If you’ve spent extended time in TikTok, you’ve probably seen a creator complain that they have been “shadowbanned” -- that their content is being suppressed by the app’s algorithm instead of being shared with users. Reports of shadowbans are especially common with videos that engage with controversial topics. For many users of the social internet, content filters and opaque algorithms that suppress certain topics are an everyday reality they’ve come to expect. Where one doesn’t expect this censorship is in library research.
Librarians were therefore shocked to discover that Clarivate’s Research Assistants, AI tools used by some libraries that help patrons discover resources, responded to controversial searches with error messages. Topics including genocide, lynchings, and COVID-19 would elicit the response “we couldn’t generate an answer for your question”. Sources related to these topics were, effectively, shadowbanned –- still technically available in the library catalog, but not presented to users.
Apparently, these topics were flagged by the content filtering system used by Azure OpenAI, the LLM underpinning the research assistant tools. Thankfully, this particular issue has been resolved by adjusting the content filter settings, but given the popularity of AI, we will undoubtedly encounter some version of this problem again. It is therefore imperative when using a new tool that we consider not only what it offers us, but also what it blocks. We must constantly advocate for and exercise our freedom to research controversial topics.
Sources:
Singley, Jay. “‘We Couldn’t Generate an Answer for Your Question.’” ACRLog, 21 July 2025, https://acrlog.org/2025/07/21/we-couldnt-generate-an-answer-for-your-question/.
Tay, Aaron. “The AI Powered Library Search That Refused to Search.” Substack newsletter. Aaron Tay’s Musings about Librarianship, 28 July 2025, https://aarontay.substack.com/p/the-ai-powered-library-search-that.
0 Comments.