# Echo chamber or filter bubble: Which one are you in? > Echo chamber and filter bubble get used as the same word for the same feed, and the mix-up hides who's doing the filtering. The research separates the room you chose from the room an algorithm built, and shows which one most readers are in. Source: https://informed.now/blog/echo-chamber-vs-filter-bubble By Kira Shishkin · October 26, 2026 · Psychology of News **Summary:** An echo chamber is a room you built by choosing like-minded sources. A filter bubble is a room an algorithm built for you by guessing what you'd click. The research says both are rarer than the headlines suggest, and the one most people sit in is the one they built themselves. The filter worth worrying about is the one you built yourself. That's the uncomfortable finding behind a decade of research into echo chambers and filter bubbles, and it cuts against the account most of us give of our feeds. We like to blame the algorithm. The data keeps pointing back at our own choices. ### What's the difference between an echo chamber and a filter bubble? The difference is who does the filtering. An echo chamber is a media space you chose. You follow people who agree with you, read outlets that share your politics, and mute the ones that don't. The Reuters Institute's 2022 [literature review](https://reutersinstitute.politics.ox.ac.uk/echo-chambers-filter-bubbles-and-polarisation-literature-review) describes it as a bounded, enclosed media space that magnifies the messages inside it and insulates you from the ones outside. The walls are yours. A filter bubble is a media space built for you. The term comes from Eli Pariser's 2011 book, and the same review treats it as an echo chamber produced by ranking algorithms rather than by the reader. You never chose the walls. A system watched what you clicked, guessed what you'd click next, and quietly stopped showing you everything else. The 2 words get used as synonyms, and that's where the confusion starts. One describes a habit, the other a machine. A habit you can fix tonight; A machine you can only fix by leaving it or by changing its settings. ### How many people are actually in one? Fewer than you'd guess, by a wide margin. The 2022 Reuters Institute review pulled together surveys and web tracking from several countries. In the UK, roughly 6% to 8% of the public sat inside a politically partisan online news echo chamber. Web tracking during the 2019 UK election found 2% of Labour voters and 4% of Conservative voters in like-minded chambers. The US was the outlier, at more than 10%. The review's own summary line is that echo chambers "are much less widespread than is commonly assumed." The number stays small partly because a lot of people read almost no online news at all. The review cites a 2021 figure of around 25% of UK internet users who touch no online news in a given week. You can't be sealed inside a room you never enter. The filter bubble fares even worse than the echo chamber in the evidence. Flaxman, Goel, and Rao published a 2016 [study](https://5harad.com/papers/bubbles.pdf) in Public Opinion Quarterly using the browsing records of 50,383 active US news readers, 2.3 billion page views in total, collected in 2013. Direct visits to news sites made up 76% of all news reading, search brought 13%, and social media brought 6%. Readers who arrived through social or search did read a slightly narrower slice of outlets than readers who typed the address in: On the study's segregation scale, direct browsing scored 0.11 for news and social scored 0.12, with the gap widening for opinion pieces (0.13 against 0.17). The same readers were also exposed to more content from the opposing side than direct browsers were. The authors called the effects "relatively modest." The Reuters review reached the same conclusion from a different direction. Across the studies it covered, search engines and social media "generally lead to slightly more diverse news use," not less. The mechanism has a name in the literature: Automated serendipity. A feed built to keep you scrolling shows you things you'd never have typed into a search box, including things you disagree with. ### Why your own choices matter more than the algorithm's The chamber most people sit in has a human architect. The Reuters review found that many readers show some selective exposure, meaning they lean toward sources they like, but little selective avoidance, meaning they rarely go out of their way to dodge the other side. The small group that does both is made up of highly partisan readers, and interest, education, and income predicted news habits more reliably than politics did. A 2021 [paper](https://pmc.ncbi.nlm.nih.gov/articles/PMC7936330/) in the Proceedings of the National Academy of Sciences by Cinelli and colleagues adds the sharpest detail. The team analyzed more than 100 million posts from more than 1 million users across 4 platforms on topics like vaccination, gun control, and abortion. On the 2 largest social networks in the study, users clustered tightly with people who shared their leaning. On the 2 forum-style sites, no such split appeared. The authors argue that the dividing line was whether the platform gave users a feed they could tune themselves. Where readers held the dials, the chambers were weaker than on the 2 networks without that control. Put the 3 studies together: The algorithm widens your diet a little, and your own hand narrows it a lot. Following 40 like-minded accounts and muting the rest builds an echo chamber with no help from any ranking system, and the ranking system is, if anything, the thing sneaking in the occasional dissent. This is the same machinery that drives [confirmation bias](https://informed.now/blog/what-is-confirmation-bias): We reach for what agrees with us because agreement feels like accuracy. A feed is a mirror of that reflex before it's a cause of it. ### Isn't this research too old to trust? The strongest case against all of this is timing. The browsing data in the 2016 study came from 2013, before short video feeds took over. The UK figures in the Reuters review come from 2019 and 2020. Feeds today are more aggressively personalized than the ones those readers used, and some of the newest platforms give you almost no way to follow anyone at all. Everything you see is a guess about you. If filter bubbles were ever going to show up in force, the argument goes, it would be now, and the research hasn't caught up. We'd concede part of that. The measurement lag is real, and the review itself warns that many of its findings come from the US or from a single platform and may not generalize. Nobody has published a tracking study of a fully algorithmic video feed at the scale of the 2016 work. Yet the 2021 4-platform study already sits on the near side of that shift, and it found the same thing: Algorithms amplified clusters that users had formed, and the platforms where users could adjust the feed showed weaker clustering. The mechanism held when the feeds got smarter, and it got easier to see. ### How to check which room you're in Audit the architect, not the room. 3 steps, no app required. 1. For 1 week, note where each news item came from and who picked it: You, a friend, a search, or a ranking. A newspaper and an RSS reader count as you. If the "you" column dominates and every source leans the same way, you've built an echo chamber and you can rebuild it. If the "ranking" column dominates, you're in a bubble and the fix is a setting or an exit. 2. Add 1 accurate source you'd never choose. Accuracy matters more than leaning, so pick it by its correction record and its sourcing, which is the test our post on [media bias](https://informed.now/blog/what-is-media-bias) lays out. 1 such source a week is enough to break the insulation the Reuters review describes. 3. Read the primary report behind 1 claim a week. Secondhand summaries carry the leaning of whoever summarized them. The primary document doesn't, and the habit of going to it is the single most reliable way out of any chamber, chosen or built. The common thread is that the choosing moves back to you. A [curated](https://informed.now/blog/curated-news-vs-algorithmic-feeds) source with a stated method beats a feed with a hidden one, because you can read the method and argue with it. We run [**informed.now**](https://informed.now) on that principle: A short daily news brief by text, with the news chosen by a stated process rather than a ranking that learns you, and with the sources named so you can go check. It's a way to put a human architect back in charge. The newspaper on your doorstep is another. The filter worth worrying about is the one you built 1 follow at a time, and the same hand can take it down.