Abstract
This study investigates how Google’s AI Overviews (AIOs) – a major shift in the search engine landscape – are discursively framed by three key actors: Google, tech journalists, and Search Engine Optimisation (SEO) marketers. Drawing on platform studies and discourse theory, we combine Leximancer concept mapping with Critical Discourse Analysis to examine how these stakeholders construct, legitimise, and contest the meaning of AIOs across a broad corpus of grey literature, media coverage, and corporate blogs. Our findings reveal four dominant thematic frames: Generative AI Technologies and AI-Platform Wars, Reconfiguring Search: Let Google Do the Searching for You, Commercial Implications of AIOs, and AI Overviews – Utopia versus Dystopia. While Google frames AIOs as seamless innovations that enhance the user experience, journalists highlight issues of misinformation and epistemic opacity, and SEO marketers focus on the economic and strategic disruptions to visibility and discoverability. Crucially, our analysis underscores the absence of the user’s voice and a lack of attention to democratic, societal and environmental implications of AI-generated summaries. By analysing the discourse surrounding AIOs, this paper demonstrates that AIOs are not merely a new feature, but rather part of a broader shift in platform power and control over online knowledge. We argue that AIOs represent a key moment in the transformation of search, reshaping how trust, authority, and visibility are defined and challenged in the age of AI.
Introduction
The evolution of search technologies reached a pivotal moment in May 2024 with Google’s public rollout of AI Overviews (AIOs). In under 1 year, AIOs are available to more than a billion users across more than 200 countries (Reid, 2024b; Venkatachary, 2024). AIOs are generative AI summaries that synthesise information from across the web in response to search queries (Google, 2025a, 2025b), appear predominantly at the top of the Search Engine Results Page (SERP), and push down the widely familiar 10 blue links list (Berry, 2025; Williams, 2024). This shift not only marks a technological innovation but also a major inflection point in the evolution of web search; redefining epistemic norms around search, authority, and public knowledge. Google describes this development as a leap forward in information accessibility, promising users quick, coherent answers synthesised from various sources (Google, 2025a; Reid, 2024a). This signals a major shift in how digital platforms mediate public knowledge.
In generating these AI search summaries, Google positions itself not just as a gateway to information but as an epistemic authority. It does so by sharing updates in the form of blogposts, media releases, and key events such as the annual Google I/O event. These narratives are supported and contested by various actors central to the technology and media industry.
This paper examines the discourses and debates surrounding the propagation and adoption of Google’s AIOs. Rather than taking AIOs as a neutral and inevitable technical upgrade, we examine them as a site of discursive construction, where technological innovation is being positioned, legitimised, contested, and branded by key actors. Drawing on a corpus of over 600 English-language articles, during 1 year coverage of AIOs, from Google, news media, and marketing and SEO experts, we ask: How are key industry actors (Google, technology journalists and SEO marketers) discursively positioning Google’s AIOs? We explore the themes that emerge in their responses by combining computational concept mapping using Leximancer with qualitative discourse analysis of representative texts. This mixed-methods approach allows us to identify key patterns and narratives in industry discourse while also attending to the rhetorical and ideological work being performed by specific language choices.
We draw on two strands of scholarship: platform studies and discourse theory to understand how AIOs are being understood and shaped by the key actors. Platform studies allow us to examine the influence of digital platforms in shaping social, political, economic, and cultural aspects in society. Discourse theory highlights the power of language in shaping meaning, understanding, and legitimacy. We build on foundational prior work that understands technology not as a neutral tool, but as embedded within broader socio-discursive processes, particularly as it is narrated, framed, and justified by those who build and profit from it (Gillespie, 2010; Howarth, 2000; Plantin et al., 2018; van Dijck et al., 2018). We argue that AIOs represent more than a technical reconfiguration of search engine functionality; they signal a contested redefinition of what it means to find, evaluate, and trust information online. As the gatekeeping role of the search engine becomes increasingly generative, rather than indexical, platform actors are competing not only to shape technological futures but also to frame the public’s understanding of what those futures entail.
By examining the industry’s discourse, this paper contributes to a growing body of research on platformisation, AI communication, and the politics of search infrastructure. In doing so, we aim to make visible the narrative and strategic work through which tech journalists, SEO marketers, and Google are shaping perceptions of AI-powered information retrieval at a critical moment of transition in the search landscape.
The Google platform
The 1990s marked a paradigm shift from broadcast media to digital web-based media, ushering in a new era for the information industry and society. The web is a paradox: designed to be open and participatory (Berners-Lee, 1999) yet lacking safeguards against domination by a few private entities. At the same time, private companies played a key role in shaping the web to be more usable and accessible, particularly by making information discovery easier through search. Early experiments in the 1990s focused on developing browsers that allowed users to access information more intuitively. This was soon followed by the creation of engaging websites that attracted users with entertainment and novelty. This was happening in the backdrop of platform wars between PC and Mac and adoption of niche information retrieval systems into everyday use by companies like Google and Yahoo (Gillespie, 2010).
With the AI turn, Google Search is now facing new challenges and actively seeking innovative solutions. In the early 2000s Google emerged as the dominant search engine, facing little meaningful competition and remaining largely unchallenged. With rapid innovations in generative AI, leading AI companies such as OpenAI, Anthropic, and Perplexity are offering web search results through conversational AI interfaces. These systems allow users to interact with vast amounts of data in a more intuitive and dynamic way, often bypassing traditional search engines by delivering direct, context-aware answers and offering follow-up questions. Although Google had already integrated large language models like BERT (Devlin and Chang, 2018) to improve information retrieval, increasing competition from emerging players pushed the company to further innovate, leading to the development of AI-powered search summaries. Google introduced Gemini and expanded its capabilities even further with the revolutionary integration of AIOs, an embedded generative summary feature in its search results (ACCC, 2024; Kuzub, 2024; Reid, 2024b; Senecal et al., 2025). Marking a critical transformation in the search landscape.
Understanding this transformation and the current discourse surrounding AIOs requires us to first unpack the key entity, Google, the quintessential digital platform. The term “platform” itself is a strategically ambiguous construct, carefully deployed by major online content providers (Gillespie, 2010). As Gillespie (2010) argues, companies deploy the term “platform” across various semantic areas from technical foundations to rhetorical opportunities, to strategically obscure their control and editorial power. They also increasingly assume roles traditionally occupied by public institutions, as van Dijck et al. (2018) argued.
Google Search, now augmented with AIOs, has become not only a gateway to information but a gatekeeper of truth. This can be understood further through platformisation, which describes platforms “as the dominant infrastructural and economic model of the social web” (Helmond, 2015: 43). Through platformisation, digital platforms become central in the reorganisation of cultural practices and societal perceptions (Poell et al., 2019). Google’s AIOs can be considered as an evolution of platformisation in the AI age. In introducing AIOs, Google furthers the walled garden effect (Plantin et al., 2018; Southern, 2025), where users are less likely to exit Google’s ecosystem, thereby diminishing visibility and revenue for external publishers and content creators. AIOs represent a new form of information governance, further reinforcing Google’s role as a key digital infrastructure. Given that Google operates as a vital infrastructure in the world at large (Plantin et al., 2018), the introduction of AIOs is a step towards enhancing capabilities of this infrastructure, particularly as a universal information tool that facilitate our daily inquiries. It is for these reasons that we turn our attention to how infrastructural logic is operationalised and examine its implications for the governance of information. AIOs are shaped not only through technological design but also through discourse, as market and social relations play a critical role in their adoption and reinforcement. To unpack this discursive dimension, we examine grey literature – including materials produced by Google, commentary from tech journalists, and insights from SEO marketers, who collectively play a significant role in shaping the information ecosystem.
Grey literature as discursive site
Digital platforms actively construct their legitimacy and shape perceptions of innovation through discourse. As Gillespie (2010) argues, platform companies engage in discursive work to define what they are, what they are for, and how they should be judged. The term platform itself is a rhetorical device – signifying openness, neutrality, and opportunity – while simultaneously obscuring mechanisms of control, curation, and monetisation. It allows companies like Google to present a “broadly progressive sales pitch while also eliding the tensions inherent in their service” (Gillespie, 2010: 348) (see, for example, corporate events such as Google I/O). This strategic ambiguity inherent in the term “platform” is crucial for understanding how Google positions AIOs.
Understanding these dynamics requires examining the kinds of texts that tech companies produce. These texts, often referred to as grey literature, include blogs, press releases, and whitepapers, which serve as both communicative and strategic tools. The digital turn at the end of the last century was difficult to understand through traditional media, communication, or critical theory lenses, as it was both novel and largely driven by technological developments. In the absence of empirical data from platforms or open-source technical documentation, grey literature offered critical insights into how platform companies shape both the technology and its adoption. Humanities scholars, therefore, began to act as internet historians, excavating blogs, trade publications, and online communities to understand the emergence and evolution of digital platforms. For instance, Driscoll (2022) examined text from and about Bulletin Board Systems (BBS) to find traces of social practices of early digital systems that formed foundations for social media platforms. Similarly, by analysing popular websites of the 1990s, Ankerson (2018) argues that these websites shaped the discourse of “cool” as a strategy in shaping the usability of the web for the general public. In biographical accounts of major platforms like Facebook, YouTube, and Twitter, scholars such Burgess and Green (2018), Burgess and Baym (2020), and Bucher (2021) have drawn on a wide range of materials – such as investor presentations, media interviews with founders, and blog posts – to understand the discourse these platforms sought to shape at the time. These studies position grey literature as a rich archive, offering insight into how technological systems are made legible, legitimate, and desirable.
Crucially, grey literature serves not just as a record of industry practice but also as a discursive tool of anticipation. These texts serve as tools for tech companies to create and reinforce trends. And as Powers (2019) argues, trends are crucial for tech companies to “shape perception and compel action” (Powers, 2019: 26). Thus, grey literature offers valuable insight into the stories that companies tell about their technologies – stories that seek to normalise, naturalise, and justify infrastructural and technological change. In doing so, they perform ideological work: defining what counts as innovation, who benefits from it, and how new systems should be evaluated. As this paper argues, examining grey literature allows us to critically engage with the discursive practices of AI-powered platforms like Google’s AIOs.
Media logic, platform communication and discourse theory
To further understand how Google, tech journalists, and SEO marketers construct and disseminate narratives around AIOs, it is critical to consider the role of media logic. Originating from the work of Altheide and Snow (1979), media logic refers to the formats, production practices, and communicative conventions that shape how information is structured and circulated in contemporary media environments. The media logic of a given medium not only shapes how information is structured and conveyed but also influences how individuals and institutions engage with and respond to it. These logics are not neutral; rather, they influence how institutions package their messages to be compatible with prevailing media expectations, prioritising simplicity, emotional appeal, and immediacy over nuance, ambiguity, or critique (Altheide, 2004; Altheide and Snow, 1979).
In the context of digital platforms, media logic theory provides a lens through which we can understand how digital platforms like Google, along with tech journalists and marketers, use specific communication styles and strategies to shape their messages. By systematically examining texts published on corporate, news, and marketing blogs over time, we can identify themes and rhetorical strategies that reflect each actor’s underlying values, assumptions, and perspectives. This offers insight into how these actors frame the discourse surrounding AIOs, revealing the orientations and ideologies that guide their discourse decisions. Thus, to understand the mechanisms shaping public narratives about AIOs, it is essential to examine the values and discursive practices of influential communicators within each domain, a procedure that aligns with previous work on this domain (Altheide, 2002; Hodson, 2013).
Moreover, media logic helps explain which critiques are amplified and which are marginalised in the discourse. This dynamic contributes to what Howarth (2000) and Fairclough (1995) identify as discursive exclusions, where the absence of critique is not accidental, but structurally conditioned by how media content is curated and legitimised. Thus, media logic theory complements the view of platforms as discursive actors (Gillespie, 2010; Hurcombe et al., 2025), offering insight into how their communication tools and styles influence how the public understands AIOs.
To deepen this analysis of communicative power and ideological framing, we also integrate discourse theory, specifically that of Howarth (2000). Building on Howarth’s (2000) interpretation of post-structuralist discourse theory, we approach the discourse surrounding Google’s AIOs as a hegemonic formation – one that organises meaning through both articulation and exclusion. Howarth, drawing from Laclau and Mouffe (1985), emphasises that discourse is not merely a neutral system of signs, but a field structured by difference, where the stability of meaning depends on what is included and, crucially, what is left out.
Finally, we situate AIOs within AI communication studies given how these discursive constructions around AIOs can be understood as part of broader communicative patterns in the public imagination of AI. Scholars such as Crawford (2021) and Johnson and Verdicchio (2017) emphasise that AI narratives oscillate between utopian and dystopian imaginaries. These stories perform crucial social functions: they justify technological adoption, attract investment, and influence user expectations. These sociotechnical imaginaries (Jasanoff and Kim, 2009, 2015), or AI Imaginaries (Crawford, 2021), play a powerful role in shaping not only product reception but also regulatory and ethical discourse.
Methods
We employ a mixed-methods discourse analytic approach, combining Leximancer-assisted concept mapping with qualitative discourse analysis. Our goal is to trace how Google’s AIOs and the broader transformation of search are discursively constructed by industry actors, particularly through public commentary, blogs, press releases, and articles published by tech vendors, media outlets, and industry stakeholders. We analyse the discourse about AIOs from three different lenses: Google, tech journalism, and marketing and SEO, each representing a unique narrative.
Data collection
Using a custom web scraper, we automatically 1 collected 613 English-language texts published between May 2024 and May 2025, a period that captures the emergence, rollout, and early reception of Google’s AI Overviews. Texts were collected through a combination of Google Search and domain-specific automatic scraping that appeared for the search term: “‘google AI overview’ ‘AI overview’ AND google ‘google’s AI overview’ OR ‘ai overviews’”. The corpus includes official blog posts and press releases from Google, articles and reports from reputable technology news outlets (e.g. Wired, TechCrunch, The Verge, Ars Technica), and thought pieces, whitepapers, and commentary by SEO experts and digital marketers.
Since our focus was on understanding how different actors discursively constructed AIOs, based on the domain of each URL (e.g. www.cnet.com → cnet), we classified the 613 documents into one of three actors: journalism (253), marketing and SEO (330), and Google (30).
Conceptual mapping with Leximancer
As a first step, to identify key discursive patterns in the broader industry conversation surrounding Google’s AI Overviews, we used Leximancer (Smith and Humphreys, 2006), a concept mapping software that uses a Bayes-inspired algorithm to detect and visualise major themes and concepts in large textual datasets. Leximancer works by analysing not only word frequency but also patterns of co-occurrence, constructing probabilistic semantic models that estimate how often and in what contexts specific terms appear together. These repeated co-occurrences are interpreted as indicators of conceptual connectivity, which the software uses to generate a structured visual map and ranked list of concepts, offering both quantitative insight and an intuitive representation of the discourse landscape. The concept map clusters related terms into themes, which are represented spatially, closer proximity indicates stronger semantic relationships. This allows researchers to quickly identify key concepts, their prominence, and how they relate to one another across a corpus. The interactive software dashboard facilitates in-depth analysis of concepts connected with key quotes from any source material. Leximancer can work with a range of text document formats, for this study we organised the data into csv files where each row contained the title, source, body text, author, and category (Google, journalist, marketing and SEO).
This approach and technique have proven effective in comparable discourse analyses, such as studies of framing of problematic content by Meta’s newsroom (Hurcombe et al., 2025) and comparative news framings of the COVID-19 pandemic (Ritchart et al., 2024). Leximancer is particularly suited to the scale and nature of our dataset, which includes hundreds of technology news and industry articles discussing AI-powered search features. Given the heterogeneity and volume of texts, Leximancer allowed for an automated and systematic extraction of dominant themes and conceptual clusters without imposing predefined categories, preserving the inductive integrity of the analysis. This facilitated the identification of dominant vendor narratives and industry framings, which were then examined in greater depth through qualitative discourse analysis. The output concept map provides a foundational overview of how vendors, industry commentators, and media outlets frame the emergence of AIOs, including recurrent terms and their relationships.
Discourse analysis
In the second step, we conducted a qualitative discourse analysis of selected texts representing dominant themes and discursive patterns identified through Leximancer. Inspired by critical discourse analysis (Fairclough, 1995; van Leeuwen, 2008), a robust methodological framework that views language as a social practice that reproduces power and ideology, we examine how actors frame AIOs, position themselves in relation to Google’s discourse, and mobilise specific rhetorical strategies. Fairclough’s (1995) three-level model – text, discourse practice, and sociocultural context – served as a tool to examine how different actors construct their discourses about AIOs and shape user perceptions.
Findings
The discourse surrounding Google’s AIOs is characterised by a complex interplay of competing stakeholder narratives. Google’s discourse adopts a strategically optimistic tone. In contrast, SEO marketers express a blend of excitement and anxiety, while journalists articulate a more critical stance. This section unpacks these discourses by firstly mapping their sentiment profiles and conceptual associations (see Tables 1 and 2; Figure 1), and then tracing how these patterns surface in four interrelated thematic framings: (1) Generative AI Technologies and AI-Platform Wars, (2) Reconfiguring Search – Let Google Do the Searching for You, (3) Commercial Implications of AIOs, and (4) AI Overviews – Utopia versus Dystopia.
Sentiment distribution by actor group.
Top 10 concepts by actor.
Note.The top 10 concepts by actor are determined by a weighted score = count × likelihood/100
Thematic map by actor.
Sentiments and conceptual profiles of the discourse
Before presenting the sentiment analysis, it is important to clarify how Leximancer structures its data for processing. Rather than analysing entire articles at a time, Leximancer pre-processes the dataset into context blocks, units typically comprising two to three sentences. These context blocks form the basis for its co-occurrence statistics and sentiment coding. As such, the sentiment counts reported in Table 1 reflect the number of context blocks coded with either favourable or unfavourable sentiment, rather than the number of articles. This distinction matters because articles can vary significantly in length and tone, whereas context blocks are more uniform and provide a more granular and consistent unit of analysis. Reporting counts at the context block level thus offers a clearer view of the distribution and density of sentiment across the dataset. Where relevant, we report the total number of context blocks associated with each category, as this provides a stronger basis for comparison than article count alone. Similarly, likelihood is a conditional probability measure that approximates the likelihood that any context block from category A contains a reference to concept/sentiment B.
Google’s own discourse is largely neutral or reserved, with a slight leaning towards positive sentiment (164 context blocks coded with favourable sentiment with 5% likelihood versus 46 context blocks coded with unfavourable sentiment with 4% likelihood). This is typical of official or strategic communication that aims to maintain authority and avoid controversy.
SEO marketers oscillate between positive yet conflicted sentiment (2038 context blocks coded with favourable sentiment with 66% likelihood versus 679 context blocks coded with unfavourable sentiment with 52% likelihood). This group shows high engagement and views AIOs with both excitement (opportunities) and concern (threats to traffic, control). The strong positive tone reflects optimism about adapting to or capitalising on technology, while the negative tone suggests strategic unease or loss of organic visibility.
Journalists’ discourse reveals a cautiously critical sentiment (522 context blocks coded with unfavourable sentiment with 40% likelihood versus 691 context blocks coded with favourable sentiment with 22% likelihood). While some coverage is positive; likely reporting on the innovation of this new technology, the overall tilt is sceptical, reflecting concerns over epistemic authority, information gatekeeping, and disruption to news visibility. In the next section we present the four interrelated thematic framings.
Generative AI technologies and the AI-platform wars
The introduction of AIOs is situated within a broader competitive landscape. A dominant theme about Google’s AIOs pertains to the ongoing race between AI-powered platforms and the technologies behind them. The rise of AIOs is inseparable from the broader shift toward LLM-powered platforms. Names like Gemini, ChatGPT, and terms like generative, model, language, and systems anchor this theme, indicating a heavy focus on LLMs and generative AI systems as the backbone of this transformation. Google’s Gemini and models like ChatGPT are situated in discussions about (1) competition and (2) capability and application.
The AI-platform wars and intensifying competition between companies are central to the current discourse. Industry observers argue that AI has become so intertwined with search that it is now impossible to consider one without the other. This growing dependence on AI has sparked concerns about how Google’s existing dominance in search might be leveraged to secure unfair advantages in newer, rapidly evolving AI markets. These tensions are further illustrated by challenges faced by competitors like OpenAI, who struggled to build their own search index due to the high costs and complexity involved. As a result, OpenAI sought partnerships with existing search providers, even approaching Google – who ultimately declined, viewing OpenAI as a competitive threat (see for example Diaz, 2025).
Importantly, there is a clear techno-discursive tension between how Google presents the potential of its AI and how other actors describe its real-world performance. Google, for example, highlights the advanced abilities of its system, with Head of Search Reid (2024b) stating: “This is all made possible by a new Gemini model customized for Google Search. It brings together Gemini’s advanced capabilities, including multi-step reasoning, planning and multimodality with our best-in-class Search systems”. In contrast, SEO professionals point out that the technology still has limitations. As Sanger (2024) explains: “Even with these improvements, it still faces challenges in these areas. Why is this important? Understanding how LLMs and Google Gemini function is necessary to understand how to improve your contents visibility in AI Overviews”.
These representative quotes demonstrate that the framing around AI is deeply entangled in the escalating competition among AI-powered platforms, while the technology that drives those platforms is framed not only as a product of technological advancement but as a strategic instrument in a broader platform race, where models like Gemini and ChatGPT symbolise both capability and competitive edge. Google positions its Gemini model as a cutting-edge innovation seamlessly integrated into its search infrastructure. Meanwhile, SEO professionals and journalists highlight the frictions between AI’s promises and its practical limitations, pointing to ongoing challenges.
Reconfiguring search: Let Google do the searching for you
The discourse surrounding Google’s AIOs also focuses on how they are reshaping the search experience, replacing traditional organic links with AI-generated answers, and changing the manner and way in which information is retrieved. Google is no longer just indexing web pages – it is now directly responding to queries. This signals a significant shift in how knowledge is produced and delivered online. Rather than simply presenting links, AIOs generate summarised responses, shifting the role of search engines from gatekeepers to authors of information. Terms like queries, answers, summaries, and AI-generated content dominate this narrative, pointing to a redefinition of search as a curated and constructed knowledge system.
This transformation highlights a broader tension between how different actors – Google, tech journalists, and SEO marketers – talk about AIOs. Google frames AIOs as a seamless upgrade that improves user experience, using reassuring language like ease, help, and better experience. As the company explains, AIOs are designed to “be helpful, and [. . .] prioritize quality and safety” (Google, 2025b). Reid (2024b), Head of Search at Google, reinforces this optimistic tone by stating, “Now, with generative AI, Search can do more than you ever imagined. So you can ask whatever’s on your mind or whatever you need to get done [. . .] and Google will take care of the legwork”.
However, other actors frame the feature in more ambivalent or critical terms. Journalists, for instance, caution that AIOs may discourage users from engaging with the underlying sources, thereby undermining trust in the accuracy and legitimacy of the information. Similarly, marketing and SEO professionals express concern over the redistribution of attention and value within the search ecosystem. Their comments mention features, such as “stealing attention from both organic and paid results” (Awkward-Media, SEO Company, n.d.).
These quotes highlight the contested framing of Google’s AIOs across different actors. Google portrays AIOs as a user-friendly innovation that simplifies search through AI-generated assistance, legitimising the shift with language centred on help, quality, and seamless experience. In contrast, journalists express concern about the erosion of source engagement and the implications for informational trust, while SEO marketers critique the displacement of organic and paid content, warning of a redistribution of visibility and value within the search ecosystem. In this way, the transformation of search is not just epistemological, but also deeply commercial, which leads us to the next dominant theme.
Commercial implications of AIOs
The rise of AIOs has fundamentally challenged norms, business models, and SEO ecosystems, triggering concerns about the visibility of publishers and businesses. SEO actors are particularly vocal, expressing both strategic anxiety and opportunism. This reflects broader tensions between algorithmic governance and economic survival in the platform economy.
Thus, this theme is dominated by actors trying to understand, predict, or exploit changes in visibility. Terms like SEO, site, traffic, visibility, organic, ads, keywords, brand, businesses, and publishers suggest a discourse of optimisation under uncertainty, and highlight AIOs not just as tools but as gatekeepers of discoverability. The stakes here are commercial and existential: How do you remain visible when search becomes generative, not referential? As one SEO commentator notes: “if you’ve noticed your website traffic has dropped dramatically since 2023 or 2024, this shift from featured snippets to AI Overviews is likely the culprit. Google has changed the game, but there are still ways to adapt” (Jamieson, 2025).
Indeed, many in the industry are already pivoting toward new strategies like Generative Engine Optimization (GEO) to sustain and improve their online visibility (Dirks, 2025; Banik, n.d.). As AIOs begin to reshape how search results are displayed, businesses are being forced to rethink their SEO and digital marketing approaches. This shift, however, has been met with considerable uncertainty and frustration. One media executive noted that when AIOs appear, their website’s average clickthrough rates dropped sharply – over 50% on desktop and nearly as much on mobile. Even when their link appeared at the top of the AIO, the decline in user engagement remained significant, highlighting just how disruptive this new format can be to traditional traffic flows (Tobitt, 2025).
This signals a massive redistribution of attention, leaving brands with less control over how their content is displayed and interpreted, making it crucial to understand “how AI Overviews function and implementing strategies to enhance brand visibility within this ecosystem is now necessary for digital marketing” as mentioned by one marketing company (Brussow, 2025). Importantly, while SEO and marketing communities are trying to pave their way in this new ecosystem and adapt swiftly to the new norms, Google has tended to frame these concerns in a positive light or otherwise marginalise them. For example, Google argues that: “when people click from search result pages with AI Overviews, these clicks are higher quality for websites [. . .] And businesses get a chance to connect with those users in a more meaningful way, with ads that deeply match their needs” (Think with Google, 2024).
The integration of AIOs into Google Search has triggered profound shifts in the SEO and digital marketing landscape, reshaping how visibility, value, and control are distributed. While SEO professionals scramble to adapt through strategies like GEO and many express concerns over declining traffic, reduced control over content display, and an uncertain future in a generative-first search environment, Google maintains a largely optimistic framing, emphasising higher quality user engagement and encouraging content creators to align with AI systems. This tension underscores the commercial stakes and emerging power asymmetries within a platformised search economy.
AI Overviews – Utopia versus Dystopia
Finally, as “AI Overviews move the burden of processing search results from human users to AI service” (Brussow, 2025), a clear narrative emerges from the discourse: one that oscillates between utopian promise and dystopian concern.
At one end lies Google’s vision, saturated with terms like helpful, supersmart, superfast, and satisfied. As mentioned by Brendon Kraham, VP of search and commerce, global ads solutions: “I think of AI Overviews as a supersmart, superfast librarian who has read everything on the web. Users are loving it. [. . .] we’ve found that people who use AI Overviews search more and are more satisfied with their results; they visit a greater diversity of websites for help with more complex questions; and we even see higher engagement from younger users aged 18 to 24 [. . .] It’s all about making search smarter and your ads more helpful” (Think with Google, 2024). This quote not only reflects Google’s optimistic framing of generative AI, but also exemplifies the platformisation of knowledge and the ongoing Googlization of everything. Google presents AIOs as the best solution to search fatigue and informational legwork (Reid, 2024b), positioning the feature as an intelligent assistant capable of enhancing both user satisfaction and ad relevance.
At the other end, journalists and SEO marketers construct a more critical discourse, invoking language such as disaster, problem, mistakes, and wrong. They reveal a more sceptical and cautionary perspective, rooted in the errors and hallucinations of generative AI. Some AIOs claimed that “astronauts have met cats on the moon,” while others alarmingly recommended that “you should eat at least one small rock per day,” citing nutritional benefits. These highlight that “One fundamental problem is that generative AI tools don’t know what is true, just what is popular” (Walsh, 2024). Other errors include advice to put glue on pizza or claims that dogs play in the NBA and cats lick hands to check if humans are edible (Vjestica, 2024). As one journalistic headline bluntly put it, “Google AI Overviews are still a disaster unless you like glue on your pizza” (Smith, 2024). These examples underscore a deeper issue: generative AI does not determine truth, it reproduces what appears plausible or popular online. This has led many to question not just the accuracy of the outputs, but the underlying sources and editorial logic shaping them.
Moreover, these examples underscore a fundamental tension between Google’s claims to intelligence and the actual epistemic risks of delegating knowledge production to AI. Google’s official response to such inaccuracies appears measured, but somewhat dismissive, as quoted:”Google spokesperson Meghann Farnsworth said the mistakes appear from ‘generally very uncommon queries, and aren’t representative of most people’s experiences [. . .] Google also includes a label that its ‘Generative AI is experimental’ at the bottom of any given answer” (Vjestica, 2024).
Putting it all together, these quotes reveal that, while Google couches this paradigmatic shift in terms of a better search experience with unprecedented convenience, other stakeholders foreground the risks, challenges and implications for epistemic authority, content visibility, and the economy of attention. Their discourse highlights the broader systemic consequences of AIOs, consequences that are downplayed or entirely omitted from Google’s own messaging.
Discussion
This study set out to examine how the emergence of AI-powered search – specifically Google’s AIOs – is discursively constructed by three key actors: Google, technology journalists, and SEO marketers. Our findings reveal somewhat unsurprisingly that each actor frames AIOs through distinct lenses, shaped by their institutional interests and roles within the digital ecosystem. These framings are not random but reflect deeper motivations tied to authority, visibility, and adaptation within an evolving information landscape. From our analysis, we identified four dominant themes that structure the discourse around AIOs: Generative AI Technologies and AI-Platform Wars; Reconfiguring Search – Let Google Do the Searching for You; Commercial Implications of AIOs; and, AI Overviews – Utopia versus Dystopia.
Google strategically framed AIOs as a seamless upgrade to the search experience – what it describes as a “supersmart, superfast librarian” (Think with Google Editorial Team, 2024). This metaphor encapsulates the company’s shift from indexing information to curating and synthesising it, echoing what Plantin et al. (2018) term the “platformisation of knowledge.”
To reinforce this positioning, Google relies on emotionally and efficiency-driven language such as “helpful,” “quick,” and “efficient.” These terms reflect what Jasanoff and Kim (2009) define as sociotechnical imaginaries: visions of progress that legitimise and normalise emerging technologies. Yet, this language is tempered by rhetorical hedging. By labelling AIOs “experimental,” Google constructs a dual narrative, one that advances the product’s authority while shielding it from accountability. Such strategy falls in line with Gillespie’s (2010) argument that platforms present themselves as open and neutral while obfuscating their editorial and economic influence.
These strategic choices resonate with dominant media logics (Altheide, 2004). Google’s public messaging around AIOs draws on values such as speed, convenience, and personalisation, with terms like “powerful,” “instant,” and “let Google do the work for you”. These framings encourage trust and engagement while downplaying concerns such as misinformation, bias, or the erosion of source diversity. In doing so, Google not only builds trust in AIOs but also sustains its epistemic authority in an evolving information ecosystem (Altheide, 2004; Nguyen and Hekman, 2024). While Vaidhyanathan (2012) originally introduced the term Googlization of knowledge in the context of book digitisation, AIOs extend the concept to the web more broadly. These tools reshape how people search, interpret, and trust information – redefining the public’s relationship to search engines and digital knowledge.
In contrast, journalists and SEO marketers adopt more critical framings. Journalists highlight risks such as misinformation, source invisibility, and flawed AI outputs – ranging from recommending glue as pizza topping to suggesting that cats live on the moon. These errors, while anecdotal, underscore broader concerns about the reliability and accountability of generative systems. SEO professionals express both concern and adaptation: they report declining site visibility and engagement, prompting shifts toward Generative Engine Optimization (GEO). This signals a commercial recalibration in response to the structural changes introduced by AIOs.
Importantly, the discursive field is shaped as much by what is present as by what is absent (Howarth, 2000). Aside from Google’s generalised claims that users are happier with results, the voice of the user is largely missing. Rather than active participants, users are cast as passive recipients of algorithmically curated answers. There is little reflection on how users make sense of AI-generated content, how they evaluate credibility, or how their epistemic habits may shift in response to AI’s increasing role in search. Also, notably absent are discussions of the environmental costs of training large language models – raised only in passing in two texts – as well as deeper engagement with the ethical implications of AI deployment.
Building on the previous discussion of how key actors frame AIOs, we now turn to the broader implications of Google’s positioning – specifically, how AIOs illustrate the platform’s hybrid identity as both a commercial entity and information infrastructure. Plantin et al. (2018) argue that influential digital platforms like Google now seem to function as vital infrastructures in the world at large. According to Plantin et al. (2018), modern digital systems such as the search giant, Google, blur the lines between “platforms” – modular, profit-driven systems – and “infrastructures” – public-facing, standardised backbones of daily life. Google’s AIOs sit at the intersection of platforms and infrastructures; while built within Google’s proprietary, commercial ecosystem (a platform logic), they are discursively positioned as universal information tools that facilitate our daily inquiries (an infrastructural logic). This hybridity allows Google to benefit from the trust traditionally afforded to infrastructure while maintaining control typical of platforms. It reflects broader processes of platformisation (Poell et al., 2019), through which Google operates simultaneously as a profit-driven platform and a de facto public infrastructure.
This interplay of language, power, and platform logic underscores Google’s dual role – not only as a dominant search engine but also as a communicative actor. The launch of AIOs is not simply a technical evolution, but a discursive project aimed at shaping public understanding. As the company increasingly replaces organic links with machine-generated summaries (Robison, 2025; Wolfenstein, 2025), it raises fundamental epistemological questions: who gets to speak, what counts as authoritative knowledge, and how public understanding is shaped. These concerns are further amplified by the Googlization of knowledge, in which the company assumes both infrastructural and editorial power.
This study has several limitations. First, by focusing solely on Google, we are constrained in our ability to fully capture the emerging competition among AI platforms and how other AI vendors are framing Google and its new features. Second, although we have collected data for a 1-year coverage of AIOs, which captured the emergence, rollout, and early reception of AIOs, there is an ongoing and evolving discourse around this feature, as such our findings may be somewhat temporally bounded. Third, although our analysis concentrated on Google’s AIOs, which is a unique phenomenon in the search landscape, Google continues to develop its search features and has recently introduced “AI Mode” (Stein, 2025). The discourse surrounding this new feature may also reference AIOs and examining this could provide a broader understanding of the evolving search landscape. Fourth, while Leximancer was valuable in identifying dominant concepts and thematic clusters, it has limitations in capturing irony, nuance, and contested meanings that do not appear in high-frequency patterns. Finally, our discourse analysis emphasises representational strategies rather than user reception outcomes, so we cannot conclude how users experience or respond to AIOs in practice based on this data.
These limitations point to directions for future research. There is an urgent need for user-centred studies that explore how people encounter and evaluate AI-generated search results. What epistemic habits are being encouraged or eroded? How do users distinguish credible from spurious information when it is pre-packaged by an AI system? And what does it mean for a search engine to “answer” rather than “refer”? Additionally, comparative studies across vendors, such as Google, OpenAI, Microsoft, and emerging players like Perplexity, could illuminate how different corporate actors frame AI summarisation.
Conclusion
Google’s AIOs signal more than a technical upgrade to search – they represent a transformation in how knowledge is curated, delivered, and legitimised in the digital era. Through strategic framings grounded in dominant media logics, Google positions AIOs as intuitive, seamless tools that meet user expectations, while masking deeper shifts in information control and editorial authority. This interaction between platform strategy and communicative conventions explains how particular narratives about AIOs come to dominate public discourse, reinforcing Google’s epistemic power.
At the same time, AIOs fundamentally alter the relationships between users, content creators, and platforms. By synthesising information from multiple sources while heavily curating the output, AIOs reconfigure how information flows and how market relations are structured (ACCC, 2024; Poell et al., 2019). As Google increasingly acts as both infrastructure and editor, critical questions arise about transparency, accountability, and the politics of platform-driven knowledge infrastructures.
Ultimately, the contrasting narratives, one of which positions, in a utopian manner, AIOs as “the next big thing in search”, and the other of which positions it as a “disaster”, underscore the need for a comprehensive investigation of this new and revolutionary technology. AIOs are not simply technical enhancements to search but also represent a continuation of broader platformisation logics and discursive and infrastructural interventions that reshape what it means to search, to know, and to trust in the digital age. Understanding these transformations requires not only examining the technologies we use but also interrogating the stories we are told about them, and who gets to tell them.
Footnotes
Ethical considerations
This study utilises publicly available data and does not involve human subjects. In line with the National Health and Medical Research Council’s National Statement on Ethical Conduct in Human Research (2023), the research is considered low risk and does not require formal ethical review.
Consent to participate
As the study is based solely on publicly available news articles and does not involve human participants or identifiable personal data, consent to participate is not applicable.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Australian Research Council (ARC) for the ARC Centre of Excellence for Automated Decision-Making and Society (CE200100005).
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data availability statement
This research utilised publicly available data. The titles or URLs of all news articles can be made available upon request. However, due to copyright constraints, we cannot reproduce, host or share the complete article texts.