Etsy Just Banned AI-Generated Recipes. Its Sellers Grew 5.9% Last Quarter While Its Buyers Fell 0.4%

The policy itself is narrow. Etsy is not banning generated art, and it is not banning AI-assisted photography. It is banning the category of the product rather than the label on it.

The June quarter shows the mechanism the policy is aimed at. Marketplace GMS grew 7.5% to $2.58 billion and Etsy marketplace revenue grew 9.3% to $611.5 million, yet active buyers slipped 0.4% to about 87.0 million while active sellers grew 5.9% to 5.71 million. Sellers are being added to a marketplace whose buyer base is not growing, which means more listings are competing for the same finite number of impressions. Etsy says the compounding started with AI content that buyers cannot distinguish from the real thing; the ratio does the rest.

Here is the uncomfortable arithmetic. Etsy's revenue take rate rose 130 basis points to 25.9% in the June quarter, and the company attributes the expansion first to Etsy Ads and Offsite Ads. The Form 10-Q is blunter: advertising revenue rose $17.8 million year over year, driven primarily by higher click volume on Etsy Ads, while active buyers declined. More listings mean more impressions to sell. The policy shrinks the listing pool that feeds its own ad engine.

That makes the policy expensive, because Etsy monetizes each dollar of volume roughly twice as hard as its closest comparable: 25.9 cents against about 13.8 cents at eBay in the same quarter, computed from each company's reported revenue and volume. A marketplace earning 26 cents on every sale depends on that sale converting, not on how many listings exist to fail to convert. eBay can absorb AI-assisted listings and still grow; Etsy cannot. Read as inventory management rather than moral posturing, the trade is explicit: fewer, better listings in exchange for a take rate that survives the flood. The June quarter also carried a $46.4 million net loss on discontinued operations from the Depop sale, against $114.3 million of continuing-operations profit.

A marketplace does not create its own supply. The AI listings arrive through tools that write copy and images for sellers, get paid out through rails that settle cross-border sellers, and then land in discovery surfaces that decide what buyers see. Every link in that chain is a separate company's revenue line, and each one is currently being paid to make supply cheaper. That is the structural driver behind Etsy's problem, and it is why no single policy can fix it.

The most useful number in this story has not been published, and the gap matters more than the rules themselves. Nobody outside Etsy knows how large the AI-generated inventory on its marketplace actually is, which is why the removal wave cannot be modeled and why the October 28 print cannot settle it.

Three months before the ban, Etsy sold Depop to eBay for $1.4 billion in cash, closed July 30, 2026, and on August 3 authorized a new $2 billion buyback. Read together, the divestiture, the ban and the capital return describe one strategy: shed the volume tier, keep the curated core, and hand the cash back to shareholders. Etsy has also said it will personalize results for buyers who want more or less AI, which is the same trade stated from the demand side.

The supply-side restriction sits oddly beside the demand side. Etsy has sold through ChatGPT since September 29, 2025, when OpenAI launched instant checkout with Etsy and Shopify merchants, and OpenAI walked back from that checkout model in March 2026 to reposition toward product discovery. Meanwhile Etsy's own beta includes an agentic Gifting Assistant, and its buyer profiles now cover more than 65 million shoppers with roughly 30 interests each. So Etsy is restricting machine-made supply at the same moment it is routing demand through machine-made discovery. The uncomfortable implication is that a language model cannot judge craftsmanship from a thumbnail, which is exactly why the unedited-photo rule may matter more than any label: it is the only signal that is checkable rather than declared.

Four dates do most of the near-term work. Three of them fall in one week: Etsy, eBay and Alphabet all report on October 28, and Amazon follows on October 29. None of those quarters includes a single day of the new rules, so the prints measure the marketplace before the policy, not after it. The first print that can price the ban is the February 2027 report.

Over one to three years, the question is whether verified becomes a category with a price. Etsy's matching investments are working in the near term, with feed freshness suppressing over 80% of previously viewed and engaged listings without costing GMS and the 30-day repeat purchase rate improving, while trailing-twelve-month frequency is still modestly below the prior year. If the ban removes supply that does not deserve to be found, those matching gains compound. If it only removes supply without restoring frequency, Etsy will have shrunk a marketplace whose take rate needs volume. The bull case for marketplace stocks rests on AI search lifting discovery, the sell-side view advanced in mid-September 2026, and the ban is the supply-side hedge against that same narrative.

The lesson of October 22 is not that AI listings are bad. It is that a marketplace's product is the quality of the answer, and once generation is free, the differentiator moves from supply to verification. Etsy says its own buyers cannot tell which items are machine-made, and calls that a top-three barrier to buying more. At 3.55 times trailing revenue and a 7.7% trailing free cash flow yield, the policy is close to free if buyers return and a margin leak if they do not. For investors, the near-term trade is a margin question and the long-term trade is a moat question, and the same evidence answers both.