Field guide · Research explained
Are filter bubbles
real?
A filter bubble is a useful name for a concern: that the information reaching someone becomes narrow or repetitive. It is not a diagnosis you can make from a list of followed accounts, and no single study settles how often it happens everywhere.
OffFeed field guide · Published October 7, 2026
What would we need to measure?
First distinguish available information, what a network shares, what a platform displays, and what a person opens. Each stage can narrow or widen the next. A follow list tells us something about potential sources. It does not show every ranking decision, impression, click, or article read. The answer also changes when the question is about political viewpoints, news sources, or broad subject areas.
What have studies found?
Bakshy, Messing, and Adamic studied 10.1 million U.S. Facebook users in a 2015 study of ideologically diverse news. They measured friend networks, ranked-feed exposure, and clicks separately. In that population, individual choices limited exposure to cross-cutting content more than algorithmic ranking did. The finding concerns that platform, period, and type of content; it does not mean algorithms never matter.
Flaxman, Goel, and Rao studied online news consumption using browsing data. Their 2016 paper found mixed effects: social media and search were associated with greater ideological segregation than direct visits, while also exposing people to some articles from opposing views. These results resist a simple yes-or-no story. The study’s data and definitions also cannot describe every modern feed.
A later experiment assigned consenting Facebook and Instagram users to chronological rather than ranked feeds during the 2020 U.S. election. The change altered what they saw, but the researchers did not detect significant changes in polarization or political knowledge over the three-month study. That finding does not show that ranking has no effects in every setting.
A separate experiment on X, published in 2026, found that moving active U.S. users from a chronological to an algorithmic feed for seven weeks shifted some political attitudes and increased engagement. Moving in the other direction did not have comparable effects. The researchers also observed changes in which accounts participants followed. A follow list can therefore reflect earlier recommendations as well as a person’s deliberate choices; it still cannot reveal every post they actually saw.
What can you do in practice?
For a week, note which stories you actually opened and where they came from. Compare subjects and independent sources, not just account counts. Look for a well-sourced piece that adds context to a familiar subject, then a credible source covering an adjacent subject. If a piece challenges your view, check its evidence rather than treating disagreement itself as proof of value. Our information diet guide gives a short audit you can use.
How OffFeed uses the idea
OffFeed estimates topic mix from ten answers and, if you opt in, public X follow descriptions or YouTube channel descriptions. It does not label people as being inside or outside a filter bubble. It does not infer ideology. Its map can suggest topics to explore, with the uncertainty explained in the methodology.
Original research
- Bakshy, Messing & Adamic, “Exposure to ideologically diverse news and opinion on Facebook”, Science (2015).
- Flaxman, Goel & Rao, “Filter Bubbles, Echo Chambers, and Online News Consumption”, Public Opinion Quarterly (2016).
- Guess and colleagues, “How do social media feed algorithms affect attitudes and behavior in an election campaign?”, Science (2023).
- Gauthier and colleagues, “The political effects of X’s feed algorithm”, Nature (2026).