Within Mythcraft
How Platforms Help Myths Travel Faster
Digital platforms can reward emotional, repeatable claims before careful correction has time to catch up.
On this page
- Speed and scale
- Engagement incentives
- Design choices that reduce harm
Page outline Jump by section
Introduction
Social media algorithms amplify myths by deciding which claims are seen, repeated and rewarded at scale. A false story no longer needs to travel slowly through conversation, newspapers or local rumour networks. It can be ranked into millions of feeds because it attracts clicks, comments, watch time, anger, curiosity or group loyalty before anyone has had time to check it. That does not mean algorithms invent every myth, or that users are passive victims. It means platform design changes the speed, reach and incentives of myth-making.
The clearest mechanism is simple: ranking systems often optimise for engagement, and myths are frequently built for engagement. They compress uncertainty into a memorable claim, trigger emotion, flatter identity and invite repetition. Research on false news, recommender systems and platform experiments shows that algorithmic feeds can increase the visibility of divisive or unreliable content, although the size and direction of effects vary by platform, audience and design. The governance question is therefore not whether “the algorithm” is always guilty. It is which design choices predictably reward myth-like content, how platforms measure those risks, and what obligations should follow.
Why algorithmic feeds change the life of a myth
A myth can spread without a platform: people have always repeated vivid stories that make the world feel simpler. What social media adds is automated selection. A feed-ranking system scores posts, videos or links against signals such as past behaviour, predicted clicks, comments, likes, shares, watch time, freshness, network connections and inferred interests. That ranking determines what appears first, what disappears and what is recommended to people who never asked for it.
This matters because visibility itself becomes a reward. A claim that gets early attention can be shown to more people, producing more reactions, which can justify still more distribution. Researchers describe this as a feedback loop between user behaviour and algorithmic ranking: social drivers and algorithmic mechanisms interact, making it hard to separate “what people want” from what platforms repeatedly place in front of them. [PMC]pmc.ncbi.nlm.nih.govSource details in endnotes.
For myths and misconceptions, the crucial feature is not only reach but timing. False or misleading claims can spread during the gap between event and verification: after a violent incident, a celebrity rumour, a public-health scare or a dramatic political clip. By the time officials, journalists or fact-checkers clarify what happened, the false version may already have become familiar, emotionally charged and socially useful to a community.
The classic study of Twitter rumours by Soroush Vosoughi, Deb Roy and Sinan Aral found that false news diffused farther, faster and more broadly than true news in their dataset; MIT’s summary notes that false stories were 70 per cent more likely to be retweeted than true stories. The study did not prove that ranking algorithms alone caused the effect, but it showed why engagement systems face a structural problem: falsehoods can have features that make people want to pass them on. [Science]science.orgSource details in endnotes.
Speed and scale
The first way algorithms help myths travel faster is by collapsing the distance between a small spark and a mass audience. A misleading claim that once needed repeated word of mouth can now move through recommendation surfaces, trending modules, “For You” feeds, short-video queues and reshare chains. The platform is not just a pipe; it is a sorting machine.
This sorting can turn small early differences into large visibility differences. A post that provokes outrage or fascination may generate comments from supporters and opponents alike. To a ranking system, both can look like evidence of relevance. A correction that is careful, conditional and less emotionally charged may arrive later and attract less interaction. The result is a lopsided race: the myth is short, punchy and repeatable; the correction is slower, more qualified and less socially exciting.
The same mechanism appears in crisis misinformation. A UK parliamentary report on social media, misinformation and harmful algorithms examined how platform incentives and recommender systems can contribute to the spread of harmful misleading content, including during fast-moving events where false claims can trigger offline consequences. The report also noted concerns about advertising models and monetisation around unreliable material, because visibility can be converted into revenue as well as influence. [UK Parliament]publications.parliament.ukSource details in endnotes.
The scale problem is especially acute for myths that are not a single false post but a repeatable template: “the media are hiding this”, “this cure is being suppressed”, “this event was staged”, “this group is secretly responsible”. Platforms can remove or label one item while many near-duplicates, reaction videos, screenshots and coded references keep the narrative alive. Algorithmic systems trained on engagement may then recommend the theme even when individual pieces are moderated.
Engagement incentives
The second mechanism is the incentive structure created around attention. Many social platforms make money from advertising, subscriptions, creator programmes, shopping links or data-driven targeting. Even where the exact business model differs, attention remains valuable: more time on the platform creates more chances to show ads, collect signals, sell services or keep users from migrating elsewhere.
Ranking systems therefore tend to privilege content predicted to keep users active. A 2025 PNAS Nexus audit of Twitter’s engagement-based ranking found that optimising for what users engage with can amplify divisive content more than ranking based on what users say they want to see. The authors frame the problem as a gap between revealed preferences, such as clicks and likes, and reflective preferences, such as what users later judge to be valuable or healthy. [OUP Academic]academic.oup.comSource details in endnotes.
That gap is central to myth amplification. People may click a conspiracy thread because it is disturbing, reply to a false claim because they are angry, or watch a misleading video because it is outrageous. Those actions do not necessarily mean they endorse the content. But if the system reads engagement as satisfaction, it may learn to surface more of the same.
This is why “people chose to share it” is only a partial explanation. Users provide signals, but platforms decide how much weight to give those signals. A like, a hostile comment, a long pause, a share into a private group and a completed video view are not neutral facts; they become governance choices once a platform turns them into ranking power.
A useful way to think about the mechanism is:
- Emotion becomes a ranking signal when anger, shock or amusement drives interaction.
- Repetition becomes credibility when users see the same claim across multiple creators, groups or formats.
- Identity becomes distribution when a claim performs well inside a community and is then recommended to similar users.
- Monetisation becomes reinforcement when creators learn that myth-like content brings attention, followers or income.
None of these steps requires a platform to intend deception. The harm can emerge from ordinary optimisation: recommend what keeps people watching, commenting and returning.
When the evidence is mixed
A careful account has to avoid a myth about algorithms too: the idea that every social problem caused by misinformation can be traced neatly to a feed-ranking system. The evidence is more complicated.
Some platform experiments show strong algorithmic influence. Twitter’s own large-scale study, later published in PNAS, found that its home-timeline algorithm amplified political content unevenly across countries, with the mainstream political right receiving higher algorithmic amplification than the mainstream political left in six of seven countries studied. The authors did not find support for the simple claim that the algorithm mainly amplified political extremes over moderates, which is a reminder that algorithmic effects can be specific rather than universally sensational. [PNAS]pnas.orgOpen source on pnas.org.
Other experiments complicate the story. A 2023 Science study on Facebook and Instagram tested chronological feeds during the 2020 US election period and found that switching users away from algorithmic feeds changed what they saw and how they used the platforms, but did not significantly reduce affective polarisation, issue polarisation or political knowledge gaps during the study period. Nature’s coverage summarised the result as evidence that tweaking feeds is “no easy fix” for political polarisation. [Science]science.orgSource details in endnotes.
The most reasonable conclusion is not that algorithms are harmless. It is that their effects depend on the platform, the ranking objective, the user’s existing network, the topic, the time window and the outcome being measured. A feed can affect exposure to unreliable content without immediately changing a person’s political identity. It can increase engagement with a myth without being the original source of the belief. It can amplify some kinds of misinformation while suppressing others.
For myth amplification, this distinction matters. The governance problem is not solved by asking whether algorithms single-handedly “cause” false belief. The more practical question is whether a design predictably increases exposure, repetition, monetisation or social proof for claims that are false, misleading or unverified.
The myth-making loop
The algorithmic amplification of myths usually works as a loop rather than a one-way broadcast.
First, a claim appears in a form that is easy to react to: a short caption, a provocative video, a screenshot, a misleading chart, a confident thread or a clipped quotation. The claim does not need to be fully believed by everyone who sees it. It only needs to invite enough attention to register as lively.
Second, users respond. Some share because they agree. Others comment to mock or correct it. Some save it for later, stitch it into a reaction video, quote-post it angrily, or send it privately. The platform can interpret many of these behaviours as evidence that the content matters.
Third, the system tests the content on wider or adjacent audiences. If it performs well among users with similar interests, networks or viewing histories, it may be recommended beyond the original audience. This is where a misconception can jump from a niche group into mainstream visibility.
Fourth, creators adapt. Once people learn which claims gain reach, they produce more content with the same shape: sharper claims, stronger emotion, more certainty, more “hidden truth” framing. The myth becomes a format, not just a belief.
Finally, repetition hardens the story. The claim appears from multiple accounts, in multiple styles, across platforms. Even users who doubt it may remember the association. The correction has to fight not only a false statement but a feeling of familiarity.
This loop helps explain why platform governance cannot rely only on deleting individual false posts. By the time a specific item is removed or labelled, the attention pattern may already have taught creators what works and taught the system what holds attention.
Design choices that reduce harm
Reducing algorithmic myth amplification does not require a single universal fix. It requires design choices that change what gets rewarded, how quickly risky content is boosted, and how much users and regulators can understand about the system.
The first choice is friction before virality. Platforms can slow the spread of rapidly accelerating content when it concerns breaking news, public safety, elections or health. This can mean prompts before resharing, limits on forwarding, temporary downranking of unverified claims, or routing borderline material into review before it is recommended widely. The point is not to ban uncertainty; it is to stop the system treating early outrage as proof of reliability.
The second choice is ranking quality, not just engagement. Meta’s Transparency Center says the company demotes categories including fact-checked misinformation and some borderline content, showing that large platforms already use distribution reduction as a moderation tool rather than relying only on removal. The unresolved issue is transparency: outsiders often cannot tell how strong those demotions are, how consistently they apply, or whether engagement objectives later compensate for them. [Transparency]transparency.meta.comTransparency Types of content we demoteTransparency Types of content we demote
The third choice is user control that is meaningful by default. Under the EU Digital Services Act, online platforms using recommender systems must explain the main parameters of those systems in clear language and provide options for users to modify or influence them. The wider DSA framework also imposes stronger obligations on very large online platforms and search engines because of their scale and systemic risks. [Digital Services Act]eu-digital-services-act.comSource details in endnotes.
The fourth choice is risk assessment before deployment. Ofcom’s Online Safety Act work in the UK requires regulated services to assess and mitigate how algorithms affect the likelihood that users encounter illegal content or content harmful to children. UK parliamentary material also highlights recommender-system pre-testing and children’s protections as part of the regulatory response to algorithmic harm. [UK Parliament Committees]committees.parliament.ukSource details in endnotes.
The fifth choice is independent access for researchers and auditors. Without access to platform data, public debate falls back on anecdotes, leaks, small external audits or company-selected studies. That is inadequate for systems that shape public knowledge at national and global scale. The mixed evidence from Twitter, X, Facebook and Instagram shows why independent, repeated and platform-specific auditing matters: different systems can produce different effects, and design changes can alter results over time.
What good governance should measure
The governance challenge is to measure amplification as a system outcome, not just misinformation as a content category. A platform can truthfully say it removed millions of posts while still running a ranking system that rewards the next wave of emotionally optimised falsehoods. A better assessment asks what the system made more visible, more profitable and more repeatable.
[Useful measures include:]publications.parliament.ukSource details in endnotes.
- Exposure: how many users saw false, misleading or unverified claims, and how often.
- Velocity: how quickly a claim moved from a small audience to a large one.
- Recommendation share: how much of the reach came from algorithmic recommendation rather than direct following.
- Correction parity: whether accurate corrections reached the same audiences as the myth.
- Monetisation: whether accounts or sites spreading myths earned ad revenue, creator payments or traffic.
- Repeat-offender reach: whether accounts repeatedly associated with unreliable claims continued to receive broad distribution.
- User welfare: whether users later judged the recommended content useful, trustworthy or harmful, rather than merely engaging with it.
These measures shift the conversation away from a sterile choice between censorship and doing nothing. The question becomes whether a platform is designing for public reliability as well as attention.
What readers should remember
Social media algorithms do not create myths from nothing. They amplify the claims, formats and creators that perform well under the rules of the platform. When those rules reward speed, emotion, repetition and engagement, myths gain a structural advantage over careful correction.
The strongest evidence does not support a simple claim that algorithms always radicalise everyone or that chronological feeds would fix misinformation. It supports a more precise conclusion: algorithmic ranking can shape what people encounter, which claims gain momentum, how creators adapt, and whether false stories become familiar before they are checked. That is enough to make platform design a central part of modern myth-making.
A myth travels fastest when it feels true, asks little of the reader, and gives people a reason to react. Social media algorithms can turn that reaction into distribution. Governance is the work of changing those incentives so that platforms do not treat the most repeatable claim as the most reliable one.
eBay marketplace picks
Marketplace Samples
Example marketplace items related to this page. Use the search link to explore similar finds on eBay.
Endnotes
-
Source: pmc.ncbi.nlm.nih.gov
Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC11373151/ -
Source: academic.oup.com
Link: https://academic.oup.com/pnasnexus/article/4/3/pgaf062/8052060 -
Source: news.mit.edu
Title: study twitter false news travels faster true stories 0308
Link: https://news.mit.edu/2018/study-twitter-false-news-travels-faster-true-stories-0308 -
Source: publications.parliament.uk
Link: https://publications.parliament.uk/pa/cm5901/cmselect/cmsctech/441/report.html -
Source: publications.parliament.uk
Link: https://publications.parliament.uk/pa/cm5901/cmselect/cmsctech/1397/report.html -
Source: pnas.org
Link: https://www.pnas.org/doi/10.1073/pnas.2025334119 -
Source: nature.com
Link: https://www.nature.com/articles/d41586-023-02420-z -
Source: transparency.meta.com
Title: Transparency Types of content we demote
Link: https://transparency.meta.com/features/approach-to-ranking/types-of-content-we-demote/ -
Source: transparency.meta.com
Title: Transparency Misinformation
Link: https://transparency.meta.com/policies/community-standards/misinformation/ -
Source: eu-digital-services-act.com
Link: https://www.eu-digital-services-act.com/Digital_Services_Act_Article_27.html -
Source: committees.parliament.uk
Link: https://committees.parliament.uk/writtenevidence/140786/pdf/ -
Source: committees.parliament.uk
Link: https://committees.parliament.uk/writtenevidence/132995/pdf/ -
Source: committees.parliament.uk
Link: https://committees.parliament.uk/publications/48745/documents/258221/default/ -
Source: committees.parliament.uk
Link: https://committees.parliament.uk/writtenevidence/132906/pdf/ -
Source: nature.com
Link: https://www.nature.com/articles/d41586-018-02934-x -
Source: nature.com
Link: https://www.nature.com/articles/s41586-026-10098-2 -
Source: facebook.com
Link: https://www.facebook.com/franceintheus/posts/fake-news-receives-70-more-retweets-than-the-facts-when-algorithms-prioritize-us/1307434554763003/ -
Source: facebook.com
Link: https://www.facebook.com/groups/WakeUpCarlsbad/posts/881120962066918/ -
Source: facebook.com
Link: https://www.facebook.com/the.university.of.oxford/posts/new-an-oxford-study-has-analysed-10-million-posts-across-seven-social-media-plat/1278837157624263/?locale=es_LA -
Source: facebook.com
Title: internal research documents some previously unreported help to explain why faceb
Link: https://www.facebook.com/seattletimes/posts/internal-research-documents-some-previously-unreported-help-to-explain-why-faceb/10159654162841215/ -
Source: facebook.com
Link: https://www.facebook.com/NaturePortfolioJournals/posts/a-study-in-nature-human-behaviour-found-that-deceptive-networks-reached-over-37-/1437233908432871/ -
Source: facebook.com
Link: https://www.facebook.com/TheNationThailand/posts/meta-has-unveiled-tougher-measures-on-facebook-to-support-original-creators-upda/1358838822936115/ -
Source: facebook.com
Title: the powerful algorithms used by facebook and instagram to deliver content to use
Link: https://www.facebook.com/8News/posts/the-powerful-algorithms-used-by-facebook-and-instagram-to-deliver-content-to-use/686709300168280/ -
Source: pnas.org
Link: https://www.pnas.org/doi/10.1073/pnas.1517441113 -
Source: pnas.org
Link: https://www.pnas.org/doi/10.1073/pnas.2321584121 -
Source: science.org
Link: https://www.science.org/doi/10.1126/science.aap9559 -
Source: science.org
Link: https://www.science.org/doi/10.1126/science.abp9364 -
Source: sciencedirect.com
Link: https://www.sciencedirect.com/science/article/pii/S0047272726000253 -
Source: sciencedirect.com
Link: https://www.sciencedirect.com/science/article/pii/S0306457325001840 -
Source: people.umass.edu
Title: Science ELetter
Link: https://people.umass.edu/~cbagchi/assets/pdf/ScienceELetter.pdf -
Source: mccannfitzgerald.com
Link: https://www.mccannfitzgerald.com/knowledge/disputes/digital-services-act-obligations-of-very-large-online-service-providers-eu-transparency-database-and-enforcement -
Source: science.org
Link: https://www.science.org/doi/10.1126/science.adu5584 -
Source: osome.iu.edu
Title: science eletter social media algorithms can curb misinformation but do they
Link: https://osome.iu.edu/research/blog/science-eletter-social-media-algorithms-can-curb-misinformation-but-do-they -
Source: library.queens.edu
Link: https://library.queens.edu/misinformation-on-social-media/algorithms
Additional References
-
Source: youtube.com
Title: The Feedback Loop: How Social Media Algorithms Shape Reality
Link: https://www.youtube.com/watch?v=QqQkjAOIbAoSource snippet
How Recommender Systems Fuel Online Echo Chambers...
-
Source: youtube.com
Title: How Social Media Algorithms Amplify Misinformation
Link: https://www.youtube.com/watch?v=F7J4b7042aESource snippet
The Feedback Loop: How Social Media Algorithms Shape Reality...
-
Source: youtube.com
Title: The Economics of Engagement: Why Myths Go Viral
Link: https://www.youtube.com/watch?v=Aqk5jF8jN5wSource snippet
Understanding How Social Media Ranking Systems Influence Belief...
-
Source: disinfocode.eu
Link: https://disinfocode.eu/reports/download/38 -
Source: brookings.edu
Link: https://www.brookings.edu/articles/how-misinformation-spreads-on-social-media-and-what-to-do-about-it/ -
Source: knightcolumbia.org
Link: https://knightcolumbia.org/content/engagement-user-satisfaction-and-the-amplification-of-divisive-content-on-social-media -
Source: reddit.com
Link: https://www.reddit.com/r/centrist/comments/1i5gag0/leftwing_bias_on_twitter_vs_rightwing_bias_on_x/ -
Source: 5rightsfoundation.com
Link: https://5rightsfoundation.com/wp-content/uploads/2024/08/ofcom-illegal-harms-consultation-final-submission-2-3-2.pdf -
Source: ifo.de
Link: https://www.ifo.de/en/cesifo/publications/2026/working-paper/ranking-engagement-how-social-media-algorithms-fuel-misinformation -
Source: fph.org.uk
Link: https://www.fph.org.uk/media/hoejpp0s/social-media-consultation-fph-response.pdf
Topic Tree