and is moving intake onto an AI system called Athena. What it has disclosed, and what the model means for resale.
KEY TAKEAWAYS Summary by the editors
- The RealReal launched Athena, an AI-enabled product intake process, in Q1 2025; it processed more than 10% of items in its early phase and cut processing times by an estimated 20%.
- By the end of 2025 Athena handled 35% of items and was on pace to process close to half by the end of 2026, according to PYMNTS's report of the Q2 2026 earnings call.
- The RealReal said most units were launched with algorithmic pricing by the end of 2024, using near real-time signals such as page views, 'obsessions', primary market prices and search trends.
- The RealReal states that more than 100 in-house experts physically examine every item, supported by TRR Vision (microscopic image comparison) and TRR Shield (counterfeit risk flagging).
- In Q2 2026 The RealReal reported GMV of $617 million, up 22% year on year, and described its AI changes as 'compounding rather than incremental'.
The RealReal uses AI to process, price and protect a luxury resale inventory in which every item is unique. An AI-enabled intake system called Athena automates authentication and cataloguing steps, pricing algorithms set and adjust prices from demand signals, and image-based tools support, but do not replace, human authenticators. The company links these changes to lower processing costs and faster growth in 2025 and 2026.
Why does luxury resale need AI?
A primary retailer catalogues a style once and sells many units. A consignment business like The RealReal must receive, authenticate, describe, photograph and price each item individually, and the price of a used handbag depends on brand, model, condition and current demand. The RealReal says more than one million items pass through its authentication centres each month. At that scale, minutes saved per item and better first prices have a direct effect on margin and sell-through.
What is Athena and what has it changed?
In its Q1 2025 prepared remarks, The RealReal said it had launched Athena, an AI-enabled product intake process, during the quarter. More than 10% of items went through Athena in this early phase, and processing times fell by an estimated 20%. Athena initially covered ready-to-wear, with plans to expand to shoes and handbags later in 2025. The company also said that in 2024 increased automation had cut processing time by more than 10% while headcount stayed steady.
By the Q2 2026 earnings call, as reported by PYMNTS, Athena was automating authentication and cataloguing steps that previously required manual attribution. It handled 35% of items at the end of 2025 and was on pace to process close to half by the end of 2026. Chief executive Rati Levesque described its extension into higher-value items as 'the next phase of the system's rollout', and the chief financial officer said the technology removes multiple dollars of processing cost per unit.
| Area | Tool or method | Published detail |
|---|---|---|
| Intake and cataloguing | Athena | Launched Q1 2025; 35% of items by end 2025; aiming for close to half by end 2026 |
| Pricing | Algorithmic pricing | Most units priced algorithmically by end 2024; over 100 data points per item (2026) |
| Markdowns | Algorithmic discounting | In early testing in Q1 2025 |
| Authentication support | TRR Vision, TRR Shield | Image comparison and counterfeit risk flagging alongside human experts |
| Listings | Listing enrichment | Occasion, collection and trend data added automatically (2026) |
| Shopping | Conversational shopping agent | Being tested with Google (2026) |
How does The RealReal price resale items with algorithms?
The RealReal said that by the end of 2024 most units were launched with algorithmic pricing. The algorithms are dynamic and use near real-time signals such as page views, 'obsession' counts (the site's favourites), primary market pricing and search trends. In early 2025 it was testing algorithmic discounting, with initial results suggesting a better balance between price and sell-through.
PYMNTS reported in August 2026 that the pricing algorithm evaluates more than 100 data points per item to set an initial price and manage markdowns, and described it as setting and adjusting prices 'without human override'. That is a strong statement of automation; the company has not published error rates or how it handles items with little comparable sales history.
How does The RealReal authenticate items, and where does AI fit?
Authentication remains led by people. The RealReal's authentication page says more than 100 in-house luxury experts, each with hundreds of hours of brand or category training, physically examine every item. Two proprietary tools support them:
- TRR Vision captures microscopic images of an item and compares them with millions of images in its database to catch small inconsistencies.
- TRR Shield tracks which brands and styles are most likely to be counterfeit and flags those items on arrival for extra attention.
- Materials analysis, including XRF machines that identify metal alloys in jewellery, watches and hardware for comparison with genuine items.
The company says it has 13 years of data on more than 37 million authentic items and more than 250,000 inauthentic items. That labelled history is the core asset: image models and risk scoring need large volumes of verified examples, including fakes, to be useful.
What results has The RealReal reported?
For Q2 2026, The RealReal reported GMV of $617 million, up 22% year on year, and revenue of $193 million, up 17%, according to PYMNTS. Levesque described the shift as 'compounding rather than incremental'. In Q1 2025 the company credited AI-led efforts, including Athena, as drivers of operating expense leverage. It is not possible to separate the AI contribution from other factors such as growth in high-value consignment and new stores.
Other disclosed metrics show the business mix the AI supports. In Q1 2025 The RealReal reported GMV of $490 million, up 9%, and an average order value of $564. By Q2 2026 average order value had risen to $659, up 13%, and PYMNTS reported that sales of items above $1,000 rose 36% in the first half. Higher-value items make each authentication and pricing decision more consequential, which is why the extension of Athena into such items is a significant step. The company also said AI now adds occasion, collection and trend data to every listing, which supports search and merchandising.
What are the risks and lessons for resale and brands?
Automated pricing on one-off goods can misprice rare items, and fully automated markdowns can erode value if demand signals are thin or manipulated. Authentication tools carry the opposite risk: a false negative damages trust far more than a slow listing costs money. The RealReal's response is to automate intake and pricing heavily while keeping physical inspection.
- Build and keep labelled data, including counterfeits and returns, from day one.
- Automate the repetitive steps (attribution, measurements, descriptions) before the judgement calls.
- Use demand signals for pricing, but monitor items with little comparable history separately.
- Keep human authority where errors damage trust, and say so publicly.
Frequently asked questions
Does The RealReal use AI to authenticate items?
The RealReal says more than 100 in-house experts physically examine every item. AI-supported tools such as TRR Vision (microscopic image comparison) and TRR Shield (counterfeit risk flagging) assist them, and its Athena intake system automates some authentication and cataloguing steps.
What is Athena at The RealReal?
Athena is an AI-enabled product intake system launched in Q1 2025. It automates cataloguing and attribution steps; it handled 35% of items by the end of 2025 and was on pace to process close to half by the end of 2026.
How does The RealReal set prices?
Most items are priced by algorithms that use near real-time signals such as page views, favourites, primary market prices and search trends. In 2026 PYMNTS reported that the algorithm evaluates more than 100 data points per item and also manages markdowns.
Has AI improved The RealReal's results?
The RealReal links AI to lower processing costs and operating leverage, and reported 22% GMV growth in Q2 2026. The company has not isolated how much of that growth comes from AI rather than other initiatives.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
