# How Information Superspreaders Shape What Goes Viral

An information superspreader is a highly connected or hyperactive entity—whether a human influencer, a political pundit, or an automated bot—that disproportionately drives the dissemination of specific content to massive audiences within a social network. By leveraging the underlying architecture of digital networks and exploiting algorithmic recommendation systems, these entities act as the critical ignition points that transform isolated pieces of information, memes, or disinformation into global viral cascades.

## The Architecture of the Modern Attention Economy

To understand how a single post can capture the global imagination or derail a democratic election, it is essential to look at the sheer scale of the modern information ecosystem. As of early 2024, the digital landscape hosts a staggering 5.35 to 5.61 billion internet and mobile users, meaning that more than 62% to 69% of the global population is active online [cite: 1, 2]. The average internet user engages with these platforms for roughly two hours and twenty minutes daily, navigating across an average of 6.7 different social networking applications each month [cite: 2, 3]. The advertising machinery supporting this ecosystem is similarly colossal, with social media ad spending projected to reach nearly $220 billion in 2024 [cite: 3].

Despite this sprawling, decentralized web of billions of interconnected individuals, information does not flow evenly. The digital landscape is characterized by severe asymmetry. A microscopic fraction of users wields a disproportionate amount of influence over what the rest of the world sees, reads, and ultimately believes. 

Studies into digital misinformation have consistently highlighted this immense imbalance. During the 2016 United States Presidential Election, researchers discovered an extreme power-law distribution in information sharing: a mere 0.1% of Twitter users were responsible for circulating approximately 80% of all fake news on the platform [cite: 4, 5]. This defining characteristic of a superspreader network demonstrates that the vast majority of problematic content is driven by a highly concentrated cluster of actors. Similar patterns of concentrated dissemination were observed throughout the 2020 election cycle, as well as during the height of the COVID-19 pandemic [cite: 4, 6]. These superspreaders are not necessarily household names or traditional celebrities. They frequently include hyper-partisan political commentators, coordinated automated bot networks, and highly active individual citizens who act as central hubs in the daily flow of data [cite: 4, 5].

Understanding how these entities operate requires moving beyond the colloquial concept of "going viral." It requires a rigorous examination of network science, the mathematics of epidemiology, the psychology of complex contagion, and the opaque recommendation algorithms that dictate modern digital visibility.

## The Epidemiology of Information: Why Ideas Are Not Viruses

For decades, theoretical models of information diffusion have been framed using analogies to the contagion models of infectious diseases [cite: 7]. When a piece of media spreads rapidly, the intuitive assumption is that it moves from person to person exactly like a biological pathogen. However, network scientists and computational sociologists have increasingly realized that classic epidemiological models have severe limitations when applied to human behavior and digital social media [cite: 8, 9, 10].

### The Limitations of the SIR Model

The most famous of these biological frameworks is the SIR model, first introduced in the early twentieth century by researchers Kermack and McKendrick to describe the spread of pathogens like cholera or influenza [cite: 11]. The model divides a population into three primary compartments: Susceptible (S), Infected (I), and Recovered or Removed (R) [cite: 8, 11]. 

In this mathematical framework, Susceptible individuals are those who have not yet contracted the disease but remain at risk. Infected individuals actively carry the pathogen and are capable of spreading it to others. Recovered individuals have either survived and gained permanent immunity, or they have died, effectively removing them from the transmission chain entirely [cite: 8, 11]. 

The standard SIR model relies on a linear, fixed force of infection. It assumes that every person in a population is mixing freely and has an equal, random chance of coming into contact with any other person, completely ignoring the spatial boundaries and social distances between different groups [cite: 9, 12]. It also operates on a strict binary logic: when a Susceptible person comes into physical contact with an Infected person, there is only one variable—whether the pathogen successfully jumps hosts [cite: 8]. 

When applied to a political rumor on a microblogging site or a trending video on a short-form video platform, this biological assumption rapidly breaks down. Ideas are not biological pathogens, and human beings are not passive cellular hosts. When a digital user is exposed to a piece of misinformation, they process it through complex pre-existing beliefs, political biases, and varying levels of skepticism [cite: 8]. To address this reality, modern network researchers have developed modified mathematical frameworks, such as the SEIZ (Susceptible, Exposed, Infected, Skeptic) model. Unlike the rigid SIR model, the SEIZ framework introduces an "Exposed" compartment for users who require time to think before reacting to a post, and a "Skeptic" compartment for those who view the information but actively choose not to adopt, believe, or share it [cite: 8].

### Echo Chambers and Structural Trapping

Another critical flaw in using standard biological models for digital virality is the failure to account for what sociologists call "community structure" [cite: 9, 10]. Social networks are highly clustered environments. People naturally tend to follow, friend, and interact with others who share their political ideologies, geographic locations, or personal interests, forming dense digital neighborhoods. 

In a standard SIR epidemic model, an outbreak inside a densely populated cluster rapidly explodes outward into the broader population. But in a human social network, high clustering can actually cause a phenomenon known as "crowding" or the "protection effect" [cite: 9]. If a piece of information enters a highly insular community, it may bounce around intensely within that specific echo chamber but fail to cross over into neighboring, distinct communities—a dynamic network scientists refer to as "structural trapping" [cite: 10, 13]. A biological virus does not care about a host's political ideology or media preferences, but a highly partisan meme certainly does. 

## Simple vs. Complex Contagion: Why Ideas Need Reinforcement

To explain why some digital ideas overcome structural trapping while others fizzle out in isolation, sociologists Damon Centola and Michael Macy pioneered a vital distinction between two fundamentally different types of spread: simple contagion and complex contagion [cite: 14, 15].

[image delta #1, 0 bytes]



### The Mechanics of Simple Contagion

A simple contagion requires only a single exposure for successful transmission [cite: 14, 16]. Biological diseases are the quintessential simple contagions; sitting next to one person carrying a highly contagious strain of the flu is sufficient to contract the illness. In the digital realm, a universally appealing piece of breaking news, a major sports score, or a simple, uncontroversial fact acts as a simple contagion [cite: 10]. 

Simple contagions thrive in random networks connected by "weak ties" [cite: 14]. If an individual has a massive, loose network of distant acquaintances spanning many different social circles, a simple contagion will rapidly jump across these weak ties. The weak ties act as efficient shortcuts, allowing the information to bypass local clusters and infect remote regions of the global network almost instantly [cite: 14, 17].



### The Mechanics of Complex Contagion

A complex contagion, however, requires multiple sources of exposure and social reinforcement before an individual actually adopts the behavior or shares the information [cite: 14, 16]. Adopting a new political ideology, participating in a high-risk social movement, migrating to a new social media platform, or sharing a highly controversial conspiracy theory are all forms of complex contagion [cite: 14, 15, 18]. 

If a digital user sees only one person in their network sharing a radical political opinion, they are highly unlikely to adopt it themselves. However, if they witness three, four, or five trusted friends or prominent voices sharing that exact same opinion, the cumulative social reinforcement eventually triggers adoption [cite: 14, 15]. Because of this inherent need for validation, complex contagions spread completely differently than simple contagions. They struggle to cross loose "weak ties" because a single connection to an outside group is mathematically insufficient to convince that group to adopt the behavior. Instead, complex contagions thrive in highly clustered, tightly knit communities where dense, redundant social ties provide the necessary peer pressure and echo-chamber validation to push individuals past their resistance threshold [cite: 17, 18].

### Predicting Virality Through Community Structure

Researchers have discovered that closely observing how a piece of content interacts with network communities early in its lifespan can accurately predict its eventual virality. By calculating the "entropy" of a meme's spread—measuring the proportion of its initial adoption across different distinct clusters—analysts can determine if it is acting as a simple or complex contagion [cite: 10, 13]. 

The vast majority of internet memes and political behaviors behave as complex contagions, requiring heavy, redundant social reinforcement and often remaining confined to specific demographic or ideological clusters. However, the most universally successful, hyper-viral memes transcend this limitation entirely. They manage to escape their structural traps and permeate multiple diverse communities simultaneously, effectively spreading across the network exactly like a highly infectious disease [cite: 10, 13].

## Structural Virality: Broadcasts vs. Network Cascades

When content reaches tens of millions of people, it is colloquially said to have "gone viral." But network scientists draw a sharp, mathematical distinction between simple popularity and actual virality. To rigorously quantify this difference, researchers introduced the concept of "structural virality," a metric that calculates the average pairwise distance between all nodes in a given diffusion tree [cite: 19, 20]. 

Structural virality interpolates between two conceptual extremes of information diffusion: broadcast and cascade [cite: 19].

### The Anatomy of a Broadcast

Broadcast diffusion represents the lowest end of the structural virality spectrum. In this scenario, a massive central hub—such as a major traditional news organization, a global celebrity, or an established institutional account—posts a message to millions of direct followers. Millions of people see the content and share it directly from the original source. 

If drawn as a network graph, this diffusion pattern looks like a starburst: a single, massive central node surrounded by millions of shallow, one-step branches [cite: 19, 21]. Even though the content is immensely popular and widely viewed, its structural virality score is practically zero because almost no person-to-person contagion actually occurred. The reach of the message was dictated entirely by the pre-existing size of the broadcaster's audience [cite: 19, 22].

### The Anatomy of a Deep Cascade

Viral cascade diffusion represents the opposite extreme. In this scenario, an obscure or average user posts a message. One friend shares it. A friend of that friend shares it next. This localized chain reaction continues across dozens or even hundreds of distinct generations. 

The resulting network graph looks like a deep, sprawling root system. In a true cascade, no single individual is directly responsible for a large fraction of the total adoption [cite: 7, 19, 21]. Information passes hop-by-hop through interpersonal connections.

| Diffusion Feature | Broadcast Model | Viral Cascade Model |
| :--- | :--- | :--- |
| **Origin Point** | Single, highly connected central hub (e.g., major media outlet, celebrity). | Often obscure or average users, relying entirely on peer-to-peer transmission. |
| **Cascade Depth** | Extremely shallow (usually 1 step directly from the source). | Deep (multi-generational sharing chains). |
| **Structural Virality Score** | Low. Popularity is driven entirely by the size of the initial broadcast audience. | High. Driven by social contagion and generational branching processes. |
| **Cross-Ideological Reach** | Low. Followers of the central hub usually share the same broad interests or ideology. | High. Deep cascades are significantly more likely to cross rigid ideological lines [cite: 7, 23]. |
| **Network Graph Shape** | Star graph (hub-and-spoke). | Complex tree or branching root system. |

In a landmark computational analysis of one billion diffusion events on the platform formerly known as Twitter, researchers found a surprisingly low correlation between the ultimate size of an event and its structural virality [cite: 19, 22]. Simply put, knowing that a video or news story was seen by ten million people tells an analyst almost nothing about how it actually spread. 

The data revealed that most massively popular online events are driven by large broadcasts, not viral cascades. However, when true viral cascades do occur, they produce a unique sociological effect. Deep cascade trees are significantly more likely to cross ideological lines, exposing users to political diversity and differing viewpoints that broadcast media rarely achieves [cite: 7, 23]. 

## Algorithmic Amplification: When the Platform is the Superspreader

In the early days of social media, information spread organically based on a user's chosen social graph—individuals saw what their friends and the specific people they followed chose to post. Today, the platforms themselves act as automated, systemic superspreaders through a process known as algorithmic amplification [cite: 24]. 

Algorithmic amplification is the mechanism by which automated ranking systems push content far beyond a user's organic social graph, displaying it to out-of-network users who did not request it [cite: 24]. Rather than relying on simple chronological feeds, major platforms like TikTok, YouTube, Facebook, and X deploy proprietary algorithms to evaluate content based on high-frequency engagement signals. These signals include watch time, likes, replies, profile clicks, and shares [cite: 24]. Content that triggers strong engagement on these metrics is systematically promoted to progressively larger audiences.

### The Engagement Trap and Affective Polarization

The core, overriding directive of these recommendation systems is to maximize user retention and continuous attention, which underpins the digital advertising business model [cite: 24]. Consequently, algorithms heavily favor and amplify content that evokes strong emotional reactions, regardless of its factual accuracy or social benefit.

A comprehensive 2025 algorithmic audit of X during the lead-up to the 2024 U.S. Presidential Election revealed the profound societal effects of this design architecture. Academic researchers deployed 120 automated "sock-puppet" monitoring accounts to rigorously compare X's algorithmic "For You" feed against a reverse-chronological baseline feed [cite: 24, 25]. The audit found that the algorithmic feed disproportionately amplified content that was emotionally charged, highly partisan, and explicitly hostile toward members of opposing political groups [cite: 24, 25]. Furthermore, a parallel study focusing on the German political spectrum found that X's algorithm vastly overrepresented populist extremes. Posts from far-right and far-left politicians appeared in the "For You" feed at rates massively disproportionate to their actual tweet volume, overshadowing moderate voices [cite: 26].

The psychological impact of this systemic amplification is severe and rapid. In a groundbreaking field trial involving over 1,200 human participants, researchers subtly tweaked users' X feeds. When users were exposed to an algorithmic feed that slightly boosted anti-democratic attitudes and partisan animosity, their feelings of "affective polarization"—the visceral dislike, distrust, and anger directed at opposing political factions—spiked dramatically [cite: 25, 27]. 

The research indicated that just one week of exposure to the amplified hostile feed resulted in an increase in political polarization equivalent to three years of historical, baseline societal shifts [cite: 25]. Interestingly, when the research team utilized a custom browser extension to dynamically downgrade hostile and anti-democratic posts, users' attitudes toward the opposing party became measurably more positive [cite: 27]. This demonstrates that platform algorithms are not neutral mirrors of society; they hold direct, real-time influence over social trust and democratic discourse [cite: 27]. 

## Cultural Cascades: Meme Mechanics on TikTok

The power of algorithmic amplification is not limited to political polarization; it fundamentally shapes consumer culture, entertainment, and commercial success. Because recommendation systems like TikTok's "For You" page excel at identifying and targeting users with highly specific shared traits—creating dense, homogeneous network clusters virtually overnight—they are incredibly efficient at facilitating complex contagions.

### The #GentleMinions Phenomenon

In July 2022, a massive viral trend emerged globally around the release of the animated film *Minions: The Rise of Gru*. Sparked by an 18-year-old user and his friends in Sydney, Australia, the trend involved large groups of teenage boys attending the film dressed in formal business suits, steepling their fingers to mimic the film's primary villain [cite: 28, 29]. The hashtag #GentleMinions quickly amassed over 66 million views, while related franchise hashtags generated billions of total impressions across the platform [cite: 28].

Sociologically, this was a textbook complex contagion. It relied heavily on shared generational nostalgia—specifically Gen Z users who had grown up watching the original 2010 film in early childhood—combined with a layer of absurdist, ironic humor that resonated deeply with that demographic [cite: 28, 30]. TikTok's algorithm quickly recognized the high engagement metrics specifically among teenage boys and rapidly amplified the content across homogeneous networks of similar youth globally, reinforcing the behavior [cite: 28, 29]. 

The real-world commercial impact was staggering. The organic, algorithmic trend helped drive the film to a record-breaking $125 million opening weekend in the United States [cite: 28, 31]. Demographic data showed that 34% of the opening weekend audience was aged 13–17, a massive demographic shift compared to just 8% for previous installments in the same franchise [cite: 28].

### The Grimace Shake Horror Trend

A year later, the fast-food giant McDonald's experienced a nearly identical algorithmic windfall. To celebrate the 52nd birthday of their legacy mascot Grimace, the company released a limited-edition purple berry-flavored milkshake. Users on TikTok spontaneously began creating cinematic, absurdist horror short films in which drinking the shake resulted in demonic possession, fainting, or gruesome death [cite: 32, 33]. 

Despite the morbid and violent premise of the videos—which traditional corporate brand marketers would typically avoid at all costs—the trend acted as a massive algorithmic amplifier for the product. Search interest for the shake skyrocketed in the summer of 2025, and consumer tracking data showed that nearly one in three McDonald's customers cited the Grimace Shake as their favorite flavor, outperforming traditional staples like chocolate and vanilla [cite: 34, 35]. The viral trend successfully engaged Millennials and Gen Z, increasing their overall share of wallet at the restaurant, attracting consumers who hadn't visited the chain in months, and ultimately contributing to a 14% rise in global revenue for the quarter [cite: 34, 35]. Both the Gentleminions and Grimace Shake trends highlight how proprietary algorithms act as non-human superspreaders, capable of turning niche, ironic inside jokes into global commercial events [cite: 28, 34].

## The AI Threat: Social Bots and Information Warfare

While human superspreaders and platform algorithms wield immense influence, the digital ecosystem is increasingly populated by fully automated entities. Social bots—automated accounts programmed to mimic human behavior—have long been a fixture of digital platforms, driving engagement through excessive posting, rapid retweeting of emerging narratives, and algorithmic gaming [cite: 24, 36]. 

### From Brute Force Bots to AI Sleeper Agents

Historically, bots operated via brute force. They functioned by amplifying existing, human-created content to hijack trending algorithms [cite: 36]. During the early stages of the COVID-19 pandemic, bot networks were found to be hyper-social, acting within seconds of a low-credibility article being published to ensure it achieved viral momentum [cite: 36]. A longitudinal study analyzing 5.8 million tweets regarding the COVID-19 vaccine discourse found a fascinating interplay between human and bot superspreaders. The data showed that bots reached their highest appeal and scope during the "Pre-Vaccine" period, effectively pre-seeding the digital landscape with doubt, conspiracy, and misinformation [cite: 6]. However, during the actual week of the vaccine's launch, human-generated misinformation tweets took over, displaying significantly higher appeal and reach than their automated counterparts [cite: 6]. This suggests a symbiotic relationship: bots lay the groundwork and game the algorithm, but human superspreaders are ultimately required to provide the emotional resonance and peer validation needed for a true complex contagion to take hold in the broader population [cite: 6, 14].

The advent of Large Language Models (LLMs) and Generative AI (AIGC) has fundamentally altered the capabilities of these bot networks. Previously, automated accounts were often identifiable by their rigid, repetitive text patterns, stolen profile pictures, and lack of conversational nuance. Today, generative AI can produce highly persuasive, contextually aware, human-like prose, imagery, and audio at a massive, nearly infinite scale [cite: 4, 37].

### Misinformation in the 2024 Election Cycle

Researchers monitoring the 2024 U.S. Presidential Election ran extensive tests to gauge the specific threat of these next-generation bots. In an enclosed experiment mimicking a live social media platform, university researchers deployed "sleeper social bots" powered by the GPT-4 language model. These bots were given detailed background personas and tasked with aggressively arguing for a fictional political ballot measure regarding age restrictions on social media [cite: 38]. 

The experimental results were deeply alarming: the AI bots consistently passed as human to observers. They were uniquely capable of reframing falsehoods convincingly, defending their political viewpoints in deep, multi-reply arguments, and strategically redirecting off-topic conversations back to their core disinformation narratives [cite: 38]. They did not just broadcast; they debated.

While actual deployment of AI-generated misinformation during the 2024 elections—such as fake audio of political candidates, deceptive robocalls, or deepfake imagery—definitely polluted the digital waters, assessing its exact impact on raw voting behavior remains difficult [cite: 37, 39]. Independent intelligence audits found no conclusive, empirical evidence that AI-generated deepfakes measurably flipped a major election result in the US, UK, or Europe [cite: 39]. 

However, public perception tells a different story. Surveys showed that nearly 80% of the US public expressed some level of worry about AI's role in election misinformation, with concern notably higher among older demographics, the highly educated, and those working in STEM fields [cite: 40]. Instead of directly flipping votes, the primary, measurable effect of AI superspreaders in 2024 was to amplify existing polarization, entrench prior beliefs, and cast a generalized shadow of doubt over the authenticity of all digital media [cite: 37, 39]. This phenomenon is known as the "liar's dividend"—when the public knows AI fakes exist, bad actors can easily dismiss genuine, damning evidence as artificially generated, severely undermining objective truth [cite: 37].

## The Dark Web of Misinformation: Encrypted Networks

Not all superspreader activity happens in the public, observable squares of X or TikTok. In many parts of the Global South, particularly in massive democracies like India and Brazil, WhatsApp is the dominant digital communication platform. Because WhatsApp is end-to-end encrypted, researchers, journalists, and government regulators cannot map the networks or see the viral cascades forming in real-time [cite: 41, 42]. 

### WhatsApp Tiplines in India and Brazil

To combat this opaque spread, independent fact-checking organizations have established WhatsApp "tiplines" where everyday citizens can voluntarily forward suspicious messages for expert verification [cite: 41, 43]. Analysis of these tiplines provides a rare window into encrypted superspreader networks. 

During the 2022 Brazilian general elections, the national election authority operated a fact-checking chatbot that received over 223,000 queries in just two and a half months [cite: 43]. Similarly, during the devastating second wave of COVID-19 in India, tiplines were flooded with hyper-local misinformation. Researchers found that misinformation on these encrypted apps is intensely varied, frequently shifting regional languages, and relying heavily on visual metaphors, audio clips, and manipulated images rather than just text [cite: 41, 42]. Because WhatsApp group chats often consist of close friends and family, the information carries a high degree of implied trust, acting as a potent accelerator for complex contagion [cite: 44].

### The Difficulty of De-Radicalization

Curing a network once misinformation has spread is notoriously difficult. Field experiments conducted in Brazil tested this by deactivating heavy users' access to WhatsApp during a highly polarized election period [cite: 45]. The experiment successfully and significantly reduced the users' overall exposure to false news [cite: 45]. 

However, the researchers found that this forced reduction in exposure did not change the users' underlying belief in the false news they had already absorbed prior to deactivation [cite: 45]. Furthermore, the intervention showed no measurable changes in the users' levels of political polarization [cite: 45]. This underscores a grim reality of network science: preventing the initial spread of a complex contagion is possible, but curing it once the ideology has taken root in a user's mind is a monumental challenge.

## Cognitive Defenses: How to Slow a Superspreader

With algorithms explicitly prioritizing outrage, and AI bots capable of generating infinite streams of plausible falsehoods, how can the network defend itself? Because structural interventions face immense legal and technical hurdles, researchers are focusing on disrupting the contagion mechanics at the individual user level through cognitive friction. 

### The Power of Accuracy Nudges

A massive adversarial collaboration involving prominent psychology research teams from opposing viewpoints analyzed 70 different statistical models across 21 experiments involving over 27,000 participants [cite: 46]. They sought to test a simple behavioral intervention: an "accuracy nudge." 

An accuracy nudge does not paternalistically tell a user what is true or false, nor does it fact-check a specific claim. It simply asks the user to pause and rate the accuracy of a single, politically neutral headline before they continue scrolling through their feed [cite: 46, 47]. This tiny moment of friction operates on dual-process psychological theory. It forces the human brain to shift out of fast, emotional, heuristic processing (which algorithms actively exploit) and into slow, deliberate, analytical processing [cite: 47, 48]. 

The empirical results were remarkable. The simple accuracy nudge significantly decreased the intention to share false information across the entire political spectrum. The intervention successfully reduced the sharing of falsehoods by 3.3% to 14.6% among Republicans, and by 7.6% to 19.1% among Democrats [cite: 46]. Follow-up studies demonstrated that these nudges are particularly effective in environments where the overall volume of misinformation is relatively low, providing ecologically valid support for their use on major platforms [cite: 49, 50]. This proves that while humans are highly susceptible to algorithmic manipulation, a simple contextual reminder of the concept of "truth" can severely dampen the reproductive rate of a viral falsehood [cite: 46, 51].

Conversely, studies have shown that directly labeling content as "AI-generated" or attaching explicit warning labels to specific posts can sometimes backfire. Warning labels can inadvertently increase the likelihood that unlabeled misinformation on the same feed will be perceived as highly accurate—a phenomenon known as the "implied truth effect" [cite: 51, 52]. Furthermore, depending on how AI is framed, users may either distrust it entirely or assume it is highly objective, complicating the efficacy of transparency labels [cite: 53].

### Upgrading Digital Literacy: The SIFT Method

When users do engage their analytical thinking, they need the right tools to evaluate the information in front of them. For years, educational institutions taught digital literacy using the CRAAP test (Currency, Relevance, Authority, Accuracy, Purpose), a framework that asks users to deeply analyze a source's own website and "About" page [cite: 54]. 

However, modern studies show that evaluating a sophisticated bad actor based on their own self-published website is highly ineffective. Professional fact-checkers use a different, much faster strategy, which has now been codified for students and the general public as the SIFT method [cite: 54, 55]. The framework consists of four distinct steps:

First, users must **Stop**. Before reacting, sharing, or engaging, they must pause and check their emotional reaction to the headline [cite: 54, 56]. Second, they must **Investigate the source**. Crucially, this means they do not read the site's own bio. Third, they must **Find better coverage**. They look to see if reputable, established media outlets are reporting the exact same claim [cite: 54, 56]. Finally, they must **Trace claims, quotes, and media** back to their original context to ensure they haven't been manipulated or deceptively edited [cite: 55, 56].

The core mechanism underpinning the SIFT method is known as **lateral reading** [cite: 54, 55]. Instead of drilling down vertically into an unfamiliar or suspicious website, the user immediately opens a new browser tab and searches what other, independent authorities say about that specific website [cite: 55]. By stepping outside the immediate context and framing of the superspreader's post, users leverage the broader consensus of the web to evaluate credibility in a matter of seconds, bypassing the sophisticated design traps of the misinformation source entirely [cite: 55, 56]. 

## Bottom line
An information superspreader is a highly connected or hyperactive node—whether human or automated—that leverages platform recommendation algorithms and network topologies to drive the vast majority of digital virality. While true multi-generational viral cascades are statistically rare compared to massive broadcasts, they are highly effective at crossing ideological boundaries and driving complex contagions, such as political polarization and cultural meme trends. While the exact impact of new generative AI bots on real-world voting behavior remains difficult to quantify, we know that disrupting their spread requires systemic algorithmic transparency and user-level interventions—like accuracy nudges and lateral reading—that force human users to pause before they share.

## Sources
1. [limitations of epidemiological SIR models for social media virality](https://arxiv.org/pdf/2103.12191)
2. [limitations of epidemiological SIR models for social media virality](https://www.quora.com/What-are-the-limitations-of-the-epidemic-SIR-model)
3. [limitations of epidemiological SIR models for social media virality](https://pmc.ncbi.nlm.nih.gov/articles/PMC5051690/)
4. [limitations of epidemiological SIR models for social media virality](https://pmc.ncbi.nlm.nih.gov/articles/PMC9487859/)
5. [limitations of epidemiological SIR models for social media virality](https://blogs.cornell.edu/info2040/2018/11/29/sir-model-epidemics-and-social-media/)
6. [simple vs complex contagion information spread network science peer-reviewed](https://pubmed.ncbi.nlm.nih.gov/37390429/)
7. [simple vs complex contagion information spread network science peer-reviewed](https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1012206)
8. [simple vs complex contagion information spread network science peer-reviewed](https://sociopatterns.org/assets/papers/PhysRevLett.130.247401.pdf)
9. [simple vs complex contagion information spread network science peer-reviewed](https://pmc.ncbi.nlm.nih.gov/articles/PMC12168247/)
10. [simple vs complex contagion information spread network science peer-reviewed](https://www.researchgate.net/publication/371600657_Distinguishing_Simple_and_Complex_Contagion_Processes_on_Networks)
11. [AI-powered bot networks misinformation superspreaders 2024 research paper metrics](https://arxiv.org/html/2502.11248v1)
12. [AI-powered bot networks misinformation superspreaders 2024 research paper metrics](https://www.researchgate.net/publication/391531072_Appeal_and_Scope_of_Misinformation_Spread_by_AI_Agents_and_Humans)
13. [AI-powered bot networks misinformation superspreaders 2024 research paper metrics](https://pmc.ncbi.nlm.nih.gov/articles/PMC8139392/)
14. [AI-powered bot networks misinformation superspreaders 2024 research paper metrics](https://arxiv.org/pdf/2502.11248)
15. [AI-powered bot networks misinformation superspreaders 2024 research paper metrics](https://journalqd.org/article/view/5848)
16. [quantitative reach of information superspreaders vs average users study 2023 2024](https://datareportal.com/reports/digital-2024-global-overview-report)
17. [quantitative reach of information superspreaders vs average users study 2023 2024](https://www.smartinsights.com/social-media-marketing/social-media-strategy/new-global-social-media-research/)
18. [quantitative reach of information superspreaders vs average users study 2023 2024](https://www.pewresearch.org/internet/fact-sheet/social-media/)
19. [quantitative reach of information superspreaders vs average users study 2023 2024](https://www.sprinklr.com/blog/social-media-marketing-statistics/)
20. [quantitative reach of information superspreaders vs average users study 2023 2024](https://electroiq.com/stats/social-media-marketing-statistics/)
21. [structural virality case study example viral meme or news story data](https://cs.stanford.edu/people/ashton/pubs/twiral.pdf)
22. [structural virality case study example viral meme or news story data](https://pubsonline.informs.org/doi/10.1287/mnsc.2015.2158)
23. [structural virality case study example viral meme or news story data](https://www.researchgate.net/publication/282062131_The_Structural_Virality_of_Online_Diffusion)
24. [structural virality case study example viral meme or news story data](https://pmc.ncbi.nlm.nih.gov/articles/PMC3755286/)
25. [structural virality case study example viral meme or news story data](https://www.researchgate.net/publication/256188948_Virality_Prediction_and_Community_Structure_in_Social_Networks)
26. [mechanics of viral memes 2024 2025 TikTok research paper](https://www.researchgate.net/publication/395636551_When_memes_govern_A_study_of_TikTok_video_memes_in_the_2024_US_presidential_election)
27. [mechanics of viral memes 2024 2025 TikTok research paper](https://journal.dinamikapublika.id/index.php/JUMDER/article/download/28/26)
28. [mechanics of viral memes 2024 2025 TikTok research paper](https://www.researchgate.net/publication/391495333_TikTok_Meme_Analysis_Gestures_Sounds_and_Metadata)
29. [mechanics of viral memes 2024 2025 TikTok research paper](https://arxiv.org/abs/2402.13855)
30. [mechanics of viral memes 2024 2025 TikTok research paper](https://gdfm.me/assets/media/2024/05/what-we-can-learn-from-tiktok-through-its-research-api.pdf)
31. [time in Brazil](https://www.google.com/search?q=time+in+Brazil)
32. [time in India](https://www.google.com/search?q=time+in+India)
33. [broadcast virality vs network cascade virality table data comparison](https://arxiv.org/pdf/2103.04663)
34. [broadcast virality vs network cascade virality table data comparison](https://www.com.cuhk.edu.hk/wp-content/uploads/2022/11/Broadcast-Versus-Viral-Spreading.pdf)
35. [broadcast virality vs network cascade virality table data comparison](https://www.researchgate.net/publication/324551067_Broadcast_Versus_Viral_Spreading_The_Structure_of_Diffusion_Cascades_and_Selective_Sharing_on_Social_Media)
36. [broadcast virality vs network cascade virality table data comparison](https://arxiv.org/abs/2006.01027)
37. [broadcast virality vs network cascade virality table data comparison](https://www.researchgate.net/publication/341816527_Go_viral_or_go_broadcast_Characterizing_the_virality_and_growth_of_cascades)
38. [how to recognize artificially amplified information practical tips 2024 2025 research](https://en.wikipedia.org/wiki/Algorithmic_amplification)
39. [how to recognize artificially amplified information practical tips 2024 2025 research](https://www.tandfonline.com/doi/full/10.1080/07421222.2025.2561381)
40. [how to recognize artificially amplified information practical tips 2024 2025 research](https://www.researchgate.net/publication/397767926_Artificial_Intelligence_Recommendations_Amplify_the_Sharing_of_True_and_Fake_News_on_Social_Media_by_Appealing_to_Fast_Cognition)
41. [how to recognize artificially amplified information practical tips 2024 2025 research](https://www.emc-lab.org/uploads/1/1/3/6/113627673/ecker.jarmaceditorial.2025.pdf)
42. [how to recognize artificially amplified information practical tips 2024 2025 research](https://formative.jmir.org/2024/1/e60024)
43. [WhatsApp misinformation tipline India Brazil case study statistics 2023 2024](https://meedan.org/post/what-spreads-on-whatsapp-a-snapshot-of-tipline-content-from-india-and-brazil)
44. [WhatsApp misinformation tipline India Brazil case study statistics 2023 2024](https://academic.oup.com/ijpor/article/36/3/edae032/7709027)
45. [WhatsApp misinformation tipline India Brazil case study statistics 2023 2024](https://arxiv.org/pdf/2507.16298?)
46. [WhatsApp misinformation tipline India Brazil case study statistics 2023 2024](https://gvrkiran.github.io/content/WhatsApp_community_factchecking____ICWSM_May_2023.pdf)
47. [WhatsApp misinformation tipline India Brazil case study statistics 2023 2024](https://voxdev.org/topic/institutions-political-economy/reducing-exposure-misinformation-evidence-whatsapp-brazil)
48. [Gentleminions TikTok viral cascade structural analysis](https://www.accio.com/business/gentleminions_tiktok_trend)
49. [Gentleminions TikTok viral cascade structural analysis](https://reporter.anu.edu.au/all-stories/the-viral-magic-behind-the-gentleminions-tiktok-trend)
50. [Gentleminions TikTok viral cascade structural analysis](https://www.reddit.com/r/HobbyDrama/comments/wf5xh1/movies_gentleminions_how_a_viral_tiktok_trend_led/)
51. [Gentleminions TikTok viral cascade structural analysis](https://www.highsnobiety.com/p/gentleminions-tiktok-trend-minions-meme/)
52. [Gentleminions TikTok viral cascade structural analysis](https://www.insidehook.com/internet/gentle-minions-tiktok-explained)
53. [accuracy nudge 2024 research effectiveness misinformation spread](https://news.cornell.edu/stories/2024/04/accuracy-nudges-decrease-misinformation-sharing-left-right)
54. [accuracy nudge 2024 research effectiveness misinformation spread](https://pubmed.ncbi.nlm.nih.gov/38769362/)
55. [accuracy nudge 2024 research effectiveness misinformation spread](https://pmc.ncbi.nlm.nih.gov/articles/PMC11106285/)
56. [accuracy nudge 2024 research effectiveness misinformation spread](https://www.tandfonline.com/doi/full/10.1080/10410236.2025.2507676)
57. [accuracy nudge 2024 research effectiveness misinformation spread](https://academic.oup.com/jcmc/article/30/4/zmaf009/8173297)
58. [Grimace Shake trend network science research data](https://civicscience.com/the-horrifying-grimace-shake-trend-boosts-interest-plus-5-unexpected-customer-insights/)
59. [Grimace Shake trend network science research data](https://www.accio.com/business/grimace-shake-trend)
60. [Grimace Shake trend network science research data](https://www.accio.com/business/mcdonalds_grimace_shake_trend)
61. [Grimace Shake trend network science research data](https://www.numerator.com/resources/blog/grimace-effect-mcdonalds-growth/)
62. [Grimace Shake trend network science research data](https://yougov.com/en-us/articles/46901-us-how-has-the-grimace-shake-tiktok-challenge-impacted-mcdonalds)
63. [SIFT method vs lateral reading 2024 research efficacy](https://files.eric.ed.gov/fulltext/EJ1453416.pdf)
64. [SIFT method vs lateral reading 2024 research efficacy](https://www.cip.uw.edu/2021/12/07/lateral-reading-canada-civix-study/)
65. [SIFT method vs lateral reading 2024 research efficacy](https://www.researchgate.net/publication/375081144_Using_the_SIFT_strategy_to_enhance_the_Lateral_Reading_skills_of_undergraduate_students_for_detecting_digital_misinformation)
66. [SIFT method vs lateral reading 2024 research efficacy](https://www.youtube.com/watch?v=qsrXJGpxwIE)
67. [SIFT method vs lateral reading 2024 research efficacy](https://www.researchgate.net/publication/353929358_Associations_Between_Online_Instruction_in_Lateral_Reading_Strategies_and_Fact-Checking_COVID-19_News_Among_College_Students)
68. [AI bot networks 2024 US election misinformation statistics research](https://polymathic.usc.edu/ahmanson-lab/bots-vs-ballots-generative-ai-disinformation-and-2024-election)
69. [AI bot networks 2024 US election misinformation statistics research](https://cetas.turing.ac.uk/publications/ai-enabled-influence-operations-safeguarding-future-elections)
70. [AI bot networks 2024 US election misinformation statistics research](https://misinforeview.hks.harvard.edu/article/the-origin-of-public-concerns-over-ai-supercharging-misinformation-in-the-2024-u-s-presidential-election/)
71. [AI bot networks 2024 US election misinformation statistics research](https://dividedwefall.org/ai-generated-misinformation-election-2024/)
72. [AI bot networks 2024 US election misinformation statistics research](https://www.emerald.com/reps/article/doi/10.1108/REPS-12-2024-0104/1307371/The-impact-of-disinformation-generated-by-AI-on)
73. [difference between simple and complex contagion in information diffusion examples](https://pmc.ncbi.nlm.nih.gov/articles/PMC11014438/)
74. [difference between simple and complex contagion in information diffusion examples](https://en.wikipedia.org/wiki/Complex_contagion)
75. [difference between simple and complex contagion in information diffusion examples](https://sociopatterns.org/assets/papers/PhysRevLett.130.247401.pdf)
76. [difference between simple and complex contagion in information diffusion examples](https://www.researchgate.net/publication/371600657_Distinguishing_Simple_and_Complex_Contagion_Processes_on_Networks)
77. [difference between simple and complex contagion in information diffusion examples](https://pubmed.ncbi.nlm.nih.gov/37390429/)
78. [time in United States of America](https://www.google.com/search?q=time+in+United+States+of+America)
79. [2025 algorithmic audit of X Twitter political content engagement](https://en.wikipedia.org/wiki/Algorithmic_amplification)
80. [2025 algorithmic audit of X Twitter political content engagement](https://quantumzeitgeist.com/algorithm-twitter-favors-right-leaning-accounts-reveals-correlated-behaviors/)
81. [2025 algorithmic audit of X Twitter political content engagement](https://www.theguardian.com/technology/2025/nov/27/partisan-x-posts-increase-political-polarisation-among-users-social-media-research)
82. [2025 algorithmic audit of X Twitter political content engagement](https://www.thestar.com.my/tech/tech-news/2025/12/01/researchers-tone-down-polarisation-on-x-with-tweaks-to-algorithm)
83. [2025 algorithmic audit of X Twitter political content engagement](https://arxiv.org/html/2503.02888v1)

**Sources:**
1. [datareportal.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFFm_Cf16XpLStPKt8jLhT6PCmj_y3oMU1_kLfJLxyQ_kUK1PTnpRDlsyyagQT8Mc2hF374wwfUmYn5-xopEwnhQjHO916Auo7VC_pMQe9lbs5yc3yJokGn2TGZKgkJsd4RcGp7NrBPjHIXW_2UWFmhItC8P-pGJIr5Vg==)
2. [smartinsights.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEcmFSwqucNAKoQl-mWnS6WTBx4PWIxGkdu8TwpkzjMs6mG1677rlBkBFF27CyDod4UEbHvhpzg_cpfJRqsgWbwBsG99toWrC4NXogp3ya5atsO_EKOJmg3-qminEsuQe6ZlBfH26XH9-rs01SwQNqaDPSsTD8BROWGsApXY1QS2pYzQ2CFJEiapNL0dTIn_i_suu7ELJqMUIdJ34WvrfQpSwY=)
3. [electroiq.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFXPNzZ5xBTsn_gWWYrjLAvll0lPSaK8QhbjoasVSUDbnvF7hX3iPYM5PseV0h8IbvWcSW--MK96oo2_bLj5gvJdzJSd1bwzzZopcVkbjrSYtwcbzI2KVaL3FGR32ZF3I9oaRydeQJ0O0_W38vjTt-SzzylwA==)
4. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQE2WaVERuEbVHTS83RCUWNNJ_LzBRk-MMG5yobCl3ceVnbMYQQaZLZodmcwPK4FSZCBkSIvUzl7ZvDk2xrPVi5UhDe9dbO2O51ldGhO-KPwbRC9Yyfce_VThA==)
5. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQG44oSJtW_HXki3ZqaGvn4znpgkk22n0Nwjb4EUbBa-z2ya_OjJ6vx1iRUELFIyPT4Z5Vm1sltv7i9G6n2-qLzPeOi10FlXPQawffGp4tmBPXUI2CFswA==)
6. [researchgate.net](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGMIJbMUMoWkj-J-98FswVpD7ST0tnF7Z8qn_Ho-p9730FPXVq4hEmRrRCeja-uUwyHigv-CyMxBrO1XaPh4F_rCaftDs8rK7Rnl-TnbECK85eL_rAig516B1Svl1dTUoSyX8yQaqfN6L9QwAWBuLuE-1rn6UIztXJm4TODwPhoN-CYn4X8y8P-XG4oicckkGEeFYmRgiDl9bP33ulJ2j4gzgGUIbndjcW9Sg==)
7. [cuhk.edu.hk](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEJVg9yuyt76ahJO_T1s8E4H3Cc3ujEXnZfDl3TfdOQZRAs-rFUpgfWG9IBq5Vg7t2Py_OIkiyA1ESrGI9BnBmNLTqUQ9Z-lcYwweYc80CPoNgDX4DXxRpBvqiHRJem-AjRqrhxIQpmE3ryU95zAL-4pKuiRcStOUGhaP0qGIqy0kGlasqhVKkBbMIhDTq9Zaj3)
8. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGgveYkCQCLRG6ORlnIxf3aO2w31BPScMF8fhRa9yp0qKAzCqFplKvABvyZut-7U05vuO-EvAqeDX-gdWi9q6E0anUg6yxuljdBvwOMeNkdF6sKN5-9Og==)
9. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHVxV0SR7cPXnlqfDyzWReBHq_rlMynH0psgxPuRcc0z2iMTXftf08VK08xdTH_F4-FwgFP59mxkb3w9zpsgaMzroUVscJYi5mb0kqlm-CbugU_fRRZ7IjSkv0wEtH0KF1Vg9QD047-)
10. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGGlgo0CY9HKGBGmBHgWOqyizV0tYr6-3YClGp0o90EHGq30l9c3GZeBlHoRHmvyWf-zobHrCSNl6INDNJ1HeGro8cFLqPwj0oWHPqvprDwu5NgkJ4esM1L1K-seC22Ve24bENABejP)
11. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQG-B8HS8e5d64kfxXjP39eyxrH0-0RpZppzwKKRf2c8is7eXE8NmPfEi0o_XP4LRqdZ8ZGqygI5kUmqeuF8TfNzWoCI17LC7Uf3Fux7knKKMbZKJK6ZJEpA7eLnQ7MkrwQx1uIbTQ7p)
12. [quora.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEgSoug-1iAFHtodRVjw22IW02T5DLg5OuoIYMpXCHzINMVUal7Rzd25kT0SoaqJqWQQ4rMcRh0pKDCwU1Znd3QqsHkjpYh7_biCpMzTEpoO6Cqo8qKH3fbjnliEULOkEQ_5-tmR0X25N4-CCq-Q_tJowmzt0sleThGq4_wN6M=)
13. [researchgate.net](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEGJL5XDG69it_Ybl5AhGgIr56VK9bqkCIn1ejTnsI8kyXVIGt0q5KtTJ-z9pwQOI01wg2cB1TELR0h8e2XcEdfjn3-tgSYUrdngfkGz7te2sygj7TwyEI041MPDT3tFd9_tNjo1nIg4kNKT8vf2U-QLKAXpoNa4NGIBoil_gQn1_IfEBBRiuD2bfYijohJ3Z_0z8-bZ2VX0XBTiDrAcyW66BI5I-3e8A==)
14. [wikipedia.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGkgIDUsk5sq22JJ52D0zSaf3hA-R7xoMPN65yJM8TU9qwgOKYqVb-VsJeoZniV17AjM4mrz-atTFBil14MQ6f6YIlmfAU33BY9rY2ehmiMlO9BfNSIz-UKiIclLdiwoSspyFlClA==)
15. [researchgate.net](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGrfxfZMyr2-edWHr48Yqj_-pxU-1JQi8ibx-9djmDyt1cf8uxZQGaB_qK6xqD85BKr2EE8ThZn4YNcBJkyYIgPVfqK7Fv3dX6gP9bARPNSw-rTer9ki66rgM4Sn6CLHEdJCrDJ7Xt0xE06kTWLMPSvHPhAE92dKx07ZJ1N0U_Svh1zByR_E-13wC2qer_atFyP0X77TVLlbo--q0_JaVruwb4miCZD0VGu1w==)
16. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHbXjIpscJBdz_-4kRG_CrQVvx_SI8qd5hfy0MFYLXAgC3rfyWbu-HshiU_ES-71Js65DUMzZlbyTzCdZyjMaYCrEbUaIQZKZdkFyUMftTHJznB4zc-5iFeNOqFpBJ77w==)
17. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHzttVjllew3S-qrOmZHmNkszAKqewwvJF5-J0etbZ23p-iQhWY1tp50P-nbMCLifnyk5NsRNhZln636WIvZn9mLtTGWv96T_4NFgyB2Mc13k54ctEuu4wCMfFHQTR8ehbprhwT0mKyug==)
18. [sociopatterns.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQETY7SCPM_wVfNyqQ4wLkDH-T0sabV5PMeCAYkhGhDQ-YaBwz-YY0tusi-3hbhqvwQqLohutV1GrCY3UjOcbCje6ezK-MHaHSF1RidBuNdNB0kTIK8zbjAj6eZKn74lDGInntYdF1Xco5WRxOqyWPGCrtjK9lrpCbo=)
19. [stanford.edu](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEGP442GHo080ho4AheeB3CXoNvoBT3pQurIaxG3CD7PNHtonzrdjeuVyL5wkhHq3qM-xZ7rvPuGUUwoT5oDfB8CozCPOisx0Yb7wARnIx8r2OIVvsQLhi6qNifVoCnF3DG0cFSzkKzM8sycQ==)
20. [researchgate.net](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHeQKUDxJjSOLlWOQmVV1Vo_-HdOjKMcjN3ZoVQudpJMio6sCcMUU0-1JT1_pUk0xHIr0h9anoNl3S1ek9NSvBDElID2-PbTB_jqoHUuri519yoZ8hromaDD5EAeBcYqIagmHfYuj95ZE8SNAzV7EpPje2Av_NzSTCBKttVPK6ksrdJL2UX8b5fikogJjY_SO9Hyq-P)
21. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEfNagWXnq7FVy4piiGAZhJuNd2-uGGUquGhu3I1s-Kpt2eywsHYXxUVedmbEKPFZ7VnPg39ApxlcNkZsvCYZZNCCPIStGdhckwuvB_OetrYiJZZ6Vqww==)
22. [informs.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEXW1m0mnJ6KO1Srj7r1RMj0QWpL3sorz0Ru0K_yIRVpJL_fg3SBIlXoOgd2nufJL7XebG8AxdulsjVxYyX_zalKXABKwLul7d3-dpLVlrNTRjfq-RL2J_G9JqSD245gpIFHCzovGTljUUPyNx7cMY=)
23. [researchgate.net](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFjNpIOMFFZiKXuQu7jEqShnpz30C4o0VY-NZiF-RuXQuli7dSQEYIHffOTDPVSd-VyzTF7xZMqxUEWnBSfGuH1nQq2aTRiJd3LCgkMElw5N__6MmwWjrtWeb7VrEqN101A45SkpNRU6ehHHygJoSK9SlHH4SyG0iE-CzJU3m0rvACqyuR336-UJWQaRYSTNcNOFVQIyiRpvycgYvoR1lUKL3dnTX1Z4_bnx0K93WDpwnRsmjpMJHZOWlYwAepfmcQ60_ey5n6X8dCmRXh53QqmOfEFCw==)
24. [wikipedia.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGzidtUACwDBEm4DMePWrew20zUXn4xf8DNVqf8gYrfmjADwyU9I9kxokuEty0d27GgCxgDj3976235UCmCqr6tvR4F1C6IvzejHCOArbcxc47MzGQEjpPc86yQntHRrtuvFpD6VGFumkz7DIo7)
25. [theguardian.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGJfPb2xYydMC-Fxz6FhmJ5ftjYcp2mkCFFEIBkTASaDVmwOZkBE7uKbj3dGk9VEgFFAVqmq02V0tdu5tUxlMkV3oXwrNJTxQT5SX_dVpLhYr8_A_brajqUhw5-431ndocQTh6pDmul4-10c6a2I8JgG_YTXjmtF5GOWR-WQyZmzl3E-AcH17zQhAdnEfBYx1kk7kIBXbRDSuOWsffEyVt-6oPQWMr0b0Ko5r8lWdRlMu8KRTQfrp6sMNM7DQ==)
26. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGqNJj7z1XXx9oK8iUw4pP7G3E3E_WUfcW001LbCqM6AKhEDpl4P3PuDy7f5e1etoG-B0Lusc1wVEn2Klg9eMR4_x9v-jIQZXp8nXvi23dJPbY9TzdpBh0tGg==)
27. [thestar.com.my](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGeYxWN9O4riO4pclA13P6CY89MXZoGmgjeddEYOF6B3VhzKAVcwWMuMadOzNrg4dkgep63-nt9Be1FgsgYXbCjIVOwInm3k1N4GxEfehslxDxSFI_VMh0N8D4jMygjcOvMZC79VIIpqV1EbSPzD7OrRJoupQRX98T4Znz4QcvSd-OPqNhRoKWv8rHIt_egy7Mqa-ROKz9J4BAbY15j2EZLrJ_fAtF8DYX-XZc=)
28. [accio.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGW9751EBC3QPzyJkrttPWwlhzqAk5PrifDktEecfLBVRS4m7sjfEBQ8y8H-mmhj5spbz0eYhkfEGazPHP5ErSShrhomi0cn1WJk2oGjqbHy0NPteeF338bpnjoBKm3NaYIPx4QMx_fGrYyGIC9zVc=)
29. [anu.edu.au](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFqFQJviLvkT74QX2jgkD93wO8JT-vbScvS8bHLdZxsgbUt99a5F0mM3lfzzsdE8wahIHhJSIohVYVRIe4wMaRGpJ3kk1AEHu2JiMxtOnGZ59Ii0hOmgEC5IVsCcD4ZNQap6_mjXgLzZNRWAgg8X_7-ANgaDI_qn2B0KZ01RViL0PBvBRkYola5nLWhCTIB7QECTdM=)
30. [highsnobiety.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGtY6_BJYyuzjiPDHzy1VHEbXJaMmGLHzIF3QvNfvFXTCzkmPtS1CGkfFjIyh2a3fUi8yZp7wehlxiiBYacWP7QAkMxCWCmsbU_-IcVJfM_iXD6x_OVeF--Jgixi6Wvk7qVsp71XZNv-f2PbsUzQf6CLvfMdb9LpYGZNP1kvA==)
31. [insidehook.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFwASVEn4M4iwmYS1gOzrSmEN-N_pAOpnY8Z1k0wV4LDCShUlzk-4k7Zkqtj70Y3NCZnF7Ud442eIbDFGY63Pg_KprEol6naiRBlYMK0rBNdrP4Q4bASWrgKFciOWRe-StwWyCGLTTlsg5aO-L7bDiVDf1-l5dn-nsn)
32. [civicscience.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEoL3cwLIndokXdqJmi9zmAbj_9qVCsfedx5uR5-exv4gKGWol5udwOBZStVz3_oaGkBaLk0luBJPw0Bqivy6ESkmzxI2gidoPSrLTXG6XmGPZWMRPGzslwC8-i6e6tef4LsTFtk5J6sClh0V-ZsTyOC2gZnKHBwn3VFESbX2cD5jh9TOm13DcleqcMfdaG977i8K6HqQFf20MiegYdx_gyVowxA7Ri)
33. [yougov.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEuAs-hLslUkRioAtYoA3LIVE3T8x7IwqTYaYKtEsGPFMsu9p4fjIYXEeoi-1MobPBpu46z7qDc6oFnBkCxntV72c15trpFVyh9ROL7M8rv6tXQ5SGxJ7O65quQlT411VHIcLQ4MdzhcewghoNLea4VHlpSzwyWQCogfIvFGXNc2MlmAgYdQDKITB2K2mSFKqiM28Bcg1G3GAc-CcnX8g==)
34. [accio.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFN0TIjNi4dHo-KH3QkdjI8Rov7dIAKHURV8zVu7aWlASCM1acgbtinKXw6WPGbRbKZ23qJrtP527fA6HZ7w4JlvWuNjhA-V3jjGU6qmMiTiKz6pSTc-rc1JfYe6uYGIuhCF0ZhSLHNO9OQ54CS_zxNbT0=)
35. [numerator.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQG_OalKccKdiWWRUXOjcN6z_N2deICwHItMY9hhgg1lwdZM_-Y0dpv5uklXU5mf9QCQjGUeZETuIHaGIgHhkeEv414wx2Nv3ImCcsMp0bAq9zZB4buLvGRUnGlYV460gKm33MGDHaE9DnaJxWtzfC21XErqIiQNkAg6oCHZI1K1)
36. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGnepRCMZ7kVBM659Ezzhv0sAuoA1Z_ucqms6UHGZj7U9ZJEwB_fu95QzQNmQc0SCjujcrD5lo0arSri2VufMF8PF6Y2rmaQbNoXQPzqtJoEA_AJeT1uMNZjuHfpc_ptT0dK1lyN9JZ)
37. [dividedwefall.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHBWLTtfSS_oJLkHjMGv43texoBr5lZ-4YY_idx8HCp-tmF6e0Q8TLaGZEDN_ePI6IcdF_o6h6-62JAs2FaIQZhmNEluyit7XPBVUNquTbV0Bpb3fh0WW-9OsjOgudTrVUBlP7W5ORXD4Fc4Vr2tdAdOPjQvjmA2KI9fQ==)
38. [usc.edu](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEIzBEnwEDYtsSmfVrMFPGPl1WlB8RZyLs_zWcFnkEJ2QsZiWSStbRp0VGezUE3Ia62ChxROvsA8OWEJYSPo2XtxP-sxqjHxw4SkiGV-4x_Lp3MC2DKl01F6OxVeid042jqaBXavbqb93-ewiXx5NdPToMAXSbcdBBkkLdl97BkSwhG5MH-RFtpAiYpOU3JN1ZbluwqsDTdKhyqktQ=)
39. [turing.ac.uk](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEX8isxq0yWMIvbFvandZ4OZ9zxLqnrdar_Q77tn_xAhkEqtWpqae4oNiQmVhSZeKc--_oC9uRl6-tbnKGaXH0V-Iqs1J_a8wlkoRhCxbyIdGCe6EPIze8wxo5FYs95L9ZH_DyQqQkRevdLiWlul5BKNbPjzWwjzGk7l66urvNWCnk_UXPDUUdyxUlMf9vzZ0wVf5G51yeizMWaSA==)
40. [harvard.edu](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGovK15bNmUPfL3KjVRqnuAYFeI4g1JM4eduVMJut-Z7rsSQR966_UJPIJbI3SEOdd_-hBvAbrsUZfRo38CInoFBghmmnNn3cB1cbPhs093rtDzTGAwN_dTPPaDvIWY4768k_IU8rN-WB40_JXuxARcqxUSI-mcr1v8NhQiR7gu4oFYro1mZ37QA3X3AFc695yv0JzE23UUcwy72TU73nStYJy9vAsQV1B-kI-XrXGeJF_pJVMLeGXY9nl105h_--2W11PhBUK5gLQPTWSv8g==)
41. [meedan.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHgqwBm6TAaYJRnPNbU2palvdS2lk22E9lDfkhzYj6REpZpQMVoZYsheGE_H6tWodC8X6y8AHEUWnLU2VxW9oX2VSKZ43znP_G4vnj8NiNAnSLiGowjUvEecAEpysSCM__xJwduBa0T37bOhuoqSpAlkgrB5sJZdj3LC1wqzwJxOKL7ofBO-TNG03ZQUQq28DlYqo7Oo7s0eFF5)
42. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFcKzDL_PkPfQ_3suDEVxk4NDlF8Fp4nrgHJpQjtVdTlW28OBaU1U5pmK1hDsUk5j1desWtSXCBq5opqD2RQo_CZbSNJpPIh_6vizcurR42m19dooHptkw=)
43. [oup.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHMPEz37_vz2GFWjq8myEKGr4PepA-JFDUkaY7ZO8KTxl2IQFEVAtF9oyyPA0N4ML18tBRM5EECsg_8GyTlqMDl-GKfDWOqBNvJQV0KxwSRTIsVxZ5iN_8yVkzOlhd0B9vMqv4mNkvNirrrTdpuKAEBpA==)
44. [github.io](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHsxdne7LN8bJh6wt92dYsHHj1IMwxAf2ejE9QTGg5rsbrBjx9r9AwBhZm9dZ8T8ilq715pvVT179uU5OatS6nEsfbuJYu-vCPZ1aZpxKiYdjF5MLAKxZ4MzrlsgWlyPe3y79mQ9oLP0Q280GkwFBQlR7fL0fYV2XRoNcGB5i9I8moIBOiMV2kPV21_uCoa)
45. [voxdev.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQH1-Up8v18BkaMwLwDafu199DmJy7WajMSALTogjG-PjoBEgpvofKgPAjcnUQ-w6mjPEUmxVTivDME8m937_bwF5PBl_N37O0MGQi-wRY_cNmQR-AmtkHxR586HJrbqES70Xb0WCEmNIBxHjU_RhZyLZ1LynL2c72Sj9FP0BGNml-u_M4AmVa8dqxAT30PTFdQtse9gIw98lBiBeRSl_u3ACiwMVSywQQ==)
46. [cornell.edu](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHo5vshltEArPn0UnRC_0_FP0Y-b3bfNKlFXpypMaGjCfsEshFQ7kiPSXCX1Scg8lbu3epLjUMeaEXHDKV4okCogZVifK0VtNKgoQ1QbzOurFmvsXaK-4psoayXcwXZTdSH5tWMaKt7I-AXs7mlrv8hMZECQdV2xRrV7m90hIYAwlFbxkiARxAQeDE2JPm5CxGzrgXffSD0WDA=)
47. [oup.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQF9qyq8-Wt0YPPZwtU-RIr2vUj_UWbcRNF99AyDNGnNvA7ahRsZuDD4l_WoxtB7KU7Ffgob6LKEs2LKWMLriXtHZPTk2B3xK4M1OLkBZArNLv3NXH9kCztMats9AlxpnrLmZDVERYjxGc0Rh2LOkcFB)
48. [researchgate.net](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFzg3PSlTxTUBJ-NJSKaSQ8L1wAxaPUyP-riOSaHUeQ7WuwffSOpZDyYJWew1pejhSRqGGTTN_zXNaMKhWxiCXHFojsEmx826klTBGHaT2VfTLoLWMewpT8QOWPFurP3_rDx6m9Hf3o8Riqukz-23FNl3b7XotdGqNBq4dXCG78CjOOEkLzzol-KJUERPK_r6HDtT8xK4VWAKQataHu2aUMmvUvnU9Ph6I87PIAIH1owzwXjUU9JzwxKHJdwjUeZjdIbKo4sa-ElmyDn-_miW_qWVRJLlYEXLT18YZgigUfix-evXWaSc_yaDA=)
49. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQF2wT3fVTe1BeOCL94yuhlGmkZALN0neNemSNn36MOSqtrEit9eBK0ek4Aodh_SpgFE1LqphZ8vpHHThImEMNnkWASMEzTvXKC0_ytIOz4JKkVkUUu_Ahm-R6Hyb2WYgw==)
50. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQF56Aw1crKFYv22-72LKs6bFCf-0ZqoV7jut4f4v-14-lkJyeOyTl71tsl9HIRPUmgF4uzohWFhKwjvNJzPHFpQOo97f169ZMIzVkY1GRlUBU_g9ojgO_uHHx-AiHAxnR2GwPozp92mOQ==)
51. [tandfonline.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFwp_eWyg2-02NtxI9Rx6aGy8dYcovrvrLcAnHD0Tfk_1MFTLXfgLHHnecVPLI7ooffwb7pg6glvO092J-2B_HEt1hHe0dcNxCRKHckzGUy1rGK-rQLfaG4gOgdX2Oqi_ruRAkaF4iINFfdiur-vZTVVvi1ySPQZMc=)
52. [jmir.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFae-fiPviM4xzsQnqkK_Fhb46EMG11otkvbA2TS5KSJl0W62YFWH_-m7jyA17BUkYi68Komj5iMeS67X6ucBk-f3N-rI-ugDEsNmcjGW9Pz_ZxDPfa6_1acB_zquly)
53. [tandfonline.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEXkGvgnFdKKAiAhA0vTUSzhKG3S4w2nKbCc8x7kDNrGRfp0nL-iaDvCD_TnFvGAxB8ktoQSzcAQVcJr0NhQ9WVK-YHNvzmdot1rfvjh2ZCcZpeDk9-2vKiyJIeD80F3YteMQd8WS2FUuAN7EpQA037gfFgF8ymYC4=)
54. [ed.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEyV1akx3EX_zThQabzGmm_4jp9iK_sjhLgZnbrStwAUn_jGNo_DyA2DjrExUbhJGrYzJFesXVcxwlhBv9GZS_lGZuUaIqwrA7iz00XVvJEorwmd7rvbdeQMDWIIG5O3gZfm5XeXII=)
55. [uw.edu](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHMSgWk24enTKBt6IvKFCsJSC9zkjSa033BHF3KX0lfo8TAhb_TamaVFd-juJ94pGN0USVbN--cmxI3cXbCkEsCyDTetSv2D5T-CNIC7-1ZBzsY0aAXG-CL_5f6k_B1oiB1-ghzYdWs8G7MV5MYoix-3FTBBZ96Pr3775o=)
56. [youtube.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGi_GrrcOAm310bSFj1dRcC5cWkPwGr6CDX1NySHncSSlg_rpNzOBURj_dxq6TY9FEmC25ctAoVA7Lhb2_6kyodXRdbxeUdEfZpRTsjnlZxTqPv7mKIYlmuyV052p3rUsX3)