Amazon tests AI for synthesised reviews

Amazon is reportedly testing AI-generated summaries of product reviews on its retail site, as brands, online platforms, and agencies work out how best to deploy the technology.

Why it matters

Product reviews are powerful, with some studies finding that 71% of shoppers say they make a difference to their purchasing decisions.

While quantity of reviews is also very important, AI presents a double edged-sword here, as the technological developments that allow these benefits are also fuelling an explosion of cheaply generated content, of which fake reviews are likely to be just one outcome.

Amazon reviews

It appears that the e-commerce giant is looking to AI to solve multiple problems around product reviews:

• CNBC reports that the company is testing a summarisation feature of customer reviews which creates a short piece of text highlighting common elements of feedback – both good and bad. The feature was first spotted by the CTO of Amazon marketing agency Fortress Brand, Mark Wieczorek, writing on LinkedIn, news later confirmed by Amazon.
• But the above feature is only useful if the reviews it draws on are truthful. The BBC reports that the company is also deploying the technology to weed out fake reviews, which can skew purchases by artificially increasing the number of reviews or the average rating. Consumer groups, meanwhile, say the measures are not yet enough to tackle the problem.

The AI context

Since the emergence of OpenAI’s ChatGPT bot in November 2022, the AI conversation has filtered into mainstream news, conference keynotes, and analyst calls, largely irrespective of sector or category. Most leaders are tentative, if broadly optimistic, about a technology that threatens as much as it promises.

• While the most prominently discussed threats surround questions of human labour, the proliferation of cheap, plentiful content unconcerned with the concept of truth – or “bullshit”, as the American Philosopher Harry G. Frankfurt termed it.
• At its more banal, recent stories in this vein come in the form of Google’s AI search engine allegedly plagiarising websites word for word without then sending the reader to the website. But at least that information stems from human editors.
• In the EU, meanwhile, legislation is now entering parliament which aims to restrict where AI applications may be used in “high risk” areas such as health or fundamental rights, where mistakes are most costly.

Sourced from CNBC, BBC, WARC, Marketing Brew, Tom’s Hardware, The Register