Member Exclusive   //   August 8, 2024

Amazon Briefing: Gen AI is taking over product listings on Amazon & sellers aren’t happy about it

This is the latest installment of the Amazon Briefing, a weekly Modern Retail+ column about the ever-changing Amazon ecosystem. More from the series →

This story has been updated on August 12 with a statement from Amazon.

Amazon sellers are bracing for more changes — and more headaches — as the e-commerce juggernaut relies more on artificial intelligence as part of its back-end operations. 

On July 31, Amazon announced on its discussion forum for sellers that the tech company will be overhauling its existing product bullet point requirements, ostensibly to make it easier for customers to compare products. Special characters, emojis and certain phrases, such as refund-related guarantees, will no longer be allowed. In other words, product listings will become more uniform across Amazon’s expansive web store.

As part of the changes, Amazon will use AI to rewrite product listings that are deemed non-compliant and replace such content with “compliant, high-quality bullet points,” according to the forum post.  “We’ll remove non-compliant content and use AI to generate compliant, high-quality bullet points. We’ll share any AI-generated quality-related improvements with you for review before we publish them to your listings,” the company also said. The changes will go into effect on Aug. 15. 

It’s all part of CEO Andy Jassy’s aggressive artificial intelligence push. Like many other tech giants, Amazon has been pouring more money into AI, ramping up spending on data centers, chips and real estate needed to power artificial intelligence. The company’s capital expenditures totaled more than $30 billion, mostly to boost AWS infrastructure, in the first half of the year, and the company expects that to increase in the second half. Amazon has also released a spate of AI-powered tools to speed up the including autogenerated product descriptions and product pages that can be created with just the copy-and-paste of a link.

“The generative AI component is in its very early days,” said Jassy on a second-quarter earnings call with analysts on Aug. 1. Amazon sellers say they’ve been gritting their teeth as the Seattle-based company tests out artificial intelligence to edit product listings, among other things, according to multiple sellers and agencies who spoke with Modern Retail. The results, they say, leave much to be desired. Until artificial intelligence is fine-tuned, agencies say sellers would more prefer to opt out of Amazon’s latest AI-focused change.

According to a statement from an Amazon spokesperson, “Amazon is dedicated to providing shoppers with accurate and informative product information, and we work to help sellers improve the quality of their product listings. The vast majority of sellers choose the option to improve their listing quality when given AI-generated suggestions, finding the suggestions largely accurate and helpful.”

The statement went on to read, “sellers can choose to accept, modify, or decline the suggestions altogether, unless there is a listing policy violation. We take all feedback and concerns seriously, and on the rare occasion that a seller believes we have made an error and they did not violate a policy, our seller support team is available 24/7 via email, phone, and chat to assist.”

“People spend a lot of time tweaking bullet points for keywords, for conversion rate, to best represent their products, their brand story and how their products compare to others,” said Gwen McShea, president of Lean Edge Marketing in Vermont, which works with about 30 Amazon sellers. “To say that AI can go out and do something better is a very far reach.”

McShea isn’t alone.

“Right now, their AI model is making updates and making a lot of mistakes,” said Jon Elder, CEO and founder of Black Label Advisor, which manages hundreds of brands. “For example, half of your data as a seller might disappear in your title, your bullet points might mention incorrect aspects of your product, whereas before, the seller had everything optimized.”

McShea has noticed recently that many clients’ listings have had brand names erroneously stripped out of the titles as a result of Amazon’s AI. “If you’re in a market competing against no-name Chinese products — that’s killer,” she said. “If you’re trying to explain that you’re the premium band and they take away your name, then you look like everyone else, and that can tank sales significantly.”

Amazon seller consultant Leah McHugh, who specializes in listing compliance at consulting firm eCommerceChris, is very familiar with the quirks of the marketplace’s AI and automation tools. For example, she has often worked on cases in which Amazon’s AI flags a listing’s bullet points as non-compliant, but upon closer inspection, McHugh finds that there hasn’t been a policy violation at all. 

She recently handled a case for a seller that sells printed products to order with more than 2 million listings that were miscategorized by the site’s AI, an error that immediately dropped the merchant’s sales by about 60% because customers weren’t able to find the products, said McHugh. It took several weeks to resolve the issue. 

McHugh said that if Amazon’s AI has trouble telling the difference between non-compliant and compliant text, it doesn’t bode well for its latest AI announcement.

For sellers who are wary of AI writing their bullet points, Amazon currently lets them unenroll. “If you don’t opt out of the program, you’re at the mercy of the AI system, and then you will have to request that Amazon change it back to what you had before,” said Elder. However, it’s unclear if this will still be an option when the changes officially go into effect on Aug. 15. But ultimately, sellers with listings violating Amazon’s policy on bullet points either have to fix them themselves or modify them using the AI-generated content suggested

Although the forum post says sellers will be able to review any changes made to a product listing by the site’s AI before it’s published, some are skeptical that this will turn out to be the case based on how Amazon currently handles AI-generated changes. It’s also unclear how long of a window sellers will be given to review such changes. 

“I wish that was the case, but their AI is already changing stuff automatically,” said Elder.

To McHugh, the latest announcement raises questions about how easily or effectively sellers will able to seek assistance from Amazon when AI gets it wrong, particularly if Amazon continues to automate back-end processes.

“If you don’t want to accept the AI-generated replacement, I’ll be curious to see how easy it is to override that,” she said. “It’s all becoming automated, and they’re not giving paths for an actual person to override it when the automation isn’t working.”

Amazon news to know

  • As it grows its media business, Amazon is positioning Amazon Prime Video similar to traditional media companies.
  • Amazon has moved up the deadline for Black Friday inventory to October 19.
  • The head of Amazon’s India operations Manish Tiwary has left the company.

What we’ve covered