AI Photos for Reselling: what the rules actually say 🔍

The question nobody answers with sources

"Can I use AI photos on Poshmark?" is one of the most searched questions in US reselling right now, and the answers you find are almost uniformly confident and almost uniformly unsourced. Some say it is fine. Some say you will get banned. Very few quote a single line of an actual policy.

So we did the boring version. We read the published photo rules on Poshmark, Depop, Mercari, eBay and Vinted US, on their own pages, on 31 July 2026. Below, each one is quoted word for word with a link, so you can check us.

Two things came out of that reading, and both are more interesting than the yes/no everyone is arguing about.

First: the platforms are much stricter and much more specific about photos than most sellers realise — but what they are strict about is authorship and accuracy, not technology.

Second: not one of the five has published a rule that mentions AI-generated images. Not a permission, not a prohibition. Meanwhile all of them have been shipping AI features of their own. That silence is the actual subject of this article, and pretending otherwise would be the easy, dishonest way to write it.

Three very different things get called "AI photo"

Most of the confusion in this debate comes from one word covering three practices that have almost nothing in common ethically. Separating them makes the rest of the question much easier.

1. Enhancement. You photographed your actual jacket. AI corrects the white balance, evens the lighting, removes the shadow your body cast across the sleeve. The item in the image is the item in the box, and nothing about it has changed.

2. Presentation. You photographed your actual jacket, and AI shows it on a neutral background, or worn on a generated model so buyers can see how it drapes. The garment is still yours and still accurate; what is synthetic is the context around it — the backdrop, the body wearing it.

3. Fabrication. An image of a jacket is generated that was never photographed in your possession. Or an image of your jacket is altered so the colour is richer, the fabric less worn, the stain gone.

Category 3 is not a grey area on any platform in this guide, and it is not a grey area ethically either. It is misrepresentation, and every published rule quoted below rules it out — under the heading of accuracy, of stock images, or of not hiding condition.

Categories 1 and 2 are where the actual question lives. Hold on to that distinction as we go through the rules, because the platforms' language turns out to speak to it fairly directly, even though none of them was written with AI in mind.

Vinted US: the most explicit rules of the five

Vinted launched in the United States in January 2026, and it brought its global catalog rules with it. They are the most quotable photo rules in this entire comparison — and the strictest. If the platform itself is new to you, our guide to how Vinted USA works and what sellers pay covers the mechanics.

Straight from Vinted's catalog rules:

"They have to be taken by you with the purpose of using the photos on Vinted."

"They cannot be replaced with stock photos, photos from ads and watermarked images."

"They have to represent the item as it is."

"No image editing is allowed. We also recommend showing any defects, scratches, wear, etc."

"The first photo should clearly represent the whole item."

"You can upload up to 20 photos per listing."

That phrase is the strictest sentence any of these five platforms publishes about images, and it is worth sitting with rather than skating past.

Read at its most literal, it would rule out cropping and brightness correction, which every seller on the platform does and which no moderation system appears to act against. Read in context — sandwiched between "represent the item as it is" and a recommendation to show defects — it reads as aimed at editing that changes the perception of the item: the colour that is not the real colour, the stain that has been retouched away.

We cannot tell you which reading Vinted applies, because Vinted has not published that, and only Vinted interprets Vinted's rules. What we can say is that the sentence sits inside a paragraph about faithfulness, and that the documented enforcement discussed publicly concerns misleading listings rather than exposure correction.

Our deep dive on this specific question — Vinted moderation and AI photos — goes through the terms, the legal analyses and the EU AI Act in detail. It was written for European sellers and the rules it quotes are the same ones that now apply in the US.

Mercari: "photos that you do not own or have not taken yourself"

Mercari turns out to have the clearest published position on photo authorship of any US-native platform here, and it is in a place most people never look: not the photo tips, but the prohibited conduct policy.

Mercari's Prohibited Conduct policy lists among prohibited listing behaviours:

"Using any stock or advertisement photos"

"Using photos that you do not own or have not taken yourself"

"Listing items not in your possession, shipping from a third party or stating the item will be shipped at a future date"

"Setting a false brand"

That second line is the one that matters for our question, and it is worth reading carefully because it contains two separate tests joined by "or": photos you do not own, or photos you have not taken yourself.

A photo you took of your own garment, then processed, remains a photo you own and took. A generated image of a garment you never photographed satisfies neither test. The policy was plainly written about lifted retailer images rather than about generative AI — but the language it happens to use maps onto the distinction from the previous section more cleanly than most.

Mercari's Marketplace Guidelines add the general standard in plainer terms: sellers should "accurately represent yourself and your stuff", and avoid "false information".

Depop: "only use photos taken by yourself"

Depop's rule lives in a help centre article with an unassuming title — What are 'stock images'?, last updated 13 May 2025 — and it is the most character-rich statement of the five:

"Photographs are the DNA of Depop. This is why we ask that all photos you use on your Depop shop, are original and not lifted from an online source."

"Stock images and other photos from the internet actually belong to the original photographer, so using them to advertise your sales is an infringement of property rights, which is why we don't allow it."

"Only use photos taken by yourself — don't be afraid to get creative and show off your personality."

Two things stand out.

The stated reason is intellectual property, not deception. Depop's objection to stock images is that they belong to someone else. That is a different argument from Vinted's faithfulness argument, and it lands differently on an AI render generated from your own photo — which is not lifted from anyone.

And the last clause is an active invitation to stylise: "don't be afraid to get creative and show off your personality". Depop's culture rewards a strong visual identity; the platform says so itself. That is the most permissive-sounding language in this comparison — though "only use photos taken by yourself" in the same paragraph is a real constraint, and Depop has not published anything about what happens when a photo you took becomes the input to a generated image.

Depop is also the platform where photo economy bites hardest: it allows notably fewer images per listing than Vinted does, so each one has to carry more information. Check the current limit in the app, since it moves with product updates. Practically, on a four-image listing there is little room for a visual that is not doing informational work — the flaw shot is not optional, and the aesthetic shot has to earn its slot.

eBay: accuracy over aesthetics, in unusual detail

eBay's picture policy is the most operationally specific of the five. Not allowed:

"Photos that don't accurately represent the item"

"Placeholder images used to convey messages"

"Stock photos for used, damaged, or defective items"

"Photos with added borders"

"Photos with added text, artwork or marketing material"

"Watermarks of any type, including those used for ownership attributions"

And the technical floor: "Photos must be at least 500 pixels on the longest side."

Three of these are worth flagging for anyone thinking about AI presentation.

"Photos that don't accurately represent the item" is the broadest accuracy standard in this comparison, and it is written as a property of the resulting image rather than of the method. That cuts both ways: it does not care that you used AI, and it does not excuse you because you used AI. The output is judged.

"Stock photos for used, damaged, or defective items" is precisely targeted. eBay tolerates catalogue imagery for genuinely new-in-box goods where a stock image is accurate. The moment an item is used — which is the entire resale market — that tolerance ends.

"Photos with added text, artwork or marketing material" and the watermark ban rule out a habit some sellers pick up from other platforms: badges, price flashes, "FAST SHIPPING" banners. Whatever your AI tool offers in that direction, eBay has published that it does not want it.

eBay is also the platform where the authenticity stakes are highest, since pre-owned designer goods are a major category. In a segment where buyers and authentication services scrutinise stitching and hardware in your photos, an image that has been smoothed or reconstructed is working against your own interest, independently of any rule.

Poshmark: the honest gap

Here is where we have to report an absence rather than a rule, because that is what we found.

We read Poshmark's Community Guidelines looking for a photo policy comparable to the four above — an authorship requirement, a stock-image prohibition, an accuracy standard for images. We did not find one published there.

What the Community Guidelines do provide is a general honesty standard, under the heading "Be Real":

"You may not use Poshmark to collect personal data or content that doesn't belong to you or to which you do not own the rights."

"Please don't engage in any behavior to impersonate or mislead others."

That is a real standard, and "don't mislead others" plainly covers a listing image that misrepresents an item. It is simply less specific than Mercari's "photos that you do not own or have not taken yourself" or Vinted's "represent the item as it is".

Poshmark is worth understanding as a platform here, because its culture is unusually social — closets, sharing, Posh Parties, relationships between regular buyers and sellers. In a market where repeat buyers matter that much, a visual that overpromises does more damage than a removed listing would. The person who is disappointed by their delivery is someone who was going to shop your closet again.

The AI question none of them has answered

Now the finding that the whole article has been building toward.

Across five platforms and every published photo rule we could locate, not one mentions artificial intelligence, generated images, or synthetic media in the context of listing photos. Not to permit them. Not to prohibit them.

That is a genuinely strange situation, and it becomes stranger when you notice what the same companies have been doing with AI. Poshmark, Depop and Mercari have all rolled out AI-assisted listing tools that generate descriptions and item attributes from an uploaded photo. These are not reluctant adopters; they are shipping the technology into their own products.

But those features work on the description side, from a photo you supplied. None of them resolves the question of AI-generated images — and none of the platforms has published a position on it.

Three honest conclusions:

1. There is no published prohibition to point to. Anyone who tells you AI photos are "against the rules" on these platforms should be asked which rule. As of 31 July 2026, none of the five has published one that says so.

2. There is no published permission either. Which means the accuracy and authorship rules that are published are the ones you will be judged against. Those rules are quoted above, and they are not permissive about misrepresentation.

3. This will change. A category this visible does not stay unaddressed indefinitely. When rules do arrive, sellers whose practice was already built on faithfulness to the real item will need to change little. Sellers whose listings depend on a flattering fiction will find out quickly.

The FTC layer, kept in proportion

US sellers sometimes hear that the Federal Trade Commission has rules about AI disclosure, and get alarmed. Worth putting this in its right size.

The FTC's core standard is genuinely broad. In the Commission's own words, when consumers see an advertisement, "federal law says that ad must be truthful, not misleading, and, when appropriate, backed by scientific evidence" — and the FTC "applies the same standards no matter where an ad appears".

That is a principle about not deceiving people, and it points in exactly the same direction as every platform rule above. If your listing image gives an accurate impression of the item, you are aligned with it.

One genuine cross-border note: if you sell into the European Union, transparency obligations for AI-generated content under the EU AI Act apply from 2 August 2026. That is a separate framework from anything US, and we cover it in the Vinted moderation article.

The honest-AI method

Everything above reduces to a practice. Here it is, and it is short.

The principle

AI is for presenting the real item, never for inventing one. Every published rule in this article is compatible with that sentence, and no future rule will contradict it.

The first photo is real, always

Your opening image is a photograph of your actual garment, whole and in focus. Vinted requires this explicitly; on the other four it is simply what a buyer needs in order to know what they are buying. This one habit resolves most of the grey area on its own, because it means the buyer always has the real item as their reference before anything else.

The flaws are documented, unprocessed

Stains, pilling, missing buttons, worn hems, stretched cuffs: photographed straight, no processing, and described in the text. This is not just honesty — it is the single most effective defence against a "not as described" claim, which is the dispute type that costs sellers the most.

A buyer who saw the mark in the listing and bought anyway has no case. A buyer who found it on delivery has one.

The AI visual is a complement

The generated image comes after the real ones and adds information the flat photos cannot: how the garment drapes, how it reads worn, what the proportions look like on a body. It never stands alone, and it never replaces the documentary photos.

Disclose in one line

"AI-generated visual — the real item is shown in photo 1." Ten seconds, no cost, and it converts a possible sense of being tricked on delivery into an expectation set in advance. No US platform in this guide requires it today. It is still the right habit, and the direction of travel across every jurisdiction is toward more disclosure rather than less.

The checklist before you publish

Which platform rewards what

A last practical note, since the same listing rarely performs identically everywhere.

Vinted US has the most photo slots — twenty by its own rules — and the strictest published language. Use the room: multiple real angles, every flaw, and a presentation image at the end if you want one.

Depop gives you very few slots and an audience that responds to visual identity. Every image must earn its place, and the platform explicitly invites creativity in the same breath as requiring your own photos.

Mercari has the clearest authorship language of the five and a fast, transactional buyer culture. Clarity beats styling here.

eBay cares about accuracy and detail above all, and bans added text, borders and watermarks outright. In the pre-owned designer segment, close, honest detail shots are worth more than any staging.

Poshmark is social and relationship-driven, with a strong emphasis on the Covershot as the first impression. It publishes the least about photo rules, which is a reason for more care rather than less.

If you are still choosing where to concentrate your effort, we put the four side by side on fees, audience and format in Vinted vs Poshmark vs Depop vs Mercari.

If you want the underlying craft rather than the rules, our complete photo guide covers lighting, angles and first images, and AI worn photos explains the drape-and-fit problem this whole category exists to solve.

FAQ: AI Photos on US Resale Platforms

Sources

All pages read on 31 July 2026, on the platforms' and agencies' own sites.

• Vinted — Catalog rules (photos taken by you, no stock photos, item as it is, no image editing, first photo, 20 photos maximum)
• Mercari — Prohibited Conduct (no stock or advertisement photos; no photos you do not own or have not taken yourself), Marketplace Guidelines and Buyer Protection (72 hours after confirmed delivery)
• Depop — What are 'stock images'? (help centre, last updated 13 May 2025)
• eBay — Picture policy
• Poshmark — Community Guidelines ("Be Real"; no photo-specific or AI-specific rule published there as of this date) and Posh Protect
• Federal Trade Commission — Truth in advertising
• European Union — Regulation (EU) 2024/1689 (AI Act), Article 50 transparency obligations, applicable from 2 August 2026

Quotations reflect the version of each page consulted on the date above. Every platform can revise these texts at any time, and several of them will. Where a platform has published no rule on a point, this article says so explicitly rather than inferring one — that distinction is the whole reason to trust the rest of it.

Companion piece: our guide to reselling scams in the US covers the seller-side security question, including why honest listing photos are the cheapest dispute prevention available.