Amazon Is Rewriting Your Listings Without Asking. Here's How to Stop It.

Amazon Is Rewriting Your Listings Without Asking. Here's How to Stop It. Speaker 1: Amazon is scraping over 2,000 brand websites, rewriting product titles, swapping images and changing bullet points across millions of listings without asking. They call it ProjectStarfish and it's already affecting over 900,000 sellers. One seller's product title got replaced with I'm sorry but I cannot fulfill this request. That's a ChatGPT error message live on a product detail page on Amazon. Another seller's AI generated review summary tanked their sales by 75%. I'm going to show you in this video exactly what Amazon Ads is doing, the horror stories you need to know about and the eight Step defensive playbook to protect your listing before it hits you. There's a new tool that Amazon quietly released called BrandCatalogLock that most sellers don't know even exists. I'm going to show you how to activate it but first you need to understand what's actually happening behind the scenes because this is bigger than a few bad title rewrites. What ProjectStarfish actually is? The $7.5 billion initiative project Starfish got reported by Business Insider in July 2025. It's Amazon's internal AI initiative and the goal is to make Amazon the best source of product information for all products worldwide. Here's how it works In four steps. So one, data collection. Amazon's AI scrapes brand websites, third-party retailer pages, manufacturer databases, product reviews, Q&A sections, anything it can find about your product online. Step two, data synthesis. When it finds conflicting information, it prioritizes brand registry content. But if your fields are empty or incomplete, it fills the gaps with whatever it scraped and it corrects what it thinks are errors, even if your version was right. So step three, content generation. Using LLMs, Starfish rewrites titles for conversations, generates bullet points, creates product descriptions and even auto generates titles for deployment. Amazon pushes these changes live, sometimes with a notification email, sometimes without. Here's the number that tells you how serious Amazon is about this. An internal document estimated Project Starfish would contribute 7.5 billion in Extra Gross Merchandise Sales 7.5 Billion. That's how much Amazon thinks better listing content is worth and the adoption numbers back it up. 90% of sellers accept AI-suggested changes. 40% increased the listing quality scores. 250 million shoppers used Rufus last year and that's up 140% year-over-year and shoppers who interact with Rufus are 60% more likely To complete a purchase. Amazon doesn't see this as a problem, they see this as the future of their catalog. They optimize for shoppers not sellers. That's the tension. Amazon AI is making decisions about your listing content based on what it thinks converts best and you might disagree with those decisions but if you don't act the changes go live anyway. Now let me walk through what's happening to real sellers right now because this is not theoretical A supplement seller used Amazon's Generate Listing content tool when it created a new listing. The AI auto-generated the titles, the bullets, the descriptions that included prohibited health claims. Three weeks later, Amazon's compliance bots flagged the AI-generated content. The listing got removed within 48 hours before the seller could review it. They've already manufactured the product and shipped the inventory into Amazon's FBA. When they tried to relist the correct content, Amazon Catalog retained the errors that AI generated. They couldn't overwrite it. Seller support kept sending scripted responses. The bad data persisted Even after the listing was deleted. Think about that. Amazon's own AI generated content that violated Amazon's own policies. Then Amazon punished the seller for it. And it gets worse. Brand registered manufacturers, not resellers, the actual brand owners are being locked out of editing their own listing. One seller reported that Amazon changed their title to something that violated Amazon's own style guide. When the seller tried to fix it, Amazon rejected the correction saying the current title doesn't meet the style guidelines. The title Amazon wrote. Another seller's only option from seller support was delete the listing and relist under the same SKU with a new title. For a title change, images are getting swapped too. One seller said Amazon Change my listing image with no way to make my own image. Their images aren't even better. They're way worse and I have no idea where they got them. Variations are getting split. One seller woke up to find 13 out of 16 child ASINs had been broken into separate listings with separate reviews. No trigger, no notification. A sports supplement got reclassified into a baby category. Another product got flagged As an adult, killing all organic visibility overnight. Here's the pattern across all of these. Amazon's AI makes the change. The seller notices something wrong. The seller tries to fix it. The fix doesn't go through because Amazon's system now treats the AI-generated content as authoritative. And seller support can't help because they're dealing with an automated system they don't control. The sellers who get hit The hardest are the ones with incomplete listings, empty back-end fields, missing attributes, gaps in data because Starfish was designed to fill gaps. If your listing has holes, Amazon's AI will fill them and once it does, getting your version back is a fight. If you're running a brand on Amazon and this concerns you, it should. Subscribe because I'm gonna keep covering what's changing and what to do about it. Now let me give you the Playbook. This is the most simple and most important thing you can do. Amazon launched BrandCatalogLock as part of their 2025 Brand Registry update. When it's active, unauthorized users, including in many cases Amazon's own AI, can't change your titles, images, bullets, or descriptions. Only approved brand roles can make edits. You need a registered or pending trademark. Brand Registry enrollment and correct contributor roles assigned. Request it through Brand Registry's content protection tools. If you don't have this turned on, stop this video right now and go and do that. Now, defense two, fill every back-end field. I did an entire video on this, the Rufus back-end optimization. Link in the description somewhere, but the short version is, if your fields are empty, Starfish fills them, often incorrectly. If your fields are complete and consistent, Amazon's Amazon AI is less likely to override because it's prioritized brand registry content as the authoritative source. Every empty field is an invitation for Amazon's AI to write something you don't approve. Flat files expose more fields than the seller central UI. They give you stronger catalog authority and critically, they serve as documentation of your intended content if you need to dispute AI changes later. If you've uploaded it via flat files, you have a record. If you typed it into the UI and Amazon overrides it, good luck proving what you originally entered. And then Defense 4, monitor your listings. Set up monitoring for at least 20% of your SKUs. Track title changes, bullet modifications, image swaps, category reassignments and variation splits. There are tools for this. Sentry Kit for Variation Monitoring, BindWise for Seller Pulse. Even a simple daily check of your top sellers catches most problems before they compound. Now, Defense 5. Audit Your Brand Website. This one's overlooked. Starfish scrapes your website. If your website has outdated specs, inconsistent descriptions, or old product information, that's what Amazon's AI pulls in. Make sure your website product pages match your Amazon listings exactly. Same name, same specs, same features. Remove anything outdated. Because if there's a conflict between your website and Amazon's listing, Starfish uses whatever it finds. As one expert put it, if your original content is bad, the AI version will be worse. Garbage in, garbage out. Now, defense 6. Respond to notification emails immediately. Amazon sometimes sends advanced notification emails about planning AI changes. You typically get a two-week window to review and reject. The problem? These emails look like routine compliance emails. They're easy to miss. Set up email filters to flag anything from Amazon about listing updates. Have someone check them daily because if you don't respond, the changes go live automatically. Defense 6. Never use generated listing content without reviewing every word. If you're creating a new listing, don't let Amazon's AI write it for you. Pre-write everything offline. Upload via flat file. If you don't use the AI-generated tool, review every single line before publishing, especially in regulated categories like supplements, Health and Beauty. The AI doesn't know your compliance requirements. Defense 8. Know your escalation points. When standard editing fails, and it will for some of you, there are specific escalation routes. Submit identical changes on multiple ASIN offers simultaneously. Use the Investigate Other Product Issues path and select My Issues Is Not Listed. That forces humans to review. Contact your advertising rep to escalate internally. Open a BrandRegistry support case requesting BrandCatalogLock specifically. Update your brand website to match your content you want on Amazon. Then, reference it in your support case. Here's the principle behind all 8 of those. The more complete and consistent your listing data is across Seller Central, FlatFiles, and your brand website, the harder it is for Amazon's AI to justify Overwriting it. Starfish fills gaps, don't give it gaps to fill. So here's where we land. Amazon built the largest product knowledge base in the world. That's what ProjectStarfish is. They're going to scrape every source they find, rewrite whatever thinks converts better and push it live across the catalog. You can fight that or you can get ahead of it. The sellers who fill every field, lock their catalog, monitor daily and keep their website clean, they don't get hit by this because Amazon's AI looks at their listing and finds nothing to fix. The best defense against AI rewriting your listing is making sure there's nothing left for it to rewrite. If you haven't seen it, I did a full breakdown on the back-end fields that Rufus uses to rank your products. That's an offensive side of this. Today was defensive. That video is how you score. So the link is in the description and somewhere here. And if Amazon's AI has already messed with your listing, drop a comment. I want to hear what happened and how you fixed it.