Price from eBay's sold data first, then give an AI assistant your photos plus the facts it can't see (model number, measurements, faults, what's included) and ask for an 80-character title that puts buyers' search words first, complete item specifics and a flaw-honest description. Check every claim before listing, because eBay holds you responsible for accuracy.
It helps to know what "faster" depends on. eBay's help pages say its Best Match ranking considers how closely a listing matches the search, the item's popularity and price, the quality and completeness of the listing, terms such as returns and handling time, and the seller's track record. AI helps with two of those: relevance and completeness. Price does more for speed than any wording, which is why the research comes first.
Before AI writes a word: price from sold listings
In Seller Hub, the Research tab has eBay's product research tool, free to all sellers, with up to three years of sales data: average sold price, sold-price range, shipping costs and sell-through. For a single item, searching sold listings works too. Either way, look at sold prices, not asking prices, and only at items in the same condition.
AI is useful for turning a messy list of comparable sales into a price decision. Paste in what you found and ask for a range:
Here are 14 sold listings from the last 90 days for a used Canon EF
50mm f/1.8 STM lens: [price, condition, what was included, days listed].
1. Give the sold-price range for used, tested examples with caps only.
2. Show whether boxed examples or ones with a hood sold for more.
3. Suggest a Buy It Now price to sell within about two weeks, and a
price to sell within a few days. Show your reasoning.An illustrative answer to that prompt:
Used, tested, caps only (9 sales): $64 to $88, median about $74.
Boxed examples (3 sales) averaged about $9 more; hood made no clear
difference (2 sales).
Buy It Now to sell within two weeks: $76.
To sell within a few days: $69.Check the arithmetic yourself; assistants misread pasted tables more often than you'd expect. In this one, the $64 low was a listing described as "AF noisy, works", which belongs in a lower bracket, not the tested range; take it out and the range starts nearer $70, so the quick-sale price moves up too. The fix is to ask for the listings it counted in each group, then glance down the list. The reasoning is usually sound once the inputs are right, and it stops you listing at a hopeful price and waiting a month.
Feeding AI the facts it can't see
An assistant looking at photos will guess at model numbers and never notices a fault that isn't obvious. Give it a short fact sheet for every item:
- Exact model or part number, copied from the label, not read from a blurry photo.
- Measurements where buyers need them: size, length, weight.
- Condition, fault by fault, with where each one is ("scuff on outer cage plate, photo 4").
- How you tested it and what happened.
- What's included, and what isn't.
- Compatibility you have actually confirmed, and nothing more.
The last point is where most AI-drafted returns start. Assistants are keen to be helpful about compatibility, and compatibility claims are what "not as described" disputes are made of.
Items with no label need a different sheet. An unbranded saddle, a vintage hand tool with the maker's stamp worn off, or a generic phone mount will tempt the assistant to name a brand it thinks the item resembles. Don't let it. Put "Brand: unbranded" (or "unknown, stamp illegible, photo 3") in the sheet, and give the measurements that let a buyer decide without a brand: rail width and length for the saddle, overall length and jaw opening for the tool. Buyers searching for a specific brand won't find it, but the ones who do find it won't return it for being the wrong one.
The 80-character title: a formula and a prompt
eBay allows 80 characters and recommends including the brand, model or type, colour, style and model number. A formula that works for most used items: brand + model or part number + item type + one or two key specs + a condition or completeness term. No filler ("L@@K", "wow"), no question marks, and only the one brand the item actually is.
Write three eBay titles for this item, each 80 characters or fewer
including spaces. Put the words buyers search first: brand, model,
item type. Use only facts from the sheet. No question marks, no
words like "rare" or "look", no other brand names.
Give the character count for each.
Fact sheet: Canon EF 50mm f/1.8 STM lens. Used, fully tested on a
DSLR, autofocus and aperture work. Front and rear caps included.
No box, no hood. Light marks on the barrel. Clean glass.An illustrative response:
1. Canon EF 50mm f/1.8 STM Lens Autofocus Prime, Caps + Hood, Tested, Used (71)
2. Canon EF 50mm f1.8 STM Prime Lens, Full Frame, Front & Rear Caps, Fully Tested (78)
3. Canon EF 50mm f/1.8 STM Lens for EOS R Mirrorless & DSLR, Caps, Hood, Tested Used (81)What to fix. Options 1 and 3 add a hood that the fact sheet says isn't included; that alone would generate a return. Option 3 is 81 characters, and "for EOS R" misleads, because EF lenses need an adapter on EOS R bodies. Option 2 is the one to use, perhaps with "f/1.8" written the usual way. The assistant's character counts were right this time, but check them: eBay will stop you at 80, and counting errors are common.
Item specifics: fill every one that applies
eBay recommends adding item specifics such as make, model, size, colour and style, and prefills some for you when it recognises the product. Specifics feed the filters buyers use, so a listing with half of them missing drops out of filtered searches entirely. The field names depend on the category. For the lens, a filled-in set looks like this (illustrative):
Brand: Canon Model: EF 50mm f/1.8 STM
Focal length: 50 mm Maximum aperture: f/1.8
Mount: Canon EF Focus type: Auto and manual
Type: Standard / prime Condition: Used
Included: front cap, rear cap (no box, no hood)AI can map your fact sheet onto a category's specifics quickly if you paste the field list. Check every value it fills in. An assistant that doesn't know a value tends to supply a plausible one rather than leave it blank.
Descriptions that answer the buyer's three worries
A used-item buyer wants to know three things: does it work, what's wrong with it, and what exactly will arrive. A before and after for a part from a bike shop's workshop:
Before: Great used derailleur, works perfectly, see pics.
After: Shimano 105 RD-R7000 rear derailleur, 11-speed, short cage.
Removed during a customer's upgrade in our workshop and tested on
our stand: shifted cleanly across all 11 gears.
Condition: scuffs on the outer cage plate (photo 4); jockey wheels
about half worn (photo 6).
Included: derailleur and mounting bolt. Not included: cable, hanger.
Compatibility: please check the manufacturer's compatibility
information for your drivetrain before buying.
Posted within one working day. 30-day returns.The second version heads off the three questions that would otherwise arrive as messages, and it points compatibility checking back to the manufacturer rather than making claims. A realistic mistake from the other direction: an assistant given the note "light dust inside the lens" wrote "pristine optics, no dust or haze" in a description, because the rest of the sheet sounded positive. The buyer opened a "not as described" return. Paste the fact sheet into every prompt and add "don't soften or omit any fault".
Faulty items are where the softening instinct does most damage. Take a studio flash head that fires at every power setting but whose modelling lamp is dead. The assistant's first draft called it "a minor issue with the modelling lamp that doesn't affect flash performance". That's an opinion, not a fact, and a buyer who needs the modelling lamp for positioning will disagree. The rewrite: "Flash fires correctly at all power settings (tested). Modelling lamp does not light: we haven't tested whether the bulb or the circuit is at fault. Priced to reflect this." An honest version like this still finds buyers, such as someone who only needs a background light, and it doesn't come back. List it under the condition eBay offers for items that don't fully work, rather than as plain "used".
eBay's own AI tools versus ChatGPT or Claude
eBay has AI built in. Its app can build a draft listing from photos, its AI bulk listing tool in Seller Hub (where available) scans batches of images and suggests categories, titles and item specifics as drafts, and the listing flow includes an AI description writer. eBay's own announcement "strongly recommends" checking all details because you're responsible for the listing's accuracy.
| eBay's built-in AI | ChatGPT or Claude | |
|---|---|---|
| Cost | Included | Free tiers, or about $20 a month for Plus or Pro |
| Knows eBay's categories and specifics | Yes, and prefills them | Only if you paste the field list |
| Works from photos | Yes, including in bulk | Yes, one chat at a time |
| Follows your own rules and voice | Limited | Fully, with a saved prompt |
| Uses your fact sheet | You edit the draft afterwards | Built in from the start |
| Best for | High volume of common, recognisable items | Items with faults, parts, anything needing care over claims |
Many sellers use both: eBay's tool to create the draft and prefill specifics, then a general assistant to rewrite the title and description from the fact sheet. For a bigger second-hand operation, how resellers list second-hand items faster with AI covers batching and photo workflows.
What gets listings demoted or returned
- Graphics on photos. eBay asks you not to add badges, logos, copyright notices or watermarks, which it says can affect search placement. You can add up to 24 photos, at least 500 by 500 pixels, with about 1,600 by 1,600 recommended. If you're improving photos cheaply, creating product photos with AI on a small budget helps, but for used items the photos must show the actual item.
- Stock photos for used items. Not allowed when you list an item as used.
- Keyword spamming. eBay's search manipulation policy bans words not directly related to the item, question marks in keywords, and several brands or models in one listing. Tell the assistant this in the prompt, because keyword-heavy titles are exactly what it produces when asked to "optimise for search".
- Loose compatibility wording. "Fits", "for" and "compatible with" have rules of their own under the same policy. Use them only for compatibility you've confirmed.
- Invented specs and softened faults. The largest single cause of AI-related returns. Online shop AI mistakes that hurt trust covers the wider pattern.
A photography studio clearing old kit
Camera gear is a good case for careful AI listing, because buyers are knowledgeable and the details that matter are easy to get wrong. Say a portrait studio upgrades and has two older camera bodies, four lenses, a set of flash heads and some stands to sell.
The fact sheet for a body should include the shutter count (from the camera's own data, not a guess), any sensor marks, battery health and what's in the box. For lenses, glass condition in plain words (dust, haze, fungus, scratches), whether autofocus and stabilisation were tested, and which caps come with it. Flash heads need their power and whether every head and cable works. A filled-in sheet for one of the bodies (illustrative):
Item: Canon EOS 6D Mark II body, used
Shutter: 41,200 actuations (read from camera data, screenshot photo 9)
Sensor: clean, no marks visible at f/16 test shot (photo 10)
Cosmetic: rubber grip lifting slightly at thumb rest (photo 5);
light wear on base plate
Battery: one original battery, holds charge for a full shoot
Tested: stills, video, autofocus; single SD slot reads and writes
Included: body cap, battery, charger, strap. No box, no manual.The first version of this sheet said "both card slots work", written from memory of a different camera. The 6D Mark II has one slot, and the error surfaced only because the sheet was checked against the body before anything was drafted; an assistant would have repeated it in the description without hesitation. From this sheet, a title such as "Canon EOS 6D Mark II Full Frame DSLR Body, 41k Shutter Count, Tested, Charger" comes in under 80 characters with the shutter count, which is the number camera buyers look for first.
Two decisions are business decisions, not AI ones. Bundling a body with a lens often sells more slowly than listing them separately, because buyers want one or the other; the sold data will tell you which. And lighting sold as a kit needs one listing with a clear list of every item, since "set of three" titles attract questions if one stand is missing. The assistant handles the wording once those choices are made.
Clearing 30 items from a bike shop's back room
Say a bicycle repair shop has 30 used parts and accessories from customer upgrades: derailleurs, wheels, saddles, a few tools. Listing them by hand took the owner about 15 minutes each, which is why they sat in a box for a year.
The routine. Photograph each item on the workbench against a plain background in daylight, six to eight photos each. Write the fact sheet on a phone note as each part is checked: label, faults, test result, what's included. Research sold prices for the ten most valuable items; price the rest from a quick sold search. Run the title and description prompt in batches of five, paste each result into the listing, check every claim against the sheet, and fill specifics.
The numbers. About 6 minutes per item instead of 15: three hours for the lot instead of seven and a half. Priced from sold data at the "sell within two weeks" level, 22 of the 30 sell within 30 days. One comes back, and it's the one listing where the fact sheet was skipped. The eight left are mostly items with few sold comparisons, which get a price cut after three weeks.
Reading the ones that didn't sell. Before cutting prices across the board, look at each unsold listing's views and watchers in Seller Hub. Plenty of views with no watchers usually means people found it and didn't like the price or the photos. Watchers who never buy suggests the price is close, and a Best Offer option can close the gap. Very few views points back at the title and item specifics: in the shop's case, two wheels had been listed without the wheel size filled in, so they were missing from every filtered search by size. Adding the specific cost nothing and fixed both.
Buyer questions. Messages still arrive, and AI drafts the replies well if the fact sheet is pasted in. A buyer asked whether a rear wheel "will fit my 2019 road bike". The first draft answered "Yes, this wheel is compatible with most modern road bikes", which is a promise the shop can't back. The reply that went out: "It's a 700c rim-brake wheel with a quick-release axle and an 11-speed freehub. If your bike uses rim brakes and a quick-release rear wheel, it should suit, but please check your frame's spacing and brake type before buying, and message us with any measurements you'd like confirmed." Buyers get the facts they need to decide, and the decision stays theirs.
The shop now lists new back-room stock every Friday afternoon. If it later sells on its own website or other marketplaces as well, syncing stock across Shopify, Amazon and eBay stops the same part selling twice. And the prompt habits carry over: using AI for Etsy titles, tags and descriptions shows the equivalent routine for a handmade marketplace.
Further reads
- How to Write Product Descriptions With AI That Actually Sell — Description principles that carry across every marketplace.
- Can ChatGPT Read PDFs, Spreadsheets and Photos? What Breaks — What assistants get wrong when reading labels in photos.
- How Small Amazon Sellers Use AI for Listings and Reviews — The same job on Amazon, with its own 2026 title rules.
- Does Writing Etsy Listings With AI Hurt Your Search Ranking? — Whether AI-written listings hurt marketplace search ranking.
- Taking a Local Shop Online With AI: Product Listings in a Weekend — Getting a whole shop's stock listed in a weekend.
- How to Track Competitors' Prices and Offers With AI — Keeping an eye on competing sellers' prices over time.
- Etsy's AI Title Suggestions: Should You Accept Them? — Etsy's title tool reads your title, first photo and description but not your sales. A four-question test for each suggestion, with worked verdicts.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: eBay seller guidance on creating listings, eBay help pages on Best Match, adding pictures, product research and the search manipulation policy; eBay Community announcement of the AI bulk listing tool, checked September 2026.