Determine whether content topics cause elevated bot suspicion

Determine whether political, crypto, promotional, financial, shopping, and related content categories increase public suspicion of an account when encountered, or whether these categories are merely encountered more often by users who query Botometer.

Background

The study finds that political, crypto/web3, promotional/professional, finance/trading, shopping/reselling, and retweet-to-win content is enriched among accounts checked through Botometer, while fandom and related content is depleted. However, the comparison with the ambient Twitter stream does not identify the mechanism behind this association.

In particular, the observed enrichment could mean that users become more suspicious after encountering these topics, or that Botometer users disproportionately encounter accounts producing this content for other reasons. The paper explicitly leaves this causal distinction unresolved.

References

This asymmetry is consistent with public suspicion tracking what an account advocates (promotion, partisanship, crypto) rather than how automated its behaviour looks, although the ambient-stream comparison cannot tell whether these topics increase suspicion once encountered or are simply encountered more often by Botometer users.

— When Does the Public Become Suspicious of Bots? Demand-Side Evidence from Botometer Query Logs  (2609.20661 - Elmas, 17 Sep 2026) in Section “RQ1: Which Accounts Attract Suspicion?”, subsection “Content Markers”