Key findings
Persistent and evolving misinformation narratives (Marwick)
Misinformation as sticky stories, not isolated claims. False narratives from 2020 persist but adapt: non-citizen voting claims, election fraud narratives that have become "common knowledge" in Republican circles—election denial is now a marker of Republican legitimacy rather than a fringe claim.
Mainstreaming of fringe narratives. QAnon conspiracy theories (e.g., the claim of a cabal of Democratic pedophiles) were largely confined to fringe online communities in 2016 but have now reached mainstream acceptability—PRRI polling finds ~20% of Americans and 28% of Republicans hold some QAnon beliefs. Similarly, trans-related conspiracy theories and wedge issues have moved from far-right fringe to mainstream political discourse (bathroom bills, drag queen story hour).
The stickiness of false stories. Stories feel true when they align with people's preexisting worldviews about dangers (immigration, liberal elites, transgender people). They resonate more than "dry facts." Once accepted as true within a political community, they create a shared false reality that maps to partisan identity.
Platform and media ecosystem shifts (Marwick, Starbird, Tucker)
The irrelevance of mainstream news media. In 2022, only 38% of Americans kept up with news at all. Mainstream news consumers have become more Democratic, but the least engaged news consumers have moved toward Trump. People now consume political information primarily from podcasts, YouTube, TikTok, Twitter/X, and Facebook—not traditional news outlets. A Bloomberg study found the top nine YouTubers have audiences that are 80% male, feature only 12% women as guests, and frequently discuss voting, platform Republican politicians, argue against political correctness, and reinforce gender binaries and political isolationism—strongly correlated with Trump's unprecedented gains among young men.
A partisan right-wing media infrastructure. Republican funders have cultivated an ecosystem: they identify talent on TikTok, incubate it in platforms like PragerU, amplify it through other influencers, and promote it on platforms (like X/Twitter) that prioritize conservative content. There is no equivalent infrastructure on the left or center-left, making the asymmetry stark.
Influencers as participatory, improvisational performers. Rather than top-down scripted messaging (the Democratic model), right-wing influencers engage in real-time improvisation, responding to audiences and platform events. Audiences shape influencer messaging through feedback loops (e.g., an audience member notes "my ballot doesn't look right" → this becomes the next day's dominant narrative → influencers push it further). This two-way communication gives everyday people agency and ownership of the messaging—something left-leaning audiences have not felt in recent years.
Content moderation rollback. After 2020, when content moderation increased substantially around politics, 2024 saw a pullback. Mark Zuckerberg has stated he is "tired of trying to do the right thing" on moderation. This has led to greater proliferation of misinformation on social media platforms.
Election-specific rumors and infrastructure (Starbird)
The infrastructure-rumors feedback loop. The GOP built an infrastructure (lawsuits, protests, legislative changes) designed to challenge election integrity and benefit their candidates. This infrastructure uses false rumors to motivate lawsuits, protests, and law changes. Each time these pseudo-events occur, they launder rumors into public discourse, making them "feel heavy" and valid, even though many are untrue. These events then generate new rumors—creating a self-reinforcing cycle.
2024 differed from 2020 in partisan response. In 2020, the GOP deployed this infrastructure aggressively post-election. In 2024, even though such infrastructure was ready, Donald Trump won, so it went unused. However, activists on the left began spreading rumors (Starlink rigging, election fraud claims), but these were not picked up by Democratic establishment, media, or political apparatus—no lawsuits, protests, or infrastructure deployed. This is a key asymmetry between how the parties have oriented themselves toward using misinformation strategically.
Platform ownership and bias (Tucker)
Explicit partisan involvement by platform owners. In 2020, platforms attempted to position themselves as nonpartisan. In 2024, this changed dramatically: Donald Trump launched Truth Social; Elon Musk purchased Twitter and became openly involved in the Trump campaign as the "biggest donor" (via engagement amplification). This represents a sea change from implicit platform bias concerns to explicit partisan alignment by platform owners.
Reduced research transparency. In 2020, researchers had unprecedented access to Facebook and Instagram data through the U.S. 2020 Election Study. By 2024, this access largely disappeared. Platforms reduced research partnerships, making it harder for academics to study misinformation at scale. This is a massive loss for understanding what actually happened on platforms during the 2024 election.
Changes in AI and generative content
Generative AI had two levels of impact (Tucker, Marwick)
Level One concern (expected): AI-generated fakes trick people. This was the widely anticipated threat—deepfake videos that would swing elections. However, this largely did not materialize in 2024. The most impactful video in the 2024 election was a real video (Biden's debate performance), not a deepfake.
Level Two concern (observed): The "liar's dividend"—real content is doubted because AI exists. By making people aware that AI can generate convincing fakes, the information environment becomes more uncertain. When Trump speculated that Harris rally images were "probably AI anyway," he weaponized awareness of AI's existence (whether or not the content was actually AI-generated) to cast doubt on real content.
Generative AI's actual election impact: Visual propaganda (Marwick)
Visual propaganda, not deepfakes. The largest use of generative AI in 2024 was not deepfakes or misleading synthetic videos claiming to be real, but visual propaganda—memes and images that reinforced narratives without claiming to depict reality (e.g., images of Trump saving pets from immigrant invaders). These images were often created by AI, reposted by influencers, amplified by Elon Musk and Trump, and spread massively. They were effective not because people believed them to be documentary truth, but because they felt true and reinforced strategic frames.
Cheap, scalable visual propaganda. Generative AI has made it extremely cheap to create visual content at scale. The result is saturation of propaganda that reinforces anti-immigrant, populist frames without any truth claim.
LLMs as unreliable information sources
Gen Z and Gen Alpha use LLMs as search engines. Young people increasingly use ChatGPT and other LLMs as search tools, or encounter AI-generated results in Google searches. Preliminary studies (in-progress replication with Perplexity) suggest that LLMs are not better than Google at helping users identify false claims—in fact, early results suggest LLMs may spread inaccurate political information. This is particularly dangerous if users rely on LLMs for local political information (e.g., local candidate info).
AI safety rollback. The Trump administration has rolled back AI safety and responsible AI initiatives in favor of promoting American hegemonic dominance in AI. This likely means less pressure on AI companies to reduce hallucinations and improve reliability in political information contexts.
Notes
This panel is valuable for capturing a moment of transition in the information ecosystem after the 2024 election. The panelists are leading researchers in misinformation, crisis informatics, and social media political behavior, each bringing complementary perspectives: Marwick's focus on cultural narratives and identity; Starbird's focus on institutional infrastructure and rumor laundering; Tucker's focus on platforms, platform ownership, and research access.
Key contributions:
- Asymmetric infrastructure: The Republican party has built a more sophisticated media ecosystem (influencer networks, participatory messaging, improvisational theater) than Democrats, who rely on traditional broadcast media. This structural asymmetry matters more than individual fact-checking.
- Misinformation is not just false facts: It's about frames, narratives, and selective evidence presentation. Stories that "feel true" spread regardless of veracity. This reframes the problem from "How do we stop false claims?" to "How do we compete on narrative resonance and community belonging?"
- Platform changes matter enormously: The absence of research access, the rollback of content moderation, and the explicit partisan alignment of platform owners (Musk at Twitter/X, Trump's Truth Social) fundamentally alter the information environment in ways we still don't fully understand.
- The hype-threat mismatch on AI: Deepfakes were hyped as the existential threat to the 2024 election, but visual propaganda and the "liar's dividend" proved more consequential. This is an important correction to the AI-doomerism narrative.
The panel represents the state of knowledge among leading researchers several weeks after the 2024 election, and captures both what we know and what we're uncertain about regarding the role of misinformation in electoral outcomes.