Amazon rolls out AI-generated product review summaries

US ecommerce giant Amazon is using generative AI to distil multiple product reviews presumably written by real people into one easy to digest summary.

As if to sugar the AI pill, the Amazon blog announcing the move explains at length that when user reviews were first introduced in 1995 they were initially viewed with scepticism before being lovingly embraced. Now, we’re told, we wouldn’t dream of buying so much as a charging cable before consulting the army of strangers who felt moved to publish their experiences of the product.

Apparently, the ability to give a produce a simple ‘star rating; from one to five, without any corresponding qualitative review, was only introduced in 2019. Since it makes sense to aggregate those ratings into an average score, Amazon doesn’t see any reason why a similar trick shouldn’t be tried with the qualitative reviews.

On the scale at which Amazon operates it wouldn’t be feasible to attempt such a thing manually but thanks to recent breakthroughs in generative AI it can now be automated. “We want to make it even easier for customers to understand the common themes across reviews, and with the recent advancements in generative AI, we believe we have the technical means to address this long-standing customer need,” blogged Vaughn Schermerhorn, Director, Community Shopping at Amazon.

“Want to quickly determine what other customers are saying about a product before reading through the reviews? The new AI-powered feature provides a short paragraph right on the product detail page that highlights the product features and customer sentiment frequently mentioned across written reviews to help customers determine at a glance whether a product is right for them.”

Superficially this seems like a great idea. Most people probably just read the first few reviews, if at all, and it seems much more efficient and comprehensive to review a distillation of all of them. On the other hand, if people thought the reviews themselves had an element of AI to them they would presumably trust them less. That’s why transparency is so important with generative AI and Amazon is sensibly making this clear with a label under the summary.

But generative AI is not perfect. What biases are built into the Amazon generative AI model? How does it go about representing the full range of reviews, including negative ones, and how does it account for their relative quantities. Furthermore, how can Amazon reassure users there are no commercial incentives involved in the programming of the AI?

As with much of the current wave of AI innovations, this powerful innovation needs to be treated with caution. The chances are the AI generated summary will instantly become a very strong influencer of purchasing decisions and with that will come customer pressure to optimise it. Amazon must be seen to be as objective and neutral as possible in the generation of these summaries, which will initially be available only to some US mobile users, if it wants them to be trusted.

Get the latest news straight to your inbox. Register for the Telecoms.com newsletter here.