How to price and list second-hand fashion stock with AI

To price and list second-hand stock with AI, teams identify each item from photos, tags or product identifiers, let a model draft the listing from the brand's product data plus a condition note, and use comparable transaction data to suggest a price that is then adjusted as items sell or sit. AI handles drafting and suggestions well; humans should still confirm identification, condition and pricing boundaries, especially for higher value items.

Why is pricing and listing the bottleneck in resale?

Every second-hand item is a single unit with its own condition, so the listing work that happens once per style in new retail happens once per item in resale. Someone has to identify the exact product, photograph it, describe its condition, fill in attributes and choose a price. When that takes too long, processing costs exceed what low and mid priced items earn. AI is attractive because it can compress these steps from minutes to seconds, provided its output is reliable.

What can AI do in resale listing today?

Marketplaces have shipped AI listing tools for several years. eBay announced in September 2023 a 'magical listing' feature that generates titles, descriptions and other product details from a photo, reporting that 30 percent of US app sellers had tried the earlier version and that over 95 percent of them adopted AI generated descriptions. eBay's next generation version, rolled out in late 2025, generates titles, categories, item specifics and pricing recommendations from images. An independent review by Value Added Resource found, however, that the tool misidentified a test item in different ways across attempts, requiring sellers to delete wrong information and correct it manually.

Resale specialists use AI further along the chain. Rebag's AI tool CLAIR is used by Bloomingdale's associates in selected stores to evaluate customers' pre-owned luxury items and make offers, as reported by Retail TouchPoints in August 2024. In February 2026 the resale technology company Archive introduced AI capabilities covering warehouse management, merchandising and pricing analytics, and a pilot of agentic commerce tools for shoppers, according to FashionUnited.

Read also

AI in resale and second-hand fashion: authentication, pricing and listing

How does AI suggest resale prices?

A resale price model needs three inputs: what the item is (exact style, original price, season), what condition it is in (grade and defects), and what similar items have recently sold for and how quickly. Demand signals, such as searches, saves and views, refine the estimate. eBay describes its pricing suggestions as based on real time transaction data, intended to balance selling speed and price realisation.

For a brand's own resale programme, a useful structure is a base price as a percentage of original retail by grade, adjusted by a demand factor per style or category, with a floor and a ceiling set by merchandising. AI helps with the demand factor and with ongoing adjustments; the bands keep prices consistent with full price positioning. Trend signals from resale data are also useful upstream: styles that hold their value second-hand are candidates for continuation or reissue.

What does a practical AI listing workflow look like?

1. Capture. Photograph each item in a standard set of angles; scan any label code, QR or NFC tag.
2. Identify. Match the item to the brand's product master in the PIM, by identifier where possible and by image recognition otherwise; flag low confidence matches for review.
3. Grade. Let AI flag visible defects; the inspector confirms the grade and adds notes.
4. Draft. Generate title, attributes and description from PIM data plus condition notes; reuse original product imagery alongside the item photos.
5. Price. Apply the grade based price band and the AI demand adjustment; route exceptions to a merchandiser.
6. Publish and learn. List on own and partner channels; reprice on a schedule based on days listed and engagement; feed realised prices back into the model.

What are the risks of AI pricing and listing?

• Misidentification. Wrong model or material leads to wrong prices and returns; independent testing of marketplace tools shows it still happens.
• Invented details. Generative models may add features or materials not present. Restrict generation to verified attributes and condition notes.
• Condition misrepresentation. Under-described defects lead to returns and complaints; condition grading should remain a human decision for higher value items.
• Price erosion. Automated markdowns can push prices below levels that fit the brand; floors protect positioning.
• Thin data. For new programmes or rare items, there are few comparable sales; suggestions should show their confidence and data basis.

Read also

How can AI find the root cause of fashion returns?

How should teams measure the results?

Track listing time per item, share of listings edited by staff, identification error rate, price realised versus suggested, days to sell and returns of resale items by reason. ThredUp reported that direct listings it launched in late 2025 achieved average selling prices more than double those of its core marketplace, a reminder that what is listed and how it is presented can matter as much as the algorithm. The aim is lower processing cost per item and higher realised prices together; if one improves at the expense of the other, the rules, not just the model, need adjusting.