Etsy leverages the latest in Large Language Models (LLMs) and Generative AI to identify potentially offending content on its marketplace, saying in a post on Tuesday that it uses LLM to remove “potentially” violating or infringing items at scale. But sellers have expressed concerns about how marketplaces deploy technology to take down their listings.
Etsy consultant Scott Voelker recently discussed those concerns on his Brand Creators YouTube channel, discussing Etsy’s practices with Chris Shaffer in the following video:
In the Etsy Code as Craft blog post, it said it had over 100 million unique items for sale; “In order to continue to scale and enhance our detections through innovative products and technologies, we also leverage state-of-the-art Machine Learning solutions which we have already used to identify and remove over 100,000 violations during the past year on our marketplace.”
The blog post described one of its systems to detect policy violations that “utilizes supervised learning, a family of algorithms that uses data to train their models to recognize patterns and predict outcomes.” But it wasn’t clear whether it still requires human review before a listing is removed, as Etsy had said was the case last year when it stated: “Now, when a listing is flagged by our automated controls for potentially violating our Handmade Policy, it will remain active but not appear in search results and recommendations in order to give our specialists time to review the listing,” it said in September, adding that that the review process typically took less than 24 hours.
Sellers’ concerns about marketplaces removing listings mistakenly identified by “bots” as infringing or in violation of policy aren’t isolated to Etsy. Sellers have also asserted that eBay and Amazon’s use of AI to police listings results in erroneous takedowns.
Ironically, people are reporting that some Etsy listings are AI-generated items that are inaccurate, such as recipes and knitting and crochet patterns, which results in wasting buyers’ time and resources, according to ZDnet Senior Contributing Editor David Gewirtz.
A seller wrote about the problem on Reddit in February, saying the listing images are AI-generated and trick beginners because they look so good, but the AI-generated patterns themselves make no sense. “It pushes people away from buying patterns online from sellers who do actually design their own patterns properly and don’t use AI,” the seller said.
Etsy allows the use of AI tools in the creative process, and it’s hard to see it could meaningfully identify misleading AI-generated items like an inaccurate crochet pattern, even using AI in its detection practices.
“…review process typically took less than 24 hours…”
hm….the review is again by the same AI brainless robot, which will repeat the initial conclusion: FAILED.
This is affecting Depop too, so many people (including myself) have been banned with absolutely no help from their support team. To make matters worse you have to fight them just to get any remaining funds. I’m crossing my fingers this results in some big class action lawsuit against Etsy!!