The Evolution of Digital Content: From Hypertext to Generative AI

1. Introduction and Method

Few phrases are used as loosely as "digital content". In industry usage it may denote anything from a scanned textbook to a fifteen-second dance clip, a Wikipedia article, a recommender-ranked news feed or a paragraph produced by a language model. This review treats digital content not as a fixed category of artefact but as a socio-technical regime: a configuration of technologies of production, channels of distribution, economic models, and norms of authorship and trust. Read this way, the history of digital content is less a story of ever-better media and more a story of shifting scarcities. When the cost of producing and copying information collapses, as it has repeatedly since 1945, what becomes scarce is attention [1, 2, 3], the ability to find and rank what matters [4, 5], and the ability to know whether what one is seeing is true [6, 7].

The review asks three questions. First, what were the defining technical and economic features of each successive regime, from hypertext to generative AI? Second, what does the empirical literature say about the consequences of those regimes for knowledge, learning, trust and wellbeing? Third, what open problems are likely to define the period 2026-2031, particularly for producers of educational and edutainment content in India?

1.1 Method

This is a narrative (integrative) review, not a systematic review or meta-analysis. The topic spans at least six disciplines whose literatures use incompatible vocabularies and designs, which makes a single PRISMA-style search protocol impractical. Instead, sources were selected through four routes:

1. Seminal works that the fields themselves treat as foundational (for example Bush, Nelson, Berners-Lee, Simon, O'Reilly, Jenkins, Mayer and Sweller).
2. Recent peer-reviewed syntheses and high-impact empirical studies published mainly between 2018 and 2025, identified through citation chaining from those foundations and through targeted bibliographic queries.
3. Authoritative institutional reports with transparent methods, including the Reuters Institute Digital News Report, Pew Research Center, the OECD, UNESCO and the ITU.
4. India-specific sources where a primary report was publicly accessible and its figures could be checked against the original document.

Every reference was verified against an external record before inclusion. Journal articles, conference papers and many books were checked against the Crossref or DataCite registries by DOI, matching title, first author and year; reports and books without a DOI were checked against the publisher's or issuing body's own page. Items that could not be verified were dropped. Every statistic quoted in the text is taken from a cited source that was opened during preparation of this review; where a figure is a survey estimate, the survey population is stated.

Three limitations follow from this design. The selection is purposive and therefore open to the author's framing. English-language scholarship dominates, which under-represents work published in Indian languages. And the field moves quickly: some findings on generative AI cited here are preprints or working papers, and they are flagged as such.

2. From Hypertext to Web 1.0

2.1 The associative dream

The intellectual prehistory of digital content is usually dated to Vannevar Bush's 1945 essay "As We May Think" [8]. Writing as director of the US Office of Scientific Research and Development, Bush worried that the scientific record was growing faster than any scholar could consult it. His proposed remedy, the memex, was a desk-sized microfilm device in which a reader could link one document to another and save those links as "trails" for others to follow. The memex was never built, but the essay contained two ideas that recur throughout this review: that the bottleneck of knowledge is retrieval and selection rather than storage, and that the paths readers make through material can themselves be valuable content.

Twenty years later Theodor Nelson gave the idea a name. In a 1965 paper to the Association for Computing Machinery he introduced "hypertext" and "hypermedia" and described a file structure for "the complex, the changing and the indeterminate" [9]. Nelson's later Project Xanadu imagined two-way links, transclusion (quoting by reference rather than copying) and micropayments to original authors. Those features matter historically because the web that eventually succeeded did not adopt them. Links on the web are one-way and break silently; copying rather than referencing became the default; and authorship and payment were left to be solved later, largely by advertising. Several of today's problems of provenance and attribution can be read as consequences of those early simplifications.

2.2 Berners-Lee and the document web

Tim Berners-Lee's 1989 memorandum at CERN, "Information Management: A Proposal", framed the problem in Bush-like terms: information about people, projects and software was being lost inside an organisation with high turnover [10]. The proposed solution combined hypertext with the internet, using three simple standards that would become URLs, HTTP and HTML. In a 1992 paper the CERN team described the World-Wide Web as a "wide-area hypermedia information retrieval initiative" deliberately designed so that anyone could publish without central permission [11].

The design choices of that period (open standards, no registration, tolerance for broken links) explain both the explosive growth of the early web and its characteristic weaknesses. Web 1.0 content was largely published rather than conversed: institutional homepages, digitised documents, personal pages and directories. Discovery depended on hand-built directories and, increasingly, on search engines. The economic model was unsettled, and the "content" category was dominated by text and static images because bandwidth was scarce.

2.3 What the first regime established

Three structural features of the hypertext and Web 1.0 era persist. First, permissionless publication separated the act of publishing from institutional gatekeeping, a precondition for every later regime. Second, link structure as signal: the realisation that links could be used to rank relevance, which made search engines the first great curators of the open web. Third, a missing layer of provenance: the web had no native way to record who made something, when, or whether it had been altered. That gap is the background to the provenance standards discussed in Section 9.

3. Web 2.0 and User-Generated Content

3.1 The participatory turn

By the mid-2000s the dominant metaphor changed from publication to participation. Tim O'Reilly's 2005 essay "What Is Web 2.0" codified a set of design patterns: the web as platform, harnessing collective intelligence, data as the core competitive asset, and a "perpetual beta" of continuously updated services [12]. Kaplan and Haenlein's widely cited definition of social media as applications built on "the ideological and technological foundations of Web 2.0" that allow "the creation and exchange of User Generated Content" captured the business-school consensus of the period [13].

Cultural theorists described the same shift in different terms. Henry Jenkins's Convergence Culture argued that media were converging not only technically but socially, as fans and audiences became active participants who remixed, annotated and circulated content [14]. Yochai Benkler's The Wealth of Networks offered an economic account: when the physical capital required for information production is widely distributed, "commons-based peer production" can rival firms and markets [15].

3.2 Blogs, wikis and the question of quality

Blogs gave individuals continuous, chronological publishing tools; wikis made collaborative editing routine. Wikipedia became the emblematic case, and the question of whether volunteer-produced knowledge could be trusted was examined early. A 2005 Nature investigation by Giles compared science entries in Wikipedia and Encyclopaedia Britannica and found the difference in the number of serious errors to be small [16]. The comparison was disputed, but it shifted debate from whether open collaboration could produce reliable reference content to under what conditions it does. A later systematic review of scholarly research on Wikipedia's content found generally favourable assessments of accuracy in many domains, alongside persistent concerns about coverage gaps, uneven quality and systemic bias in who edits [17].

3.3 The limits of participation

Two cautions emerged from this literature. First, participation was always unequal: early studies of the "second-level digital divide" showed that differences in online skills and uses persisted even when access spread [18], and later work found that inequalities shifted from physical access to material access and quality of devices and connections [19]. Second, the openness celebrated in Web 2.0 rhetoric was increasingly hosted on privately owned services. Gillespie's analysis of the word "platform" showed how companies used it strategically to present themselves simultaneously as neutral conduits to users, as opportunities to advertisers and as non-publishers to regulators [20]. That rhetorical work anticipated the platform era.

The more user-generated the web became, the more its architecture came to be owned by a handful of firms.

4. Platformisation and Algorithmic Curation

4.1 From websites to platforms

"Platformisation" denotes the process by which platforms become the dominant infrastructural and economic model of the web and penetrate other sectors [21, 22]. Helmond traced how social media companies extended their reach across the web through APIs, plug-ins and social buttons, making external web data "platform ready" [21]. Gerlitz and Helmond described the resulting "like economy", in which social buttons turned engagement into data that could be aggregated and monetised [23]. Van Dijck, Poell and de Waal's The Platform Society generalised the argument: platforms reshape public values in transport, health, news and education, not just entertainment [24]. Plantin and colleagues showed how platforms such as Google and Facebook acquire infrastructure-like characteristics while retaining platform-like business models [25].

For cultural production specifically, Nieborg and Poell argued that platformisation makes content "contingent": continuously altered, modularised and optimised in response to platform metrics and policies [26]. Later work extended this into a research agenda on markets, infrastructures and governance in platformised cultural production [22]. Gorwa's survey of platform governance distinguished the governance of platforms by states from governance by platforms over their users [27], while Gillespie showed that content moderation is not a peripheral function but constitutive of what platforms are [28].

4.2 Recommender systems

The technical core of platformisation is the recommender system. The Netflix Prize era popularised latent-factor models; Koren, Bell and Volinsky's account of matrix factorisation became a standard reference [29]. Gomez-Uribe and Hunt described Netflix's production recommender as a family of algorithms covering personalised ranking, similar-item lists and artwork selection, and reported that recommendation drives a large share of viewing on the service [30]. Covington, Adams and Sargin described YouTube's two-stage deep neural network architecture, with a candidate-generation model and a ranking model trained mainly on watch time rather than clicks [31]. That choice of objective is consequential: optimising for expected watch time rewards content that holds attention, which is not the same as content that informs.

Communication scholars studied what these systems do to visibility. Bucher's analysis of Facebook's News Feed argued that algorithmic ranking produces a "threat of invisibility" that disciplines how users and publishers behave [4]. Rieder, Matamoros-Fernández and Coromina's study of YouTube search rankings described distinct "ranking cultures" that vary by topic and over time [5]. Creators themselves develop folk theories, or "algorithmic gossip", about what the system rewards [32], and play a "visibility game" in which they adjust content to perceived algorithmic preferences [33].

4.3 Do algorithms polarise? Evidence from the 2020s

Public debate often assumes that recommender systems trap users in "filter bubbles". The empirical picture is more qualified. A Reuters Institute literature review concluded that the evidence for algorithmically induced echo chambers is weaker than commonly assumed, and that most people's news diets are more diverse than the metaphor implies, although small groups of highly partisan users do cluster [34]. Bakshy, Messing and Adamic found in 2015 that individual choices limited exposure to cross-cutting news on Facebook more than ranking did [35]. Cinelli and colleagues, comparing several platforms, found evidence of homophilic clustering whose strength varied by platform design [36].

The 2023 papers from the US 2020 Facebook and Instagram election study provide the strongest causal evidence to date. Switching users from algorithmic to chronological feeds for three months substantially changed what they saw and reduced time on platform, but did not detectably change polarisation or political attitudes [37]. Removing reshared content reduced exposure to political news but likewise did not detectably affect polarisation [38]. Reducing exposure to like-minded sources did not reduce polarisation [39]. At the same time, González-Bailón and colleagues documented substantial ideological segregation in exposure to political news, concentrated on the right of the spectrum and in content rated false by fact-checkers [40]. For YouTube, an audit by Ribeiro and colleagues found pathways from milder to more extreme channels in commenting communities [41], while a later counterfactual-bot study found that, on average, following recommendations pushed users toward more moderate content than their own choices did [42]. Bail and colleagues found that exposure to opposing views on Twitter could increase polarisation among some users [43].

These results do not show that algorithms are harmless. They show that effects on attitudes over short periods are small relative to user choice and prior disposition, while effects on exposure, time spent and what becomes visible are large. For producers of content, the second set of effects is arguably the more important.

4.4 The attention economy

Herbert Simon observed in 1971 that "a wealth of information creates a poverty of attention" [1]. Davenport and Beck turned the observation into management doctrine [2], and Wu's history of "attention merchants" traced the business of harvesting attention from nineteenth-century penny newspapers to social media [3]. Zuboff's account of "surveillance capitalism" went further, arguing that behavioural data extracted from users is used to predict and shape behaviour [44]; Couldry and Mejias framed similar processes as "data colonialism" [45]. These are contested, theory-heavy accounts, but they share an empirical premise that is not in dispute: on advertising-funded platforms the commercial value of content is measured primarily in attention captured, not in understanding produced.

5. Mobile, Short-Form Video and the Creator Economy

5.1 The smartphone as the default screen

The smartphone moved content consumption from the desk to the pocket and made video the dominant format. Global internet use is now the majority condition but not a universal one: the ITU estimates that about 5.5 billion people, roughly 68 per cent of the world's population, were online in 2024, with 93 per cent in high-income countries but only 27 per cent in low-income countries, and with a persistent urban-rural gap (83 versus 48 per cent) [46]. Among US teenagers surveyed by Pew in autumn 2024, 95 per cent had access to a smartphone, 90 per cent used YouTube, roughly six in ten used TikTok, and nearly half said they were online "almost constantly" [47].

The Reuters Institute's 2025 survey across 48 markets documents the same shift in news. The proportion consuming social video news weekly grew from 52 per cent in 2020 to 65 per cent in 2025, and in India, the Philippines, Thailand and Kenya more people now say they prefer to watch news rather than read it [48]. Over half of the Indian, Thai and Kenyan samples reported using YouTube for news [48].

5.2 YouTube and the professionalisation of amateurs

Burgess and Green's study of YouTube identified a persistent tension between the platform's "community" rhetoric and its commercial media logics [49]. Over two decades that tension resolved into a new industry. Cunningham and Craig call it "social media entertainment": a proto-industry of professionalising creators operating across platforms, financed by advertising revenue share, brand deals and fan funding, and situated between Hollywood and Silicon Valley [50].

Creator labour has its own literature. Duffy's concept of "aspirational labour" describes the largely unpaid work of aspiring creators who invest in self-presentation in the hope of future reward [51]. Abidin's ethnographic work on internet celebrity documents the vernacular strategies through which micro-celebrities cultivate intimacy and attention [52], and her study of TikTok shows how the platform's algorithmic distribution decouples visibility from follower counts, creating an attention economy driven by virality [53]. Marketing scholarship has examined why audiences trust influencers: Lou and Yuan found that informative value and perceived credibility of influencer-generated posts predicted trust in branded content [54], and a systematic review by Vrontis and colleagues organised the field's antecedents, mediators and outcomes [55].

5.3 TikTok and the logic of short-form

TikTok's design, a full-screen vertical feed populated mainly by a recommendation engine rather than a social graph, changed the grammar of digital content. Kaye, Chen and Zeng traced the "parallel platformisation" of Douyin and TikTok [56]. Zulli and Zulli argued that TikTok encourages "imitation publics" organised around replicating sounds, formats and templates [57]. Schellewald's interview study suggested that users experience TikTok's appeal through its affordances of easy entertainment and discovery rather than social connection [58]. Hutchinson's work on micro-platformisation shows that activist and civic communicators also adapt to these formats [59].

The Reuters Institute reports that TikTok is the fastest-growing network for news, used by 16 per cent of its global sample for news weekly in 2025, and that respondents across markets cite online influencers and personalities, alongside national politicians, as the largest perceived sources of false or misleading information (47 per cent each) [48].

6. Streaming, Edutainment and the Learning Sciences of Video

6.1 Streaming and the economics of catalogues

Subscription video-on-demand services industrialised long-form streaming and made the recommender a central editorial actor [30]. For education the relevant change was cultural as much as technical: audiences became accustomed to high production values, binge viewing and personalised catalogues, which raised expectations for educational video as well.

6.2 MOOCs: promise and pivot

Massive open online courses were the education sector's first platformised experiment. Early enthusiasm gave way to sober evidence. Margaryan, Bianco and Littlejohn evaluated the instructional design quality of 76 MOOCs and found that most scored poorly on established principles of instruction [60]. Hew and Cheung's review documented learner motivations alongside high attrition and challenges for instructors [61]. Reich and Ruipérez-Valiente's analysis of edX data concluded that MOOC providers had "pivoted" from democratising higher education toward professional and corporate training, with low completion rates and participation concentrated among already-educated learners in affluent countries [62].

The COVID-19 emergency of 2020 forced a far larger experiment. Reviews documented both the speed of adoption and its inequities, including connectivity, device access and teacher preparedness [63, 64]. The OECD's Digital Education Outlook 2023 argued that effective digital education requires an ecosystem (infrastructure, interoperable tools, data governance and teacher capacity) rather than isolated products [65].

6.3 What the learning sciences say about educational video

The design of educational content is the one area of this review with a mature experimental literature. Cognitive load theory holds that working memory is severely limited, so instruction should minimise extraneous load (processing caused by poor design) and manage intrinsic load (complexity inherent to the material) [66, 67]. Twenty years after its first formulation, Sweller, van Merriënboer and Paas restated the theory's architecture and its links to evolutionary accounts of knowledge [68].

Mayer's cognitive theory of multimedia learning derives design principles from dual-channel processing, limited capacity and active processing, and the evidence is collected in The Cambridge Handbook of Multimedia Learning [69]. Among the best-supported principles are coherence (exclude extraneous material), signalling (highlight essential structure), redundancy (do not duplicate narration with identical on-screen text), segmenting (break lessons into learner-paced parts) and personalisation (use conversational style). Mayer later argued that motivation must be incorporated into the model, since a learner who is not engaged will not do the cognitive work that design makes possible [70]. Fiorella and Mayer reviewed eight generative learning strategies, such as summarising, self-explaining and teaching, that move learners from passive viewing to active construction [71]. Mayer's earlier critique of "pure discovery learning" is a useful corrective to the assumption that engagement alone produces learning: guided activity outperforms unguided exploration [72].

Empirically, Guo, Kim and Rubin's analysis of 6.9 million video-watching sessions on edX found that shorter videos were more engaging, with median engagement time of about six minutes regardless of video length; that informal talking-head segments and tablet-style drawing were more engaging than high-budget studio recordings; and that speaking rate and enthusiasm mattered [73]. Brame translated these and related findings into guidelines for managing cognitive load, maximising engagement and promoting active learning with video [74]. Noetel and colleagues' systematic review and meta-analysis of randomised trials in higher education found that replacing existing teaching methods with video produced small improvements in learning, and that supplementing existing methods with video produced larger gains [75].

Engagement is necessary for learning, but it is not evidence of learning.

6.4 Edutainment and microlearning: what is and is not known

"Edutainment" and "microlearning" are industry terms with weak theoretical definitions. The evidence above supports some of their claims (short segments, conversational presentation, signalling, narrative motivation) and undermines others. Segmenting is beneficial when it reduces load and lets learners pace themselves; fragmenting a curriculum into unconnected clips is not the same thing, and it risks raising the load of integrating ideas across clips. High production value is not, in itself, a predictor of engagement [73]. And entertainment that adds vivid but irrelevant material can violate the coherence principle [69]. The defensible position is that cinematic and narrative techniques can support learning when they direct attention to essential content and motivate generative processing, and harm it when they compete with that content for attention.

7. The Indian Context: Cheap Data, Vernacular Audiences and EdTech

7.1 Scale and language

India illustrates how the economics of connectivity reshape content. The entry of Reliance Jio in 2016 with low-cost 4G data triggered a price war that made mobile data dramatically cheaper, an episode widely treated as the turning point in mass mobile video adoption in the country. Before that inflection, the KPMG-Google study Indian Languages: Defining India's Internet estimated that Indian-language internet users had grown at a compound rate of 41 per cent between 2011 and 2016 to reach 234 million, overtaking 175 million English-language users, and projected 536 million Indian-language users by 2021 [76]. The same report stated that 88 per cent of Indian-language internet users were more likely to respond to digital advertisements in their own language than in English [76]. These projections are now dated, but the direction they identified, toward a majority-vernacular, mobile-first internet, has held.

The Reuters Institute's India sample (which, it should be noted, is an online, English-speaking, younger and more urban sample than the national population) shows a strong preference for video: over half use YouTube for news, more prefer watching news to reading it, and 18 per cent reported using AI chatbots for news weekly in 2025, much higher than in the UK [48]. The Institute also notes that India is WhatsApp's largest market and that misinformation circulating in large groups has in the past been linked to mob violence [48].

7.2 Policy: Digital India

The Government of India's Digital India programme, launched in 2015, framed connectivity, digital identity and public digital infrastructure as development priorities [77]. Its educational implications operate mainly through infrastructure (connectivity, payment rails and public platforms) rather than content design. The literature on digital divides warns that access alone does not equalise outcomes: differences in skills, device quality and the material conditions of use persist after access spreads [18, 19].

7.3 Misinformation and literacy interventions in India

India has produced some of the most informative field evidence on countering misinformation. Badrinathan's field experiment during the 2019 Indian general election delivered an hour-long, in-person media literacy training to respondents in Bihar and found no significant improvement in their ability to identify misinformation, with the training possibly backfiring among strong supporters of the ruling party [78]. The result is a caution against assuming that literacy curricula developed in Western contexts transfer, and it reinforces the importance of design that fits the medium (closed messaging groups, video, local languages).

7.4 EdTech after the boom

India's EdTech sector expanded rapidly during pandemic school closures and then contracted. Peer-reviewed, independent evaluation of learning outcomes from India's large commercial EdTech platforms remains scarce relative to their scale, and many publicly cited figures come from company statements or investor materials that are not independently audited. This review therefore does not report such figures. The broader evidence on AI in education warns that most published applications have been weakly connected to pedagogical theory and to educators' perspectives [79], a gap that applies with particular force where products are designed for scale first.

8. Misinformation, Trust and Fact-Checking

8.1 Concepts

Wardle and Derakhshan's Council of Europe report proposed the influential framework of "information disorder", distinguishing misinformation (false content shared without intent to harm), disinformation (false content shared with intent to harm) and malinformation (genuine content shared to cause harm) [6]. Tandoc, Lim and Ling's review of how scholars had used the term "fake news" found six distinct types, from satire and parody to fabrication, manipulation, advertising and propaganda, and recommended analytical precision over the catch-all label [80]. Lazer and colleagues framed "the science of fake news" as an interdisciplinary agenda [7], and Lewandowsky, Ecker and Cook argued that the problem extends beyond individual falsehoods to a "post-truth" condition in which shared standards of evidence erode [81].

8.2 Spread

The most cited empirical study in this area is Vosoughi, Roy and Aral's analysis of roughly 126,000 rumour cascades on Twitter between 2006 and 2017. False news diffused "farther, faster, deeper, and more broadly" than true news, and the effect was driven by human sharing rather than bots; the authors linked this to the greater novelty of false content [82]. Allcott and Gentzkow estimated the reach of fabricated stories during the 2016 US election and concluded that, while widely shared, they were unlikely to have changed the outcome [83]. Bruns, Harrington and Hurcombe traced how COVID-19/5G conspiracy theories moved from fringe communities into mainstream attention through celebrities and media coverage [84].

A growing methodological literature cautions against overstatement. Altay, Berriche and Acerbi argue that the prevalence and effects of misinformation are often exaggerated, that most people consume little of it, and that alarmism can itself erode trust in reliable information [85]. The Reuters Institute reports that 58 per cent of its 2025 global sample were concerned about telling what is real from fake online, while noting that such concern is often driven by perceived bias rather than exposure to objectively fabricated content [48].

8.3 Why people believe

Pennycook and Rand's work has been central to the psychology of belief. In "Lazy, not biased" they found that susceptibility to partisan fake news is better explained by lack of reasoning than by motivated reasoning: people who scored higher on analytic thinking were better at discerning true from false headlines regardless of whether the headlines fit their politics [86]. Pennycook, Cannon and Rand showed that a single prior exposure increases the perceived accuracy of fake headlines (an illusory truth effect) [87]. Their 2021 review summarised the evidence that inattention, rather than inability, explains much sharing of falsehoods [88]. Ecker and colleagues' review in Nature Reviews Psychology integrated cognitive, social and affective drivers of misinformation belief and of resistance to correction [89]. Fernbach and colleagues found that extreme opponents of genetically modified foods knew the least about the science but believed they knew the most, illustrating how overconfidence can block correction [90].

8.4 What works: correction, prompts and prebunking

The evidence on interventions is cautiously encouraging. Lewandowsky and colleagues' 2012 review established the "continued influence effect", whereby retracted misinformation continues to shape reasoning, and set out evidence-based recommendations for effective correction [91]. Early concern about a "backfire effect" [92] has been substantially qualified by later work, and Walter and colleagues' meta-analysis found that fact-checking has a significant positive effect on correcting political beliefs on average, moderated by prior beliefs, ideology and message design [93]. Vraga and Bode tested correction placement and type on Instagram and found that corrections can reduce misperceptions on visual platforms [94].

Two scalable approaches have strong experimental support. Accuracy prompts (subtly asking users to consider whether a headline is accurate) improved the quality of subsequent sharing in survey and field experiments [95]. Prebunking or psychological inoculation (forewarning people about manipulation techniques with weakened examples) improved discernment; Roozenbeek and colleagues showed that short inoculation videos shown as YouTube advertisements improved recognition of manipulation techniques at scale [96]. Lewandowsky and van der Linden reviewed the inoculation literature [97], and van der Linden summarised susceptibility, spread and interventions in a public-health framing [98]. Badrinathan's null result in India (Section 7.3) is a reminder that effect sizes depend on context and medium [78].

9. Generative AI and Synthetic Content

9.1 Technical foundations

The generative regime rests on a small number of technical advances. Generative adversarial networks, introduced by Goodfellow and colleagues, trained a generator against a discriminator to produce realistic images [99]. The Transformer architecture of Vaswani and colleagues replaced recurrence with self-attention and made very large language models trainable at scale [100]. Brown and colleagues' GPT-3 paper demonstrated that scaling such models yields "few-shot" abilities across tasks without task-specific fine-tuning [101]. Instruction tuning with human feedback made these models far more usable as assistants [102], and the GPT-4 technical report described a multimodal model with markedly higher performance on professional and academic benchmarks, while disclosing little about architecture or training data [103]. In images, denoising diffusion probabilistic models offered a stable alternative to GANs and underlie most current text-to-image systems [104].

Bommasani and colleagues coined "foundation models" for models trained on broad data that can be adapted to many downstream tasks, and they emphasised that the homogenisation this creates means that defects in one model propagate into many applications [105]. Bender and colleagues' "stochastic parrots" paper warned of environmental costs, opaque training data and the tendency of fluent output to be mistaken for understanding [106]. Ji and colleagues' survey of "hallucination" in natural language generation documented the systematic production of fluent but unsupported content [107].

9.2 Synthetic media and deepfakes

Deepfakes, synthetic audio-visual media that depict people saying or doing things they did not, were examined early by legal scholars as a threat to privacy, democracy and national security [108]. Technical surveys describe a rapid arms race between generation and detection [109, 110]. Vaccari and Chadwick's experiment found that a deepfake political video did not mainly deceive viewers outright but increased uncertainty, which in turn reduced trust in news on social media [111]. Nightingale and Farid found that AI-synthesised faces were indistinguishable from real ones and were rated as slightly more trustworthy [112].

9.3 Synthetic text and persuasion

Experimental work suggests that machine-generated text is already persuasive. Kreps, McCain and Brundage found that people judged AI-generated news stories as credible at rates similar to human-written ones [113]. Spitale, Biller-Andorno and Germani found that GPT-3 could produce accurate information that was easier to understand than human-written tweets, but also disinformation that was more compelling, and that participants could not reliably tell AI from human tweets [114]. Goldstein and colleagues found AI-generated propaganda to be nearly as persuasive as real propaganda in a preregistered survey experiment [115]. Hanley and Durumeric documented a sharp rise in machine-generated articles on misinformation websites following the release of ChatGPT [116]. The Reuters Institute's 2025 survey found audiences expect AI to make news cheaper to produce but less transparent, less accurate and less trustworthy [48].

9.4 Model collapse and the data commons

A distinctive risk of the generative regime is recursive: models trained on content produced by other models. Shumailov and colleagues showed in Nature that indiscriminate training on model-generated data causes "model collapse", in which the tails of the original distribution disappear and outputs degrade across generations [117]. The finding gives an economic and technical argument, not only an ethical one, for preserving and labelling human-generated data. It also suggests that verified human-made content may become more valuable as a training and reference resource precisely as synthetic content becomes abundant.

9.5 Provenance, watermarking and detection

Three families of response are being developed. Provenance standards attach cryptographically signed metadata describing the origin and edit history of a media file; the C2PA technical specification is the most widely adopted industry effort [118]. Watermarking embeds a detectable signal in generated content:

Kirchenbauer and colleagues proposed a statistical watermark for language-model text [119], and Fernandez and colleagues' "Stable Signature" roots watermarks in latent diffusion image models [120]. Post hoc detection attempts to classify content as human or machine-made. The evidence on detection is sobering. Sadasivan and colleagues argued theoretically and empirically that reliable detection of AI text becomes harder as models improve and that paraphrasing defeats many detectors [121], and Liang and colleagues found that GPT detectors disproportionately misclassify writing by non-native English speakers as AI-generated [122], a serious fairness problem in multilingual settings such as India. Provenance metadata can be stripped, and watermarks can be removed or forged; none of these methods is sufficient alone.

9.6 Generative AI and education

Educational responses moved from alarm to guidance within two years. Early commentaries in Nature and elsewhere mapped implications for science and scholarship [123], and a multidisciplinary opinion paper led by Dwivedi collected over forty expert perspectives on opportunities and risks [124]. Kasneci and colleagues set out opportunities (personalised tutoring, feedback, accessibility) and challenges (bias, over-reliance, assessment integrity) for education [125]. Cotton, Cotton and Shipway examined academic integrity in the era of ChatGPT [126]. Systematic reviews of AI in education before and after the LLM era found rapid growth in applications but weak grounding in learning theory and limited involvement of educators [79, 127, 128]; Yan and colleagues' review of LLMs in education highlighted practical and ethical challenges including reproducibility, transparency and privacy [129]. Mollick and Mollick proposed structured ways of assigning AI to students, such as AI as tutor, coach or "student" to be taught [130].

The most important empirical caution comes from a field experiment with nearly a thousand high-school students in Turkey reported by Bastani and colleagues: unrestricted access to a GPT-4 assistant improved performance on practice problems but harmed performance on subsequent exams taken without it, whereas a version designed with tutoring safeguards largely mitigated that harm [131]. The study is a working paper and should be read as such, but it is consistent with the learning-sciences principle that learning requires effortful generative processing that can be short-circuited [71]. UNESCO's guidance calls for human-centred, age-appropriate deployment, data protection and validation of tools before classroom use [132]. Jobin, Ienca and Vayena's mapping of 84 AI ethics guidelines found convergence on transparency, justice, non-maleficence, responsibility and privacy, but wide divergence in how these principles are interpreted and implemented [133].

10. Wellbeing and the Attention Debate

10.1 The case for concern

No debate about digital content is more public than the one over adolescent mental health. Twenge and colleagues reported that increases in depressive symptoms and suicide-related outcomes among US adolescents after 2010 coincided with rising new-media screen time, and argued the two were linked [134]. Haidt's The Anxious Generation develops a broader thesis: that a shift from "play-based" to "phone-based" childhood in the early 2010s is a major cause of rising adolescent anxiety and depression, and it proposes collective norms such as no smartphones before high school [135]. Some experimental studies support a causal contribution. Hunt and colleagues found that limiting social media use to about 30 minutes per day for three weeks reduced loneliness and depression among university students [136]. Allcott and colleagues' large randomised experiment paying users to deactivate Facebook for four weeks found small improvements in subjective wellbeing, along with reduced political knowledge and polarisation [137]. Boers and colleagues' longitudinal study of Canadian adolescents found within-person associations between increases in social media and television use and increases in depressive symptoms [138]. Research on design features describes how variable rewards, infinite scroll and social metrics can encourage habitual use [139].

10.2 The case for caution

Other researchers argue that the evidence does not support strong causal claims. Orben and Przybylski applied specification curve analysis to three large datasets and found that the association between digital technology use and adolescent wellbeing was negative but very small, explaining at most 0.4 per cent of the variation in wellbeing, comparable to the association with eating potatoes [140]. Applying the same approach to time-use diary data, they again found little evidence of substantial negative associations [141]. Heffer and colleagues' longitudinal study found that social media use did not predict later depressive symptoms, although depressive symptoms predicted later social media use among girls [142]. Vuorre, Orben and Przybylski found no evidence that associations between technology engagement and mental-health problems had increased over time [143]. Odgers and Jensen's annual research review concluded that the evidence is mixed and that digital environments may amplify offline vulnerabilities rather than create them [144]; Odgers's 2024 review of The Anxious Generation in Nature argued that the book's causal claims are not supported by the science [145]. Valkenburg, Meier and Beyens' umbrella review found that most reviews report weak associations and emphasised large individual differences in how adolescents respond [146]. Ferguson's meta-analytic review of social media experiments argued that experimental effects are small and vulnerable to methodological artefacts [147].

10.3 Reading the debate honestly

Several conclusions are defensible. First, average associations between total screen or social media time and wellbeing are small, but averages can hide meaningful effects for subgroups and for particular kinds of use. Second, the "time spent" measure is a poor proxy: what people do, with whom, and at what cost to sleep and offline activity is more informative than hours. Third, experimental reductions in use sometimes improve wellbeing modestly, which is compatible with both camps. Fourth, young people themselves report mixed experiences; In Pew's autumn-2024 survey of US teens, 74 per cent said social media made them feel more connected to their friends' lives, while 48 per cent said it had a mostly negative effect on people their age, compared with 14 per cent who said so about themselves [148]. Fifth, the debate on attention is related but distinct. Ophir, Nass and Wagner found that heavy media multitaskers performed worse on tests of task switching and filtering irrelevant information [149], but the direction of causation remains unresolved.

For content producers the practical lesson is narrower than either camp's rhetoric: design choices that exploit compulsive use are a choice, not a technical necessity, and they are the features the evidence most consistently identifies as problematic [139].

11. Future Directions, 2026-2031

The following directions are offered as research and practice priorities, not forecasts.

11.1 Provenance as infrastructure

If synthetic content becomes the majority of new content in some formats, provenance will need to become infrastructural in the way HTTPS became infrastructural for security. Standards such as C2PA [118], combined with watermarking [119, 120], are likely to be necessary but insufficient, because metadata can be stripped and detectors can be unreliable and biased [121, 122]. Research priorities include measuring whether provenance labels actually change audience trust and behaviour, and how they perform on messaging apps and in low-bandwidth contexts.

11.2 Protecting the human data commons

Model collapse [117] implies that human-generated, documented and verified content has both epistemic and economic value as a reference corpus. Open questions include licensing, compensation for creators whose work trains models, and how to keep curated human archives distinguishable from synthetic ones.

11.3 AI tutors that preserve effort

The early evidence that unrestricted AI help can harm learning [131], read alongside cognitive load and generative learning theory [68, 71], points toward AI systems designed to scaffold rather than replace learner effort. Rigorous, preregistered, independent trials, particularly in Indian languages and in government schools, are a clear gap [79, 132].

11.4 Creator-led information and news

The Reuters Institute documents the rise of personality-led news creators and of AI chatbots as news sources, especially among under-25s [48, 150]. How audiences calibrate trust in individual creators, and how creators can adopt verification norms without institutional backing, deserve focused research building on influencer-credibility studies [54, 55].

11.5 Evidence-based platform governance

Large-scale collaborations with platforms produced the strongest causal evidence on algorithmic effects to date [37, 39, 40]. Sustained, independent researcher access to platform data is a precondition for answering the questions in Sections 4, 8 and 10, and governance debates [27] will shape whether such access exists.

11.6 Vernacular and multilingual AI

Most evaluation of generative models and detectors is conducted in English. For India's majority-vernacular internet [76], quality, safety and fairness in Indian languages are first-order research problems, not localisation details.

12. Conclusion

Across eighty years, the evolution of digital content shows a consistent pattern. Each regime, from Bush's memex and Nelson's hypertext through the document web, the participatory web, the platformised attention economy and now generative AI, lowered the cost of producing and distributing information. Each therefore moved scarcity elsewhere: to attention [1], to curation and ranking [4, 31], and to verification [6]. Generative AI does not create the problem of provenance, which was latent in the web's original design, but it removes the residual friction that once made fabrication costly [113, 115].

The evidence reviewed here argues against both utopian and catastrophic narratives. Open collaboration can produce reliable knowledge [16, 17]; algorithmic ranking strongly shapes exposure yet shows small short-run effects on attitudes [37]; false news spreads quickly but is consumed by fewer people than often assumed [82, 85]; corrections and prebunking work modestly [93, 96]; and average links between screen use and adolescent wellbeing are small and contested [135, 140]. In education the most robust findings are old ones: working memory is limited, active processing matters, and design should direct attention to what is essential [66, 69, 73]. The new technologies do not repeal these constraints; they make it easier either to respect them or to ignore them at scale.

13. Implications for Human-Made, Verified Educational Content

Without making claims for any particular organisation, the literature suggests several modest implications for producers of human-made, verified educational and edutainment content, including in India.

• Make provenance visible. Credit named authors and reviewers, cite sources, date content and keep correction logs. Where feasible, adopt open provenance standards [118]. Audiences still say that trusted brands with a record of accuracy are where they go to check information [48].
• Design for learning, not only for watch time. Use segmenting, signalling and coherence, keep segments short, and build in retrieval and generative activities [69, 71, 73]. Treat cinematic technique as a means of directing attention, not an end.
• Use AI as an assistant under human verification. Disclose AI assistance, verify factual claims and avoid AI features that do learners' thinking for them [131, 132].
• Respect attention. Avoid design patterns associated with compulsive use [139], and measure success by learning outcomes rather than engagement alone.
• Serve vernacular audiences well. Produce in the languages learners use, and test comprehension and misinformation resilience in those languages rather than assuming transfer [76, 78].
• Evaluate independently. Publish methods and outcomes, and invite independent evaluation, since evidence on EdTech outcomes is thin [79].

None of these practices guarantees impact. They are, however, the practices most consistent with the evidence assembled in this review.

References

1. Simon, H. A. (1971). Designing organizations for an information-rich world. In M. Greenberger (Ed.), Computers, communications, and the public interest (pp. 37–72). Johns Hopkins Press. https://iiif.library.cmu.edu/file/Simon_box00055_fld04178_bdl0002_doc0001/Simon_box00055_fld04178_bdl0002_doc0001.pdf

2. Davenport, T. H., & Beck, J. C. (2001). The attention economy: Understanding the new currency of business. Harvard Business School Press. https://archive.org/details/attentioneconomy00thom

3. Wu, T. (2016). The attention merchants: The epic scramble to get inside our heads. Knopf. https://www.penguinrandomhouse.com/books/234876/the-attention-merchants-by-tim-wu/

4. Bucher, T. (2012). Want to be on the top? Algorithmic power and the threat of invisibility on Facebook. New Media & Society, 14(7), 1164–1180. https://doi.org/10.1177/1461444812440159

5. Rieder, B., Matamoros-Fernández, A., & Coromina, Ò. (2018). From ranking algorithms to ‘ranking cultures’. Convergence: The International Journal of Research into New Media Technologies, 24(1), 50–68. https://doi.org/10.1177/1354856517736982

6. Wardle, C., & Derakhshan, H. (2017). Information disorder: Toward an interdisciplinary framework for research and policy making (Council of Europe report DGI(2017)09). Council of Europe. https://rm.coe.int/information-disorder-toward-an-interdisciplinary-framework-for-researc/168076277c

7. Lazer, D. M. J., Baum, M. A., Benkler, Y., Berinsky, A. J., Greenhill, K. M., Menczer, F., Metzger, M. J., Nyhan, B., Pennycook, G., Rothschild, D., Schudson, M., Sloman, S. A., Sunstein, C. R., Thorson, E. A., Watts, D. J., & Zittrain, J. L. (2018). The science of fake news. Science, 359(6380), 1094–1096. https://doi.org/10.1126/science.aao2998

8. Bush, V. (1945). As we may think. The Atlantic Monthly, 176(1), 101–108. https://www.theatlantic.com/magazine/archive/1945/07/as-we-may-think/303881/

9. Nelson, T. H. (1965). Complex information processing. In Proceedings of the 1965 20th national conference (pp. 84–100). ACM Press. https://doi.org/10.1145/800197.806036

10. Berners-Lee, T. (1989). Information management: A proposal. CERN. https://www.w3.org/History/1989/proposal.html

11. Berners-Lee, T. J. (1992). The world-wide web. Computer Networks and ISDN Systems, 25(4-5), 454–459. https://doi.org/10.1016/0169-7552(92)90039-s

12. O'Reilly, T. (2005). What is Web 2.0: Design patterns and business models for the next generation of software. O'Reilly Media. https://www.oreilly.com/pub/a/web2/archive/what-is-web-20.html

13. Kaplan, A. M., & Haenlein, M. (2010). Users of the world, unite! The challenges and opportunities of Social Media. Business Horizons, 53(1), 59–68. https://doi.org/10.1016/j.bushor.2009.09.003

14. Jenkins, H. (2006). Convergence culture: Where old and new media collide. New York University Press. https://nyupress.org/9780814742952/convergence-culture/

15. Benkler, Y. (2017). The Wealth of Networks. Yale University Press. https://doi.org/10.12987/9780300127232

16. Giles, J. (2005). Internet encyclopaedias go head to head. Nature, 438(7070), 900–901. https://doi.org/10.1038/438900a

17. Mesgari, M., Okoli, C., Mehdi, M., Nielsen, F. Å., & Lanamäki, A. (2015). “The sum of all human knowledge”: A systematic review of scholarly research on the content of Wikipedia. Journal of the Association for Information Science and Technology, 66(2), 219–245. https://doi.org/10.1002/asi.23172

18. Hargittai, E. (2002). Second-Level Digital Divide: Differences in People's Online Skills. First Monday, 7(4). https://doi.org/10.5210/fm.v7i4.942

19. van Deursen, A. J., & van Dijk, J. A. (2019). The first-level digital divide shifts from inequalities in physical access to inequalities in material access. New Media & Society, 21(2), 354–375. https://doi.org/10.1177/1461444818797082

20. Gillespie, T. (2010). The politics of ‘platforms’. New Media & Society, 12(3), 347–364. https://doi.org/10.1177/1461444809342738

21. Helmond, A. (2015). The Platformization of the Web: Making Web Data Platform Ready. Social Media + Society, 1(2). https://doi.org/10.1177/2056305115603080

22. Poell, T., Nieborg, D., & van Dijck, J. (2019). Platformisation. Internet Policy Review, 8(4). https://doi.org/10.14763/2019.4.1425

23. Gerlitz, C., & Helmond, A. (2013). The like economy: Social buttons and the data-intensive web. New Media & Society, 15(8), 1348–1365. https://doi.org/10.1177/1461444812472322

24. van Dijck, J., Poell, T., & de Waal, M. (2018). The Platform Society. Oxford University Press. https://doi.org/10.1093/oso/9780190889760.001.0001

25. Plantin, J. C., Lagoze, C., Edwards, P. N., & Sandvig, C. (2018). Infrastructure studies meet platform studies in the age of Google and Facebook. New Media & Society, 20(1), 293–310. https://doi.org/10.1177/1461444816661553

26. Nieborg, D. B., & Poell, T. (2018). The platformization of cultural production: Theorizing the contingent cultural commodity. New Media & Society, 20(11), 4275–4292. https://doi.org/10.1177/1461444818769694

27. Gorwa, R. (2019). What is platform governance? Information, Communication & Society, 22(6), 854–871. https://doi.org/10.1080/1369118X.2019.1573914

28. Gillespie, T. (2019). Custodians of the Internet. Yale University Press. https://doi.org/10.12987/9780300235029

29. Koren, Y., Bell, R., & Volinsky, C. (2009). Matrix Factorization Techniques for Recommender Systems. Computer, 42(8), 30– 37. https://doi.org/10.1109/MC.2009.263

30. Gomez-Uribe, C. A., & Hunt, N. (2016). The Netflix Recommender System. ACM Transactions on Management Information Systems, 6(4), 1–19. https://doi.org/10.1145/2843948

31. Covington, P., Adams, J., & Sargin, E. (2016). Deep Neural Networks for YouTube Recommendations. In Proceedings of the 10th ACM Conference on Recommender Systems (pp. 191–198). ACM. https://doi.org/10.1145/2959100.2959190

32. Bishop, S. (2019). Managing visibility on YouTube through algorithmic gossip. New Media & Society, 21(11-12), 2589–2606. https://doi.org/10.1177/1461444819854731

33. Cotter, K. (2019). Playing the visibility game: How digital influencers and algorithms negotiate influence on Instagram. New Media & Society, 21(4), 895–913. https://doi.org/10.1177/1461444818815684

34. Ross Arguedas, A., Robertson, C. T., Fletcher, R., & Nielsen, R. K. (2022). Echo chambers, filter bubbles, and polarisation: A literature review. Reuters Institute for the Study of Journalism. https://reutersinstitute.politics.ox.ac.uk/echo-chambers-filter-bubbles-and-polarisation-literature-review

35. Bakshy, E., Messing, S., & Adamic, L. A. (2015). Exposure to ideologically diverse news and opinion on Facebook. Science, 348(6239), 1130–1132. https://doi.org/10.1126/science.aaa1160

36. Cinelli, M., De Francisci Morales, G., Galeazzi, A., Quattrociocchi, W., & Starnini, M. (2021). The echo chamber effect on social media. Proceedings of the National Academy of Sciences, 118(9). https://doi.org/10.1073/pnas.2023301118

37. Guess, A. M., Malhotra, N., Pan, J., Barberá, P., Allcott, H., Brown, T., Crespo-Tenorio, A., Dimmery, D., Freelon, D., Gentzkow, M., González-Bailón, S., Kennedy, E., Kim, Y. M., Lazer, D., Moehler, D., Nyhan, B., Rivera, C. V., Settle, J., Thomas, D. R., .. Tucker, J. A. (2023). How do social media feed algorithms affect attitudes and behavior in an election campaign? Science, 381(6656), 398–404. https://doi.org/10.1126/science.abp9364

38. Guess, A. M., Malhotra, N., Pan, J., Barberá, P., Allcott, H., Brown, T., Crespo-Tenorio, A., Dimmery, D., Freelon, D., Gentzkow, M., González-Bailón, S., Kennedy, E., Kim, Y. M., Lazer, D., Moehler, D., Nyhan, B., Rivera, C. V., Settle, J., Thomas, D. R., .. Tucker, J. A. (2023). Reshares on social media amplify political news but do not detectably affect beliefs or opinions. Science, 381(6656), 404–408. https://doi.org/10.1126/science.add8424

39. Nyhan, B., Settle, J., Thorson, E., Wojcieszak, M., Barberá, P., Chen, A. Y., Allcott, H., Brown, T., Crespo-Tenorio, A., Dimmery, D., Freelon, D., Gentzkow, M., González-Bailón, S., Guess, A. M., Kennedy, E., Kim, Y. M., Lazer, D., Malhotra, N., Moehler, D., .. Tucker, J. A. (2023). Like-minded sources on Facebook are prevalent but not polarizing. Nature, 620(7972), 137–144. https://doi.org/10.1038/s41586-023-06297-w

40. González-Bailón, S., Lazer, D., Barberá, P., Zhang, M., Allcott, H., Brown, T., Crespo-Tenorio, A., Freelon, D., Gentzkow, M., Guess, A. M., Iyengar, S., Kim, Y. M., Malhotra, N., Moehler, D., Nyhan, B., Pan, J., Rivera, C. V., Settle, J., Thorson, E., .. Tucker, J. A. (2023). Asymmetric ideological segregation in exposure to political news on Facebook. Science, 381(6656), 392– 398. https://doi.org/10.1126/science.ade7138

41. Ribeiro, M. H., Ottoni, R., West, R., Almeida, V. A. F., & Meira, W. (2020). Auditing radicalization pathways on YouTube. In Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency (pp. 131–141). ACM. https://doi.org/10.1145/3351095.3372879

42. Hosseinmardi, H., Ghasemian, A., Rivera-Lanas, M., Horta Ribeiro, M., West, R., & Watts, D. J. (2024). Causally estimating the effect of YouTube’s recommender system using counterfactual bots. Proceedings of the National Academy of Sciences, 121(8). https://doi.org/10.1073/pnas.2313377121

43. Bail, C. A., Argyle, L. P., Brown, T. W., Bumpus, J. P., Chen, H., Hunzaker, M. B. F., Lee, J., Mann, M., Merhout, F., & Volfovsky, A. (2018). Exposure to opposing views on social media can increase political polarization. Proceedings of the National Academy of Sciences, 115(37), 9216–9221. https://doi.org/10.1073/pnas.1804840115

44. Zuboff, S. (2019). The age of surveillance capitalism: The fight for a human future at the new frontier of power. PublicAffairs. https://www.publicaffairsbooks.com/titles/shoshana-zuboff/the-age-of-surveillance-capitalism/9781610395694/

45. Couldry, N., & Mejias, U. A. (2019). Data Colonialism: Rethinking Big Data’s Relation to the Contemporary Subject. Television & New Media, 20(4), 336–349. https://doi.org/10.1177/1527476418796632

46. International Telecommunication Union. (2024). Measuring digital development: Facts and figures 2024. ITU. https://www.itu.int/itu-d/reports/statistics/facts-figures-2024/

47. Faverio, M., & Sidoti, O. (2024, December 12). Teens, social media and technology 2024. Pew Research Center. https://www.pewresearch.org/internet/2024/12/12/teens-social-media-and-technology-2024/

48. Newman, N., Ross Arguedas, A., Robertson, C. T., Nielsen, R. K., & Fletcher, R. (2025). Reuters Institute digital news report 2025. Reuters Institute for the Study of Journalism. https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2025

49. Burgess, J., & Green, J. (2018). YouTube: Online video and participatory culture (2nd ed.). Polity. https://www.politybooks.com/bookdetail?book_slug=youtube-online-video-and-participatory-culture-2nd-edition--9780745660189

50. Cunningham, S., & Craig, D. (2019). Social media entertainment: The new intersection of Hollywood and Silicon Valley. New York University Press. https://nyupress.org/9781479890880/social-media-entertainment/

51. Duffy, B. E. (2016). The romance of work: Gender and aspirational labour in the digital culture industries. International Journal of Cultural Studies, 19(4), 441–457. https://doi.org/10.1177/1367877915572186

52. Abidin, C. (2018). Internet Celebrity: Understanding Fame Online. Emerald Publishing Limited. https://doi.org/10.1108/9781787560765

53. Abidin, C. (2020). Mapping Internet Celebrity on TikTok: Exploring Attention Economies and Visibility Labours. Cultural Science Journal, 12(1), 77–103. https://doi.org/10.5334/csci.140

54. Lou, C., & Yuan, S. (2019). Influencer Marketing: How Message Value and Credibility Affect Consumer Trust of Branded Content on Social Media. Journal of Interactive Advertising, 19(1), 58–73. https://doi.org/10.1080/15252019.2018.1533501

55. Vrontis, D., Makrides, A., Christofi, M., & Thrassou, A. (2021). Social media influencer marketing: A systematic review, integrative framework and future research agenda. International Journal of Consumer Studies, 45(4), 617–644. https://doi.org/10.1111/ijcs.12647

56. Kaye, D. B. V., Chen, X., & Zeng, J. (2021). The co-evolution of two Chinese mobile short video apps: Parallel platformization of Douyin and TikTok. Mobile Media & Communication, 9(2), 229–253. https://doi.org/10.1177/2050157920952120

57. Zulli, D., & Zulli, D. J. (2022). Extending the Internet meme: Conceptualizing technological mimesis and imitation publics on the TikTok platform. New Media & Society, 24(8), 1872–1890. https://doi.org/10.1177/1461444820983603

58. Schellewald, A. (2023). Understanding the popularity and affordances of TikTok through user experiences. Media, Culture & Society, 45(8), 1568–1582. https://doi.org/10.1177/01634437221144562

59. Hutchinson, J. (2021). Micro-platformization for digital activism on social media. Information, Communication & Society, 24(1), 35–51. https://doi.org/10.1080/1369118x.2019.1629612

60. Margaryan, A., Bianco, M., & Littlejohn, A. (2015). Instructional quality of Massive Open Online Courses (MOOCs). Computers & Education, 80, 77–83. https://doi.org/10.1016/j.compedu.2014.08.005

61. Hew, K. F., & Cheung, W. S. (2014). Students’ and instructors’ use of massive open online courses (MOOCs): Motivations and challenges. Educational Research Review, 12, 45–58. https://doi.org/10.1016/j.edurev.2014.05.001

62. Reich, J., & Ruipérez-Valiente, J. A. (2019). The MOOC pivot. Science, 363(6423), 130–131. https://doi.org/10.1126/science.aav7958

63. Adedoyin, O. B., & Soykan, E. (2023). Covid-19 pandemic and online learning: the challenges and opportunities. Interactive Learning Environments, 31(2), 863–875. https://doi.org/10.1080/10494820.2020.1813180

64. Williamson, B., Eynon, R., & Potter, J. (2020). Pandemic politics, pedagogies and practices: digital technologies and distance education during the coronavirus emergency. Learning, Media and Technology, 45(2), 107–114. https://doi.org/10.1080/17439884.2020.1761641

65. OECD. (2023). OECD Digital Education Outlook 2023. OECD Publishing. https://doi.org/10.1787/c74f03de-en

66. Sweller, J. (1988). Cognitive Load During Problem Solving: Effects on Learning. Cognitive Science, 12(2), 257–285. https://doi.org/10.1207/s15516709cog1202_4

67. Paas, F., Renkl, A., & Sweller, J. (2003). Cognitive Load Theory and Instructional Design: Recent Developments. Educational Psychologist, 38(1), 1–4. https://doi.org/10.1207/s15326985ep3801_1

68. Sweller, J., van Merriënboer, J. J. G., & Paas, F. (2019). Cognitive Architecture and Instructional Design: 20 Years Later. Educational Psychology Review, 31(2), 261–292. https://doi.org/10.1007/s10648-019-09465-5

69. Mayer, R. E. (Ed.). (2014). The Cambridge Handbook of Multimedia Learning. Cambridge University Press. https://doi.org/10.1017/CBO9781139547369

70. Mayer, R. E. (2014). Incorporating motivation into multimedia learning. Learning and Instruction, 29, 171–173. https://doi.org/10.1016/j.learninstruc.2013.04.003

71. Fiorella, L., & Mayer, R. E. (2016). Eight Ways to Promote Generative Learning. Educational Psychology Review, 28(4), 717– 741. https://doi.org/10.1007/s10648-015-9348-9

72. Mayer, R. E. (2004). Should There Be a Three-Strikes Rule Against Pure Discovery Learning? American Psychologist, 59(1), 14–19. https://doi.org/10.1037/0003-066X.59.1.14

73. Guo, P. J., Kim, J., & Rubin, R. (2014). How video production affects student engagement. In Proceedings of the first ACM conference on Learning @ scale conference (pp. 41–50). ACM. https://doi.org/10.1145/2556325.2566239

74. Brame, C. J. (2016). Effective Educational Videos: Principles and Guidelines for Maximizing Student Learning from Video Content. CBE—Life Sciences Education, 15(4), es6. https://doi.org/10.1187/cbe.16-03-0125

75. Noetel, M., Griffith, S., Delaney, O., Sanders, T., Parker, P., del Pozo Cruz, B., & Lonsdale, C. (2021). Video Improves Learning in Higher Education: A Systematic Review. Review of Educational Research, 91(2), 204–236. https://doi.org/10.3102/0034654321990713

76. KPMG India & Google. (2017). Indian languages: Defining India's internet. https://assets.kpmg.com/content/dam/kpmg/in/pdf/2017/04/Indian-languages-Defining-Indias-Internet.pdf

77. Government of India, Ministry of Electronics and Information Technology. (n.d.). Digital India programme. https://www.digitalindia.gov.in/

78. Badrinathan, S. (2021). Educative Interventions to Combat Misinformation: Evidence from a Field Experiment in India. American Political Science Review, 115(4), 1325–1341. https://doi.org/10.1017/s0003055421000459

79. Zawacki-Richter, O., Marín, V. I., Bond, M., & Gouverneur, F. (2019). Systematic review of research on artificial intelligence applications in higher education – where are the educators? International Journal of Educational Technology in Higher Education, 16(1). https://doi.org/10.1186/s41239-019-0171-0

80. Tandoc, E. C., Lim, Z. W., & Ling, R. (2018). Defining “Fake News”. Digital Journalism, 6(2), 137–153. https://doi.org/10.1080/21670811.2017.1360143

81. Lewandowsky, S., Ecker, U. K. H., & Cook, J. (2017). Beyond misinformation: Understanding and coping with the “post-truth” era. Journal of Applied Research in Memory and Cognition, 6(4), 353–369. https://doi.org/10.1016/j.jarmac.2017.07.008

82. Vosoughi, S., Roy, D., & Aral, S. (2018). The spread of true and false news online. Science, 359(6380), 1146–1151. https://doi.org/10.1126/science.aap9559

83. Allcott, H., & Gentzkow, M. (2017). Social Media and Fake News in the 2016 Election. Journal of Economic Perspectives, 31(2), 211–236. https://doi.org/10.1257/jep.31.2.211

84. Bruns, A., Harrington, S., & Hurcombe, E. (2020). ‘Corona? 5G? or both?’: the dynamics of COVID-19/5G conspiracy theories on Facebook. Media International Australia, 177(1), 12–29. https://doi.org/10.1177/1329878X20946113

85. Altay, S., Berriche, M., & Acerbi, A. (2023). Misinformation on Misinformation: Conceptual and Methodological Challenges. Social Media + Society, 9(1). https://doi.org/10.1177/20563051221150412

86. Pennycook, G., & Rand, D. G. (2019). Lazy, not biased: Susceptibility to partisan fake news is better explained by lack of reasoning than by motivated reasoning. Cognition, 188, 39–50. https://doi.org/10.1016/j.cognition.2018.06.011

87. Pennycook, G., Cannon, T. D., & Rand, D. G. (2018). Prior exposure increases perceived accuracy of fake news. Journal of Experimental Psychology: General, 147(12), 1865–1880. https://doi.org/10.1037/xge0000465

88. Pennycook, G., & Rand, D. G. (2021). The Psychology of Fake News. Trends in Cognitive Sciences, 25(5), 388–402. https://doi.org/10.1016/j.tics.2021.02.007

89. Ecker, U. K. H., Lewandowsky, S., Cook, J., Schmid, P., Fazio, L. K., Brashier, N., Kendeou, P., Vraga, E. K., & Amazeen, M. A. (2022). The psychological drivers of misinformation belief and its resistance to correction. Nature Reviews Psychology, 1(1), 13–29. https://doi.org/10.1038/s44159-021-00006-y

90. Fernbach, P. M., Light, N., Scott, S. E., Inbar, Y., & Rozin, P. (2019). Extreme opponents of genetically modified foods know the least but think they know the most. Nature Human Behaviour, 3(3), 251–256. https://doi.org/10.1038/s41562-018-0520-3

91. Lewandowsky, S., Ecker, U. K. H., Seifert, C. M., Schwarz, N., & Cook, J. (2012). Misinformation and Its Correction. Psychological Science in the Public Interest, 13(3), 106–131. https://doi.org/10.1177/1529100612451018

92. Nyhan, B., & Reifler, J. (2010). When Corrections Fail: The Persistence of Political Misperceptions. Political Behavior, 32(2), 303–330. https://doi.org/10.1007/s11109-010-9112-2

93. Walter, N., Cohen, J., Holbert, R. L., & Morag, Y. (2020). Fact-Checking: A Meta-Analysis of What Works and for Whom. Political Communication, 37(3), 350–375. https://doi.org/10.1080/10584609.2019.1668894

94. Vraga, E. K., Kim, S. C., Cook, J., & Bode, L. (2020). Testing the Effectiveness of Correction Placement and Type on Instagram. The International Journal of Press/Politics, 25(4), 632–652. https://doi.org/10.1177/1940161220919082

95. Pennycook, G., Epstein, Z., Mosleh, M., Arechar, A. A., Eckles, D., & Rand, D. G. (2021). Shifting attention to accuracy can reduce misinformation online. Nature, 592(7855), 590–595. https://doi.org/10.1038/s41586-021-03344-2

96. Roozenbeek, J., van der Linden, S., Goldberg, B., Rathje, S., & Lewandowsky, S. (2022). Psychological inoculation improves resilience against misinformation on social media. Science Advances, 8(34). https://doi.org/10.1126/sciadv.abo6254

97. Lewandowsky, S., & van der Linden, S. (2021). Countering Misinformation and Fake News Through Inoculation and Prebunking. European Review of Social Psychology, 32(2), 348–384. https://doi.org/10.1080/10463283.2021.1876983

98. van der Linden, S. (2022). Misinformation: susceptibility, spread, and interventions to immunize the public. Nature Medicine, 28(3), 460–467. https://doi.org/10.1038/s41591-022-01713-6

99. Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., & Bengio, Y. (2020). Generative adversarial networks. Communications of the ACM, 63(11), 139–144. https://doi.org/10.1145/3422622

100. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017). Attention Is All You Need [Preprint]. arXiv. https://doi.org/10.48550/arXiv.1706.03762

101. Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D. M., Wu, J., Winter, C., .. Amodei, D. (2020). Language Models are Few-Shot Learners [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2005.14165

102. Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C., Mishkin, P., Zhang, C., Agarwal, S., Slama, K., Ray, A., Schulman, J., Hilton, J., Kelton, F., Miller, L., Simens, M., Askell, A., Welinder, P., Christiano, P., Leike, J., & Lowe, R. (2022). Training Language Models to Follow Instructions with Human Feedback. In Advances in Neural Information Processing Systems 35 (pp. 27730–27744). Neural Information Processing Systems Foundation, Inc. (NeurIPS). https://doi.org/10.52202/068431-2011

103. OpenAI. (2023). GPT-4 technical report [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2303.08774

104. Ho, J., Jain, A., & Abbeel, P. (2020). Denoising Diffusion Probabilistic Models [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2006.11239

105. Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., Bernstein, M. S., Bohg, J., Bosselut, A., Brunskill, E., Brynjolfsson, E., Buch, S., Card, D., Castellon, R., Chatterji, N., Chen, A., Creel, K., Davis, J. Q., Demszky, D., .. Liang, P. (2021). On the Opportunities and Risks of Foundation Models [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2108.07258

106. Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the Dangers of Stochastic Parrots. In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (pp. 610–623). ACM. https://doi.org/10.1145/3442188.3445922

107. Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y. J., Madotto, A., & Fung, P. (2023). Survey of Hallucination in Natural Language Generation. ACM Computing Surveys, 55(12), 1–38. https://doi.org/10.1145/3571730

108. Chesney, R., & Citron, D. K. (2018). Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.3213954

109. Mirsky, Y., & Lee, W. (2022). The Creation and Detection of Deepfakes. ACM Computing Surveys, 54(1), 1–41. https://doi.org/10.1145/3425780

110. Verdoliva, L. (2020). Media Forensics and DeepFakes: An Overview. IEEE Journal of Selected Topics in Signal Processing, 14(5), 910–932. https://doi.org/10.1109/JSTSP.2020.3002101

111. Vaccari, C., & Chadwick, A. (2020). Deepfakes and Disinformation: Exploring the Impact of Synthetic Political Video on Deception, Uncertainty, and Trust in News. Social Media + Society, 6(1). https://doi.org/10.1177/2056305120903408

112. Nightingale, S. J., & Farid, H. (2022). AI-synthesized faces are indistinguishable from real faces and more trustworthy. Proceedings of the National Academy of Sciences, 119(8). https://doi.org/10.1073/pnas.2120481119

113. Kreps, S., McCain, R. M., & Brundage, M. (2022). All the News That’s Fit to Fabricate: AI-Generated Text as a Tool of Media Misinformation. Journal of Experimental Political Science, 9(1), 104–117. https://doi.org/10.1017/XPS.2020.37

114. Spitale, G., Biller-Andorno, N., & Germani, F. (2023). AI model GPT-3 (dis)informs us better than humans. Science Advances, 9(26). https://doi.org/10.1126/sciadv.adh1850

115. Goldstein, J. A., Chao, J., Grossman, S., Stamos, A., & Tomz, M. (2024). How persuasive is AI-generated propaganda? PNAS Nexus, 3(2). https://doi.org/10.1093/pnasnexus/pgae034

116. Hanley, H. W. A., & Durumeric, Z. (2024). Machine-Made Media: Monitoring the Mobilization of Machine-Generated Articles on Misinformation and Mainstream News Websites. Proceedings of the International AAAI Conference on Web and Social Media, 18, 542–556. https://doi.org/10.1609/icwsm.v18i1.31333

117. Shumailov, I., Shumaylov, Z., Zhao, Y., Papernot, N., Anderson, R., & Gal, Y. (2024). AI models collapse when trained on recursively generated data. Nature, 631(8022), 755–759. https://doi.org/10.1038/s41586-024-07566-y

118. Coalition for Content Provenance and Authenticity. (2024). C2PA technical specification (Version 2.1). https://c2pa.org/specifications/specifications/2.1/specs/C2PA_Specification.html

119. Kirchenbauer, J., Geiping, J., Wen, Y., Katz, J., Miers, I., & Goldstein, T. (2023). A Watermark for Large Language Models [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2301.10226

120. Fernandez, P., Couairon, G., Jégou, H., Douze, M., & Furon, T. (2023). The Stable Signature: Rooting Watermarks in Latent Diffusion Models. In 2023 IEEE/CVF International Conference on Computer Vision (ICCV) (pp. 22409–22420). IEEE. https://doi.org/10.1109/iccv51070.2023.02053

121. Sadasivan, V. S., Kumar, A., Balasubramanian, S., Wang, W., & Feizi, S. (2023). Can AI-Generated Text be Reliably Detected? [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2303.11156

122. Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., & Zou, J. (2023). GPT detectors are biased against non-native English writers. Patterns, 4(7), 100779. https://doi.org/10.1016/j.patter.2023.100779

123. Stokel-Walker, C., & Van Noorden, R. (2023). What ChatGPT and generative AI mean for science. Nature, 614(7947), 214– 216. https://doi.org/10.1038/d41586-023-00340-6

124. Dwivedi, Y. K., Kshetri, N., Hughes, L., Slade, E. L., Jeyaraj, A., Kar, A. K., Baabdullah, A. M., Koohang, A., Raghavan, V., Ahuja, M., Albanna, H., Albashrawi, M. A., Al-Busaidi, A. S., Balakrishnan, J., Barlette, Y., Basu, S., Bose, I., Brooks, L., Buhalis, D., .. Wright, R. (2023). Opinion Paper: “So what if ChatGPT wrote it?” Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policy. International Journal of Information Management, 71, 102642. https://doi.org/10.1016/j.ijinfomgt.2023.102642

125. Kasneci, E., Sessler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U., Groh, G., Günnemann, S., Hüllermeier, E., Krusche, S., Kutyniok, G., Michaeli, T., Nerdel, C., Pfeffer, J., Poquet, O., Sailer, M., Schmidt, A., Seidel, T., .. Kasneci, G. (2023). ChatGPT for good? On opportunities and challenges of large language models for education. Learning and Individual Differences, 103, 102274. https://doi.org/10.1016/j.lindif.2023.102274

126. Cotton, D. R. E., Cotton, P. A., & Shipway, J. R. (2024). Chatting and cheating: Ensuring academic integrity in the era of ChatGPT. Innovations in Education and Teaching International, 61(2), 228–239. https://doi.org/10.1080/14703297.2023.2190148

127. Zhang, K., & Aslan, A. B. (2021). AI technologies for education: Recent research & future directions. Computers and Education: Artificial Intelligence, 2, 100025. https://doi.org/10.1016/j.caeai.2021.100025

128. Chiu, T. K. F., Xia, Q., Zhou, X., Chai, C. S., & Cheng, M. (2023). Systematic literature review on opportunities, challenges, and future research recommendations of artificial intelligence in education. Computers and Education: Artificial Intelligence, 4, 100118. https://doi.org/10.1016/j.caeai.2022.100118

129. Yan, L., Sha, L., Zhao, L., Li, Y., Martinez‐Maldonado, R., Chen, G., Li, X., Jin, Y., & Gašević, D. (2024). Practical and ethical challenges of large language models in education: A systematic scoping review. British Journal of Educational Technology, 55(1), 90–112. https://doi.org/10.1111/bjet.13370

130. Mollick, E. R., & Mollick, L. (2023). Assigning AI: Seven Approaches for Students, with Prompts [Preprint]. SSRN. https://doi.org/10.2139/ssrn.4475995

131. Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2024). Generative AI Can Harm Learning [Preprint]. SSRN. https://doi.org/10.2139/ssrn.4895486

132. UNESCO. (2023). Guidance for generative AI in education and research. UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000386693

133. Jobin, A., Ienca, M., & Vayena, E. (2019). The global landscape of AI ethics guidelines. Nature Machine Intelligence, 1(9), 389–399. https://doi.org/10.1038/s42256-019-0088-2

134. Twenge, J. M., Joiner, T. E., Rogers, M. L., & Martin, G. N. (2018). Increases in Depressive Symptoms, Suicide-Related Outcomes, and Suicide Rates Among U.S. Adolescents After 2010 and Links to Increased New Media Screen Time. Clinical Psychological Science, 6(1), 3–17. https://doi.org/10.1177/2167702617723376

135. Haidt, J. (2024). The anxious generation: How the great rewiring of childhood is causing an epidemic of mental illness. Penguin Press. https://www.anxiousgeneration.com/book

136. Hunt, M. G., Marx, R., Lipson, C., & Young, J. (2018). No More FOMO: Limiting Social Media Decreases Loneliness and Depression. Journal of Social and Clinical Psychology, 37(10), 751–768. https://doi.org/10.1521/jscp.2018.37.10.751

137. Allcott, H., Braghieri, L., Eichmeyer, S., & Gentzkow, M. (2020). The Welfare Effects of Social Media. American Economic Review, 110(3), 629–676. https://doi.org/10.1257/aer.20190658

138. Boers, E., Afzali, M. H., Newton, N., & Conrod, P. (2019). Association of Screen Time and Depression in Adolescence. JAMA Pediatrics, 173(9), 853. https://doi.org/10.1001/jamapediatrics.2019.1759

139. Montag, C., Lachmann, B., Herrlich, M., & Zweig, K. (2019). Addictive Features of Social Media/Messenger Platforms and Freemium Games against the Background of Psychological and Economic Theories. International Journal of Environmental Research and Public Health, 16(14), 2612. https://doi.org/10.3390/ijerph16142612

140. Orben, A., & Przybylski, A. K. (2019). The association between adolescent well-being and digital technology use. Nature Human Behaviour, 3(2), 173–182. https://doi.org/10.1038/s41562-018-0506-1

141. Orben, A., & Przybylski, A. K. (2019). Screens, Teens, and Psychological Well-Being: Evidence From Three Time-Use-Diary Studies. Psychological Science, 30(5), 682–696. https://doi.org/10.1177/0956797619830329

142. Heffer, T., Good, M., Daly, O., MacDonell, E., & Willoughby, T. (2019). The Longitudinal Association Between Social-Media Use and Depressive Symptoms Among Adolescents and Young Adults: An Empirical Reply to Twenge et al. (2018). Clinical Psychological Science, 7(3), 462–470. https://doi.org/10.1177/2167702618812727

143. Vuorre, M., Orben, A., & Przybylski, A. K. (2021). There Is No Evidence That Associations Between Adolescents’ Digital Technology Engagement and Mental Health Problems Have Increased. Clinical Psychological Science, 9(5), 823–835. https://doi.org/10.1177/2167702621994549

144. Odgers, C. L., & Jensen, M. R. (2020). Annual Research Review: Adolescent mental health in the digital age: facts, fears, and future directions. Journal of Child Psychology and Psychiatry, 61(3), 336–348. https://doi.org/10.1111/jcpp.13190

145. Odgers, C. L. (2024). The great rewiring: is social media really behind an epidemic of teenage mental illness? Nature, 628(8006), 29–30. https://doi.org/10.1038/d41586-024-00902-2

146. Valkenburg, P. M., Meier, A., & Beyens, I. (2022). Social media use and its impact on adolescent mental health: An umbrella review of the evidence. Current Opinion in Psychology, 44, 58–68. https://doi.org/10.1016/j.copsyc.2021.08.017

147. Ferguson, C. J. (2025). Do social media experiments prove a link with mental health: A methodological and meta-analytic review. Psychology of Popular Media, 14(2), 201–206. https://doi.org/10.1037/ppm0000541

148. Faverio, M., Anderson, M., & Park, E. (2025, April 22). Teens, social media and mental health. Pew Research Center. https://www.pewresearch.org/internet/2025/04/22/teens-social-media-and-mental-health/

149. Ophir, E., Nass, C., & Wagner, A. D. (2009). Cognitive control in media multitaskers. Proceedings of the National Academy of Sciences, 106(37), 15583–15587. https://doi.org/10.1073/pnas.0903620106

150. Newman, N., Ross Arguedas, A., Robertson, C. T., Nielsen, R. K., & Fletcher, R. (2024). Reuters Institute digital news report 2024. Reuters Institute for the Study of Journalism. https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2024