Branchoria Think Before Sharing Think Before Sharing

Practical judgement for social feeds, AI answers, and viral claims.

Within AI Virality

What Community Notes can and cannot prove

Community Notes data shows what users flagged as misleading, but it cannot capture every AI-made falsehood on X.

On this page

  • Why researchers use Community Notes data
  • What flagged posts leave out
  • How to read platform evidence carefully
Preview for What Community Notes can and cannot prove

Introduction

Community Notes has become one of the most useful public datasets for studying AI-generated misinformation on X. Researchers value it because it provides a large, transparent record of posts that real users considered misleading enough to annotate with corrective context. In a field where many misleading posts are deleted, altered, or difficult to identify at scale, Community Notes offers a rare window into what users actually encountered and challenged on the platform. However, it is not a complete record of AI misinformation. It captures only a subset of misleading content: posts that attracted attention, were noticed by contributors, and entered the Community Notes process. Understanding both the strengths and limitations of this evidence is essential for critical thinking about claims that AI-generated misinformation is spreading on social media.

Notes Data illustration 1

Why researchers use Community Notes data

Community Notes is a crowdsourced fact-checking system on X that allows contributors to write and rate explanatory notes on potentially misleading posts. X publishes extensive data from this system, making it one of the few large-scale moderation datasets available for independent research. [X (formerly Twitter)]communitynotes.x.comX (formerly Twitter)Community NotesCommunity Notes aims to create a better-informed world, by empowering people on X to collaboratively a…

For researchers studying AI-generated misinformation, this openness has several advantages:

  • It identifies specific posts that users considered misleading.
  • It preserves information about how those posts spread.
  • It records the community response to disputed content.
  • It allows large-scale quantitative analysis rather than isolated case studies.

One influential study analysed 91,452 misleading posts identified through Community Notes and compared AI-generated misinformation with conventional misinformation on X. The researchers found that AI-generated misleading content was more likely to focus on entertainment, came more often from smaller accounts, and was significantly more likely to go viral despite being judged slightly less believable and harmful than traditional misinformation. [arXiv]arxiv.orgarXiv Characterizing AI-Generated Misinformation on Social MediaCharacterizing AI-Generated Misinformation on Social MediaMay 15, 2025…Published: May 15, 2025

Another study used Community Notes data to track synthetic images and videos on X. Researchers identified hundreds of posts containing AI-generated media that collectively accumulated more than a billion views. The dataset revealed a sharp increase in synthetic media after major advances in image-generation tools and showed that most flagged content was not overtly political but instead centred on humour, spectacle, or popular culture. [Misinformation Review]misinforeview.hks.harvard.eduthe spread of synthetic media on xMisinformation ReviewThe spread of synthetic media on XJun 3, 2024 — Here, results show a sharp rise in the frequency of mentions of AI-g…

These findings illustrate why Community Notes data is valuable evidence. It captures misinformation that was visible enough, engaging enough, and widely shared enough for users to react to it.

Community Notes are a Bad Idea

Channel: Sabine Hossenfelder · Views: 172.1K · Uploaded: January 2025 · Length: 14 minutes

Open on YouTube

What Community Notes reveals about virality

A key strength of the dataset is that it focuses attention on content that circulated in public view rather than hypothetical threats.

Many discussions of AI misinformation focus on what synthetic media could do. Community Notes data instead shows what people actually encountered and attempted to correct. This distinction matters because the most viral AI-generated content is not always the most sophisticated deepfake.

Research using Community Notes datasets repeatedly finds that misleading AI content often succeeds through attention and shareability rather than through perfect deception. Synthetic images, fabricated screenshots, altered photographs, and humorous visual hoaxes frequently generate large audiences even when some viewers recognise them as fake. [arXiv]arxiv.orgarXiv Characterizing AI-Generated Misinformation on Social MediaCharacterizing AI-Generated Misinformation on Social MediaMay 15, 2025…Published: May 15, 2025

The dataset therefore helps answer a practical question: which kinds of AI-generated misinformation are spreading widely enough for communities to notice? In many cases, the answer is not only political propaganda but also entertainment-focused content that travels rapidly through reposts, screenshots, and algorithmic amplification. [arXiv]arxiv.orgarXiv Characterizing AI-Generated Misinformation on Social MediaCharacterizing AI-Generated Misinformation on Social MediaMay 15, 2025…Published: May 15, 2025

Notes Data illustration 2

What flagged posts leave out

The usefulness of Community Notes data should not be confused with completeness.

A post appears in the dataset only after several conditions are met. Someone must notice it, consider it misleading, write a note, and engage with the Community Notes system. Many misleading posts never enter this process at all. [X (formerly Twitter)]communitynotes.x.comX (formerly Twitter)Community NotesCommunity Notes aims to create a better-informed world, by empowering people on X to collaboratively a…

Several important categories may therefore be underrepresented:

  • Low-visibility misinformation that never reaches large audiences.
  • Misleading content shared mainly within private groups or direct messages.
  • Posts that spread quickly and disappear before contributors respond.
  • AI-generated content that users mistake for authentic material and therefore never flag.
  • Posts that receive proposed notes but never achieve the consensus required for public display.

Researchers have repeatedly noted that Community Notes can be slow relative to the speed of viral content. One large study found that fact-checking interventions often arrive after much of a post’s early engagement has already occurred, limiting their ability to capture the full lifecycle of misinformation. [arXiv]arxiv.orgDid the Roll-Out of Community Notes Reduce Engagement With Misinformation on X/Twitter?July 16, 2023…Published: July 16, 2023

This means that Community Notes data is best understood as evidence of detected misinformation rather than all misinformation.

‘You're not going to be able to trust what you see’: top fact-checkers on Meta

Channel: MS NOW · Views: 46.3K · Uploaded: January 2025 · Length: 9 minutes

Open on YouTube

How to read platform evidence carefully

Community Notes datasets can support strong conclusions, but only within clear boundaries.

A finding such as “AI-generated misinformation is more likely to go viral than non-AI misinformation” applies to the sample of misleading posts that entered the Community Notes system. It does not automatically prove that all AI-generated misinformation on X behaves the same way. [arXiv]arxiv.orgarXiv Characterizing AI-Generated Misinformation on Social MediaCharacterizing AI-Generated Misinformation on Social MediaMay 15, 2025…Published: May 15, 2025

Similarly, a rise in flagged synthetic media shows that users encountered more content they believed was AI-generated and misleading. It does not prove that every AI-generated post was harmful, deceptive, or successfully fooled audiences. The Harvard study of synthetic media found that much of the flagged content was non-political and often humorous, highlighting the difference between synthetic content and malicious disinformation. [Misinformation Review]misinforeview.hks.harvard.eduthe spread of synthetic media on xMisinformation ReviewThe spread of synthetic media on XJun 3, 2024 — Here, results show a sharp rise in the frequency of mentions of AI-g…

Critical readers should therefore distinguish between three separate claims:

  1. A post was AI-generated.
  2. A post was misleading enough to attract Community Notes attention.
  3. A post successfully changed beliefs or caused harm.

Community Notes data provides the strongest evidence for the second claim and some evidence for the first. It is much weaker evidence for the third, which requires additional behavioural or survey research. [PMC]pmc.ncbi.nlm.nih.govmisinformation flags) depended on the political congruence of the fact-checked post…. Research can help to tackle AI-generated disinfo…

Notes Data illustration 3

What Community Notes can and cannot prove

Community Notes is one of the best publicly available sources for studying viral AI misinformation on X because it records real-world cases that users collectively identified as misleading. It enables researchers to measure patterns in visibility, engagement, content type, and correction efforts at a scale that would otherwise be difficult to achieve. [arXiv]arxiv.orgarXiv Characterizing AI-Generated Misinformation on Social MediaCharacterizing AI-Generated Misinformation on Social MediaMay 15, 2025…Published: May 15, 2025

What it can show is that certain forms of AI-generated misinformation are reaching large audiences, attracting user concern, and often achieving substantial engagement before correction efforts occur. It can also reveal recurring characteristics of viral AI content, such as its strong connection to entertainment and visual novelty. [arXiv]arxiv.orgarXiv Characterizing AI-Generated Misinformation on Social MediaCharacterizing AI-Generated Misinformation on Social MediaMay 15, 2025…Published: May 15, 2025

What it cannot show is the full universe of AI-generated falsehoods on X. The dataset reflects what contributors noticed and acted upon, not everything that existed. For that reason, Community Notes should be treated as a powerful but partial source of evidence: a record of visible, user-detected misinformation rather than a complete census of all AI-generated deception on the platform. [X (formerly Twitter)+2arXiv]communitynotes.x.comX (formerly Twitter)Community NotesCommunity Notes aims to create a better-informed world, by empowering people on X to collaboratively a…

Meta replaces fact-checking with X-style community notes

Channel: Associated Press · Views: 1.3K · Uploaded: January 2025 · Length: 1 minute 50 seconds

Open on YouTube

Amazon book picks

Further Reading

Books and field guides related to What Community Notes can and cannot prove. Use these as the next step if you want deeper reading beyond the article.

Browse more on Amazon: Calling Bullshit The Chaos Machine How to Lie with Statistics

As an Amazon Associate I earn from qualifying purchases.

eBay marketplace picks

Marketplace Samples

Example marketplace items related to this page. Use the search link to explore similar finds on eBay.

Using USA
Browse more on eBay.co.uk

Example items shown for inspiration; availability and pricing can change. Branchoria may earn a commission if you purchase through outbound eBay links.

Endnotes

  1. Source: arxiv.org
    Title: arXiv Characterizing AI-Generated Misinformation on Social Media
    Link: https://arxiv.org/abs/2505.10266
    Source snippet

    Characterizing AI-Generated Misinformation on Social MediaMay 15, 2025...

    Published: May 15, 2025

  2. Source: misinforeview.hks.harvard.edu
    Title: the spread of synthetic media on x
    Link: https://misinforeview.hks.harvard.edu/article/the-spread-of-synthetic-media-on-x/
    Source snippet

    Misinformation ReviewThe spread of synthetic media on XJun 3, 2024 — Here, results show a sharp rise in the frequency of mentions of AI-g...

  3. Source: arxiv.org
    Link: https://arxiv.org/abs/2307.07960
    Source snippet

    Did the Roll-Out of Community Notes Reduce Engagement With Misinformation on X/Twitter?July 16, 2023...

    Published: July 16, 2023

  4. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC12478135/
    Source snippet

    notes reduce engagement with and diffusion of...by I Slaughter · 2025 · Cited by 16 — Upon introducing the Community Notes program, X re...

  5. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC11212665/
    Source snippet

    misinformation flags) depended on the political congruence of the fact-checked post.... Research can help to tackle AI-generated disinfo...

  6. Source: harvard.edu
    Link: https://www.harvard.edu/
    Source snippet

    Harvard UniversityHarvard University is devoted to excellence in teaching, learning, and research, and to developing leaders who make a d...

  7. Source: misinforeview.hks.harvard.edu
    Link: https://misinforeview.hks.harvard.edu/
    Source snippet

    Misinformation ReviewMar 3, 2026 — In a pre-registered experiment, we examine how properties of AI-synthesized images influence belief in...

  8. Source: misinforeview.hks.harvard.edu
    Link: https://misinforeview.hks.harvard.edu/article/research-note-examining-how-various-social-media-platforms-have-responded-to-covid-19-misinformation/
    Source snippet

    note: Examining how various social media...15 Dec 2021 — Research note: Examining how various social media platforms have responded to C...

  9. Source: misinforeview.hks.harvard.edu
    Title: volume 5 issue 3
    Link: https://misinforeview.hks.harvard.edu/article/volume-issue/volume-5-issue-3/
    Source snippet

    3This study examines the prevalence and characteristics of synthetic media on social media platform X from December 2022 to September 202...

    Published: December 2022

  10. Source: misinforeview.hks.harvard.edu
    Title: volume 6 issue 1
    Link: https://misinforeview.hks.harvard.edu/article/volume-issue/volume-6-issue-1/
    Source snippet

    Misinformation ReviewIssue 1We surveyed 1,000 U.S. adults to understand concerns about the use of artificial intelligence (AI) during the...

  11. Source: misinforeview.hks.harvard.edu
    Link: https://misinforeview.hks.harvard.edu/article/category/twitter/
    Source snippet

    harvard.eduTwitter/X | HKS Misinformation ReviewThis study examines the prevalence and characteristics of synthetic media on social media...

  12. Source: misinforeview.hks.harvard.edu
    Title: research note
    Link: https://misinforeview.hks.harvard.edu/explore/research-note/
    Source snippet

    Note | HKS Misinformation ReviewThis research note investigates the aftermath of YouTube's global ban on Russian state-affiliated media c...

  13. Source: misinforeview.hks.harvard.edu
    Link: https://misinforeview.hks.harvard.edu/article/misinformation-reloaded-fears-about-the-impact-of-generative-ai-on-misinformation-are-overblown/
    Source snippet

    Fears about the impact of...by FM Simon · 2023 · Cited by 177 — We argue that current concerns about the effects of generative AI on the...

  14. Source: arxiv.org
    Link: https://arxiv.org/pdf/2510.00650
    Source snippet

    Threats to the sustainability of Community Notes on Xby Z Arjmandi-Lari · 2025 · Cited by 3 — The Community Notes system pioneered by Twi...

  15. Source: arxiv.org
    Link: https://arxiv.org/abs/2510.24810
    Source snippet

    A Dataset for Exploring the Helpfulness of Fact-Checking...by R Xing · 2025 · Cited by 2 — Abstract page for arXiv paper 2510.24810: COM...

  16. Source: arxiv.org
    Link: https://arxiv.org/html/2604.02592v2
    Source snippet

    4.1 Dataset. Our analysis dataset comprised 2,946 Community Notes written between November 1, 2025 and January 31, 2026...Read more...

    Published: November 1, 2025

  17. Source: communitynotes.x.com
    Link: https://communitynotes.x.com/guide/en/about/introduction
    Source snippet

    X (formerly Twitter)Community NotesCommunity Notes aims to create a better-informed world, by empowering people on X to collaboratively a...

  18. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC11044002/
    Source snippet

    of X (Formerly Twitter) Community Notes...by MR Allen · 2024 · Cited by 31 — This study evaluated the topics, accuracy, and credibility...

  19. Source: Wikipedia
    Title: Community Notes
    Link: https://en.wikipedia.org/wiki/Community_Notes
    Source snippet

    Community NotesInitially shown to U.S. users only, notes were popularized in March 2022 over misinformation in the Russian invasion of...

    Published: March 2022

  20. Source: socialmediatoday.com
    Title: x formerly twitter tests collaborative ai powered community notes
    Link: https://www.socialmediatoday.com/news/x-formerly-twitter-tests-collaborative-ai-powered-community-notes/811562/
    Source snippet

    X Tests Collaborative AI-Powered Community Notes5 Feb 2026 — AI-generated notes are only created when a user requests one, but they can b...

Additional References

  1. Source: help.x.com
    Link: https://help.x.com/en/using-x/community-notes
    Source snippet

    Help CenterAbout Community Notes on XCommunity Notes aim to create a better informed world by empowering people on X to collaboratively a...

  2. Source: linkedin.com
    Link: https://www.linkedin.com/posts/billymarino_the-spread-of-synthetic-media-on-x-activity-7203713814167437313-IXER
    Source snippet

    The spread of synthetic media on X | Bill MarinoHarvard Kennedy School Misinformation Review just published new work by Giulio Corsi, Wil...

  3. Source: today.ucsd.edu
    Link: https://today.ucsd.edu/story/study-finds-xs-formerly-twitters-community-notes-provide-accurate-credible-answers-to-vaccine-misinformation
    Source snippet

    Finds X's Community Notes Provides Accurate...24 Apr 2024 — A new UC San Diego-led study published in JAMA finds that X's Community Note...

  4. Source: researchgate.net
    Link: https://www.researchgate.net/publication/381151252_The_spread_of_synthetic_media_on_X
    Source snippet

    The spread of synthetic media on XLeveraging crowdsourced annotations identifying synthetic content, our analysis reveals an increase in...

  5. Source: shorensteincenter.org
    Link: https://shorensteincenter.org/research-initiative/the-hks-misinformation-review/

  6. Source: oii.ox.ac.uk
    Link: https://www.oii.ox.ac.uk/news-events/new-study-finds-republicans-flagged-for-posting-misleading-tweets-twice-as-often-as-democrats-on-x-twitters-community-notes/
    Source snippet

    study finds Republicans flagged for posting misleading...Their analysis shows that the Community Notes program identifies twice as many...

  7. Source: Tech Policy Press
    Title: community notes alone wont beat disinformation why factcheckers are essential
    Link: https://www.techpolicy.press/community-notes-alone-wont-beat-disinformation-why-factcheckers-are-essential/
    Source snippet

    Community Notes Alone Won't Beat Disinformation3 Mar 2026 — Community Notes were designed to "democratize" moderation, but research shows...

  8. Source: theguardian.com
    Title: fears ai factcheckers on x could increase promotion of conspiracy theories
    Link: https://www.theguardian.com/technology/2025/jul/02/fears-ai-factcheckers-on-x-could-increase-promotion-of-conspiracy-theories
    Source snippet

    Fears AI factcheckers on X could increase promotion of...2 Jul 2025 — Researchers have found that people perceive human-authored communi...

  9. Source: researchgate.net
    Title: 389176167 Can Community Notes Replace Professional [Fact Checkers]({{ ‘fact-checkers/’ | relative_url }})
    Link: https://www.researchgate.net/publication/389176167_Can_Community_Notes_Replace_Professional_Fact-Checkers
    Source snippet

    Can Community Notes Replace Professional Fact-Checkers?19 Feb 2025 — Two commonly-employed strategies to combat the rise of misinformatio...

  10. Source: propublica.org
    Title: x verified accounts misinformation israel hamas conflict
    Link: https://www.propublica.org/article/x-verified-accounts-misinformation-israel-hamas-conflict
    Source snippet

    Verified Accounts on X Are Thriving As They Spread Israel...Dec 20, 2023 — We found over 1,300 verified accounts that posted misleading...

Topic Tree

Follow this branch

Parent topic

AI Virality Why AI Misinformation Travels So Easily

Related pages 5