How Resellers Use AI to List Second-Hand Items Faster

Resellers list second-hand items faster by batching photos, letting the marketplace's own AI (eBay's magical listing, Poshmark's Smart List AI, Depop's description generator) draft the title and item details, adding measurements and flaws themselves, pricing from sold listings, and crosslisting with a tool that delists automatically when something sells.

The step AI cannot take for you is the one that decides your feedback: condition. A photo model can tell a denim jacket from a chore coat, but it cannot feel a thinning elbow, smell smoke or measure a pit-to-pit. Every workflow that speeds up listing has to keep a person on those details, or the time saved on listing is lost to returns.

Follow me on Instagram@sagnikteaches

Where listing time goes on one-off items

Second-hand listing is slow because every item is unique; there is no product page to copy from. A quick way to see your own bottleneck is to time ten listings with a stopwatch. A typical pattern for a clothing reseller, before any AI, looks like the left column below.

Connect on LinkedInSagnik Bhattacharya

These timings are illustrative. The point is the shape: measurements and flaws do not get faster, and that is correct. Everything around them can.

Subscribe on YouTube@codingliquids

Step 1: Batch the photography

Shoot 20 to 40 items in one session with a fixed set-up: the same wall or backdrop, the same light, a hanger or mannequin at a marked spot, and a tape measure in shot for anything where size matters. Shoot in the same order every time: front, back, label, care tag, close-up of any flaw, measurement shots. Consistent order means an AI tool reading the photos sees the label in the third frame every time, and you can find the flaw photo without scrolling.

Name or number each item on a slip of paper in its first photo, and write the storage bin on the same slip. That number becomes your SKU, the stock code you will need when it sells.

Step 2: Let the marketplace's own AI draft the listing

All three big resale apps now draft from photos, free, inside the listing flow.

• eBay magical listing. Upload photos and it identifies the item, fills category and item specifics, suggests a condition and a price from recent transactions, and builds a title. eBay also has a bulk version in Seller Hub that turns batches of photos into listing drafts, in the markets where it has launched.
• Poshmark Smart List AI. Launched in early 2025, it generates the title, brand, size, colour and other details from a photo and guides you through the rest.
• Depop. Tap "Generate Description" after uploading a photo and it writes a description in Depop's casual style, with hashtags, and fills category, colour and brand.

Treat every draft as a suggestion. A reseller news site testing eBay's new version in February 2026 saw a mousepad identified as a ceramic decorative plaque, then as a fridge magnet, and the wrong identification carried through into the title and item specifics. With clothing, the common errors are subtler: a vintage label read as a modern reissue, a size taken from a sizing label that has been cut, or "wool" assumed from a photo of an acrylic knit. Check brand, size and material against the labels you photographed in Step 1.

A 30-second check on every AI draft catches most of this:

• Brand matches the label photo, including sub-lines (a diffusion line is not the main label).
• Size comes from the label or your measurements, never guessed from the photo.
• Material matches the care tag, word for word.
• Category is the one buyers browse, not the AI's nearest guess.
• Condition has not been set higher than your own grade.

A homeware and bric-a-brac seller has a slightly different job. Ceramics, glass and small collectables have fewer labels and more maker's marks, and a photo AI is far less reliable at reading a worn backstamp than a clothing tag. Photograph the base of every piece in sharp focus, let the AI suggest a maker, then confirm it against a reference guide or recent sold listings showing the same mark before the name goes in the title. Two extra fields matter here that clothing sellers rarely think about: packed weight and dimensions, because postage is often the difference between profit and loss on a $15 jug, and "chips, cracks, crazing and restoration", which buyers of old ceramics check first. Put both into your Step 3 shorthand every time.

The postage point is worth a quick sum. Say a stoneware jug sells for $15 and the listing charged $5.50 postage, based on the jug's own weight. Packed in a box with enough padding to survive the journey, it weighs almost twice as much and costs $7.80 to send. Take off the marketplace's fees and the $4 the jug cost at a car-boot sale, and the profit has mostly gone before the buyer opens the box. Weigh a few items packed and boxed, then write the packed weight into the shorthand rather than guessing from the item alone. An assistant can't see the box.

Step 3: Write the condition notes yourself, then let AI tidy them

This is where your expertise lives. A shorthand log from one batch (illustrative), written on a phone as each item comes off the rail:

#041 bin C  jumper, tag says lambswool, p2p 20 len 25, moth hole 3mm
            back hem, no smell, light pilling underarms
#042 bin C  cord trousers, W32 L30 on tag, measured W31 flat x2,
            rise 11, hem let down (crease line visible)
#043 bin D  jug, stoneware, mark on base unclear (photo 6), no chips,
            fine crazing inside, packed 1.4 kg

Two habits in that log save returns later. The trousers' measured waist sits next to the label size, because vintage labels often don't match modern sizing and buyers go by the tape. And the unclear maker's mark is logged as unclear rather than guessed, so the AI draft won't be allowed to fill it in. Jot flaws and measurements in shorthand like this as you handle each item, then use a chat assistant to turn shorthand into clean, honest text. A prompt and the kind of output it gives:

Turn my notes into a condition and measurements section for a
second-hand listing. Be precise and neutral. Never soften or omit a
flaw. Don't add anything I didn't write. Measurements in inches and cm.

Notes: levis type 3 jacket, label says 40, p2p 21, length 24,
sleeve 25 from shoulder. small bleach mark left cuff 1cm. top button
slightly loose. no rips. light fade on collar - normal wear.

Illustrative output:
Condition: Pre-owned, good vintage condition. A small bleach mark
(approx. 1 cm) on the left cuff, pictured. Top button is slightly loose
but secure. Light, even fading to the collar consistent with age.
No rips or holes.
Measurements (flat): Pit to pit 21 in / 53 cm. Length 24 in / 61 cm.
Sleeve (shoulder seam to cuff) 25 in / 64 cm. Label size 40.

Good, with two fixes. "Good vintage condition" is a grade you did not give, so replace it with your own grading scale. And "Levi's type 3" in your notes became nothing: the assistant dropped the model name because the prompt covered only condition. Keep identification in the title fields from Step 2 and use this output only for the condition block.

Your own grading scale is worth writing once and pasting into every prompt, so the assistant uses your words, not its own. A filled-in version for a clothing reseller:

• Excellent: no visible wear; could pass for unworn. No flaws to list.
• Very good: light signs of wear only (slight fading, softened fabric). No marks or damage.
• Good: worn, with minor flaws, each listed and pictured, none affecting wear.
• Fair: noticeable flaws (holes, stains, repairs) listed and pictured; priced for it.

With that scale in the prompt, the jacket above becomes "Good", because it has a bleach mark and a loose button, and the grade matches the photos a buyer is looking at. Buyers who have bought from you before learn what your "very good" means, which is its own kind of repeat business.

Depop's generated descriptions need the same scrutiny for a different reason: they're written in the app's casual voice and lean on era and style words. An illustrative draft for a patterned shirt read "vintage 90s oversized shirt, y2k vibes, perfect for festival season". The care tag had a modern layout and a website address printed on it, which usually points to a much later shirt; "90s" was a guess that a vintage buyer would spot at once. The edited version kept the casual tone, dropped the era, and added the chest measurement and the fabric from the tag. Era claims are one of the quickest ways to get a "not as described" return from buyers who know their labels.

Step 4: Price from sold listings, not asking prices

Asking prices tell you what people hope for; sold prices tell you what buyers paid. Filter the marketplace for sold or completed listings of the same item, copy 10 to 20 of them (title, condition, price, date) into a chat assistant, and ask for a summary:

Here are recent sold listings for the item I'm pricing. Group them by
condition, give the median sold price for each group, and flag any
outliers with the reason. My item: good condition, small flaw on cuff.

In one illustrative run on 14 sold jackets, the assistant reported a median of $68 for "good" condition, but it had counted a "new with tags" deadstock jacket at $140 in that group, lifting the median. After removing it the median fell to $61. Scan the grouped list yourself before you price, and set your price with room for offers if the platform encourages them.

Some items have no sold comps at all: a one-off studio pottery vase, a workwear jacket from a small maker, a job lot of old tools. Ask the assistant for search terms rather than a price: "What would a buyer type to look for this, and what similar items could I compare it with?" It might suggest the maker's type of mark, the glaze style or the period, which turns up comparable sold items you wouldn't have found by name. Then price at the upper end of those comparisons with offers switched on, and review after two weeks. Items with no comps sell slowly at any price; the aim is to learn from the offers you receive, not to guess right first time.

eBay's magical listing also suggests a price; compare it with your sold-comps median and trust the one whose evidence you can see.

Step 5: Crosslist, and delist the moment something sells

A one-off item listed on three platforms can sell twice. Crosslisting tools post one listing to several marketplaces and, crucially, detect a sale and remove the others. Two examples with published prices:

Vendoo's yearly billing brings Starter down to $12.49 a month. Before paying for a tool's AI extras, check whether the marketplace's free drafting from Step 2 already covers you.

Sale detection isn't instant, so plan for the rare double sale anyway. If a jacket sells on Depop and eBay within the same few minutes, cancel the second order quickly and tell the buyer straight. An assistant's first draft for that message, "Unfortunately due to unforeseen circumstances this item is no longer available", reads like an excuse. The version to keep: "I'm sorry: this jacket sold on another site a few minutes before your order and I only have the one. I've cancelled and refunded you in full, and if you'd like, I'll message you if I find another in your size." Buyers accept an honest one-off; the vague version is the one that earns a complaint.

For sellers who also run a shop of their own, syncing stock across Shopify, Amazon and eBay deals with the wider stock-sync problem.

Photo rules that catch out AI-assisted sellers

This is the part of AI listing most likely to get an account in trouble.

• Vinted. Its catalogue rules say photos must be taken by you for the purpose of selling on Vinted and must represent the item as it is, with no image editing allowed, and no stock photos, ad images or watermarks. AI background removal or virtual try-on sits awkwardly with "no image editing", so read the current rules before using either there.
• eBay. Its picture policy prohibits stock photos for used, damaged or defective items and bans added borders, text or marketing artwork. Photos must be at least 500 pixels on the longest side.
• Everywhere. Never let an enhancement hide a flaw. An AI "clean-up" that removes a stain from the photo is misrepresentation, whatever the platform's wording.

If you want better photos within those limits, creating product photos on a small budget covers lighting and set-up first. A realistic slip: a seller ran a batch through a background-removal tool that also smoothed fabric texture, and the pilling on two jumpers vanished from the photos. Both came back as "not as described". Keep one untouched close-up of every flaw in every listing.

A week of 60 items, before and after

For a part-time vintage clothing reseller who lists 60 items a week across eBay and Depop, the before-and-after looks roughly like this, taking the timings from the first table as a guide.

• Before: 60 items at 18 minutes is 18 hours a week, done in scattered evening sessions, with one or two double sales a month from delisting by hand.
• After: two photo sessions of 30 items (about 45 minutes each), marketplace AI drafts on eBay with Depop filled by the crosslister, condition notes from shorthand, sold-comps pricing for anything expected above $40. About 6.5 minutes per item, or 6.5 hours a week.
• Cost: the marketplace AI is free; a crosslister from $14.99 a month; a chat assistant for condition notes and pricing summaries at about $20 a month, or the free tier to begin with.

The saved hours are worth most if they go into sourcing, which is where resale profit is made. And the measure that tells you the workflow is safe is not listing speed but your return rate: if "not as described" returns rise after switching, the condition step is being rushed. For eBay titles and item specifics in more depth, see writing eBay listings with AI that sell faster.

Reselling with AI: common follow-ups

Does using AI descriptions hurt how listings rank in marketplace search?

There is no public evidence that the marketplaces penalise text simply because AI wrote it, and several of them now offer their own AI drafting. What does hurt is thin or inaccurate listings: missing measurements, wrong item specifics and vague condition notes lead to returns and poor feedback. Judge AI text on whether it is accurate and complete, not on who wrote it.

Can I use AI to put clothes on a model or clean up backgrounds?

Check each marketplace's photo rules first. Vinted's catalogue rules say photos must be taken by you and represent the item as it is, with no image editing allowed, and eBay bans stock photos for used items. If a platform permits enhanced images, keep an unedited photo of the actual item as the main one and never hide a flaw.

How many items a week does a crosslisting tool make sense at?

Once you list on two or more marketplaces and sell a few items a week, manual delisting becomes the risk: a one-off item sells twice and you cancel an order. At that point a crosslister with sale detection and auto-delisting, from about $15 a month, usually pays for itself in avoided cancellations alone. Below that, list on one platform and keep it simple.

Further reads

• How to Answer Etsy Customer Messages Faster With AI — Faster, polite replies to buyer questions and offers.
• How Small Amazon Sellers Use AI for Listings and Reviews — How AI listing help differs on a catalogue marketplace.
• How to Audit Your AI Subscriptions and Cut Wasted Spend — Stop paying twice for AI features your tools already include.
• Does Writing Etsy Listings With AI Hurt Your Search Ranking? — Whether AI-written listings affect search ranking, tested on Etsy.
• Best AI Tools for Etsy Sellers, Ranked by Time Saved — If you sell vintage on Etsy too, the tools ranked by time saved.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: eBay Innovation announcements on magical listing and eBay's picture policy; Poshmark announcement of Smart List AI; Depop newsroom on AI-powered listing from one photo; Vendoo and Crosslist pricing pages; Vinted catalogue rules; Value Added Resource testing of eBay's magical listing (Feb 2026).