Amazon’s Risky Recommendations

In a landmark research collaboration, CheckFirst and AI Forensics publish a comprehensive study examining the intricate facets of Amazon's recommendation algorithm, unveiling concerning content amplification and user entrapment in dubious narratives.

Picture this. Your 13 year-old daughter is innocently browsing Amazon’s book section in search of her favorite manga and stumbles upon sexually explicit content. This is precisely one of the issues at the heart of our in-depth research. Looking for an informative book about vaccines? Just type “vaccine” in the search box and Amazon suggests 90% of anti-vaccination or conspiracy theories narratives among the top 10 results. Our common study delves into concerning aspects of how algorithms may be undermining the plurality of views, amplifying disinformation and failing to adhere to the company’s own policy on sexually explicit content. The report, titled The Amazing Library: An Analysis of Amazon’s Bookstore Algorithms within the DSA Framework also suggests Amazon may be breaching the EU’s Digital Services Act.

We naturally contacted Amazon with the results of our investigation. We found out on Friday December 8 that some problematic content had been removed from the platform including explicit content as well as DVD.

Released on Monday, December 11th, this study is the result of a meticulous analysis encompassing over 60,000 book recommendations and search queries across Amazon’s Belgian and French stores. This study emerges from our collaborative effort with AI Forensics team, an European non-profit that investigates influential and opaque algorithms to defend digital rights.

Amazon Store, a leading marketplace, is being scrutinised for issues including selling non-compliant or counterfeit products and breaching consumer protection laws. The focus on Amazon should also encompass its amplification of misleading content and societal risks, beyond sheer illegal content.

Guillaume Kuster, CEO of CheckFirst

Our research exposes how Amazon’s recommendation and search algorithms are not only promoting misleading books on Health, Immigration, Climate Change and Gender Issues, but trap users into such narratives. One a user clicks for instance on an anti-vaccination book, it would take more than 20 clicks to escape a recommendation bubble of like-minded narratives.

Amazon has 181 million users in the European Union alone. It is concerning to observe that search results for queries like “COVID” or “Vaccination” almost exclusively (8 books out of ten in average) yield anti-vaccination or conspiratorial results. In France, half of the search results related to Climate Change mislead the public on scientific consensus.

With this study, CheckFirst aims to provide insights that can inform future developments in Amazon’s algorithmic design and implementation. We are proud to contribute to the ongoing conversation about responsible algorithmic practices and their impact on information integrity, fundamental rights, civic discourse, protection of minors and democracy. Join us as we shed light on the need for a safer online environment within the framework of the DSA and its definition of systemic risks.

Do you have further questions about our work? Please feel free to contact us at any time.

🔎In June, we released another study, Facebook Hustles, an extensive month-long investigation in which our team uncovered a large-scale scam operation that recruits its victims on Facebook through ads. You can read it here.

About us

Check First is a leading Finnish software and methodologies company, spearheading adversarial research techniques. We believe that everyone should be able to understand how and why content is presented to them. We advocate for online clarity and accountability, building solutions to attain this goal. Partnering with leading institutions, regulators, NGOs and educators, we aim at curbing the spread of disinformation and foreign influence manipulations.

Our story