How AI is transforming online platforms from gateways into intelligent intermediaries that increasingly shape discovery, recommendations, commerce and digital experiences.
Artificial intelligence is changing how people discover information online. As search engines, social platforms and commerce systems become increasingly AI-mediated, publishers and brands face a new challenge: remaining visible when algorithms and AI agents increasingly decide what audiences see, read and buy. The organisations that succeed will be those that combine expertise, trust and direct audience relationships with the productivity advantages of AI.
Artificial Intelligence Has Become the Operating System of Online Platforms
How AI Is Rewiring Content Discovery Across the Internet
Generative AI Is Transforming Platform Economics
The Publisher Crisis: AI Search and the Rise of Zero-Click Discovery
AI Creates New Risks Alongside New Opportunities
Why Regulation and Digital Provenance Are Becoming Strategic Assets
The Next Platform Era: Agentic AI and the Internet Beyond Search
The defining characteristic of the modern web is no longer search. It is mediation.
For more than two decades, online platforms acted primarily as gateways. Search engines indexed information, social networks distributed content, and e-commerce platforms connected buyers with products. The user remained the decision-maker. Platforms organised information, but people ultimately determined what to click, read, watch or buy.
That model is changing.
Artificial intelligence is increasingly becoming the intermediary between users and the digital world itself. Rather than presenting a list of options, platforms are beginning to interpret intent, make recommendations, generate responses, prioritise information and, in some cases, take action on a user's behalf.
This shift is visible across every major platform. Google is transforming search through AI Mode and AI-generated answers that synthesise information rather than simply linking to it. OpenAI has moved beyond conversational interfaces towards systems capable of research, planning and task execution. Microsoft is embedding AI assistants across productivity workflows, while Meta is integrating AI into content discovery, advertising, messaging and commerce. Amazon continues to expand AI-driven personalisation throughout its retail ecosystem, influencing everything from product discovery to pricing and fulfilment.
The result is a fundamental change in how information flows across the internet. Increasingly, users are interacting not with websites, publishers, brands or products directly, but with AI systems that interpret and curate those interactions on their behalf.
For publishers, marketers and platform operators, this represents a profound strategic shift. Visibility is no longer determined solely by rankings, followers or advertising budgets. It is increasingly determined by whether AI systems choose to surface, recommend or reference your content.
Artificial intelligence is increasingly becoming the intermediary between users and the digital world itself.
Previous waves of digital transformation changed how information was distributed. Artificial intelligence is changing how information is processed.
That distinction matters.
Recommendation engines, machine learning models and automation tools have existed for years. What makes the current moment different is the convergence of large language models, generative AI and foundation model infrastructure at global scale. AI is no longer operating at the edges of online platforms; it is becoming embedded within their core operating systems.
The scale of investment reflects this reality. Google, Meta, Microsoft, Amazon and OpenAI are committing tens of billions of dollars annually to AI infrastructure, model development and data centre capacity. These are not experimental initiatives. They are strategic bets on the future architecture of the internet.
The business response has been equally significant. Research from the World Economic Forum, McKinsey Global Institute and Deloitte shows that AI adoption has moved rapidly from experimentation to deployment. Generative AI is now being embedded into customer service, content operations, software development, marketing workflows and decision-making processes across industries. What began as a productivity tool is evolving into a new layer of digital infrastructure.
This explains why forecasts such as PwC's estimate of a $15.7 trillion contribution to the global economy by 2030 continue to attract attention. The value does not arise solely from automating individual tasks. It comes from redesigning entire systems around AI-mediated decision making.
The most important implication for business leaders is that artificial intelligence should no longer be viewed as a standalone technology initiative. It is becoming the operating system through which information is discovered, products are evaluated and commercial relationships are formed.
Artificial intelligence should no longer be viewed as a standalone technology initiative. It is becoming the operating system through which information is discovered, products are evaluated and commercial relationships are formed.
The organisations that understand this shift earliest will be best positioned to compete in a digital environment where AI increasingly determines what people see, what they trust and ultimately what they choose.
The most successful online platforms no longer compete primarily on content, products or features. They compete on prediction.
Over the past decade, digital discovery has undergone a profound transformation. Search engines once dominated the process by helping users find information they actively sought. Today's largest platforms increasingly rely on recommendation systems that surface information before users know they want it.
This distinction matters because recommendation systems fundamentally alter the economics of attention.
Instead of responding to explicit intent, platforms now analyse billions of behavioural signals—clicks, dwell time, watch history, scrolling patterns, purchases and social interactions—to predict what a user is most likely to engage with next. Every interaction becomes training data. Every recommendation becomes a prediction.
The result is a shift from search-driven discovery to algorithm-driven discovery.
This model now underpins much of the modern internet. Whether a user is watching a film, listening to a podcast, shopping for trainers or scrolling through social media, artificial intelligence increasingly determines what appears next.
For businesses, this creates a new competitive reality. Success is no longer determined solely by producing high-quality content or products. It depends on whether platform algorithms judge that content worthy of distribution.
In practical terms, recommendation engines have become the primary gatekeepers of digital attention.
The most successful online platforms no longer compete primarily on content. They compete on prediction.
The clearest evidence of this shift can be seen in the world's most successful consumer platforms.
Netflix is often described as a streaming business, but its competitive advantage is arguably its recommendation engine. The company has long reported that approximately 80 per cent of viewing activity is influenced by recommendations. Its ability to match audiences with content reduces churn, increases viewing time and improves return on content investment. In a catalogue containing thousands of titles, discovery has become as important as production.
Spotify operates according to a similar principle. Features such as Discover Weekly, Daily Mix and personalised recommendations are powered by a combination of collaborative filtering, content-based filtering and natural language processing. Spotify has reported that Discover Weekly alone generated billions of track streams within its first years of operation, demonstrating the commercial value of algorithmic discovery at scale.
TikTok accelerated the trend further by placing recommendation systems at the centre of the user experience. Unlike earlier social networks that relied heavily on follower graphs, TikTok's For You feed prioritises predicted relevance over social relationships. Content from an unknown creator can reach millions of users if engagement signals indicate strong audience resonance. This approach has produced some of the highest engagement rates in social media and fundamentally changed expectations around content distribution.
YouTube, meanwhile, has refined recommendation systems for more than a decade. The platform has repeatedly stated that recommendations drive the majority of viewing activity, with billions of hours of watch time influenced by machine-learning models designed to maximise viewer satisfaction and retention.
Across these platforms, the strategic pattern is remarkably consistent. The winners are not necessarily those with the largest content libraries or the biggest creator communities. They are the organisations most capable of predicting attention.
The winners are not necessarily those with the largest content libraries. They are the organisations most capable of predicting attention.
Netflix provides perhaps the most important lesson for publishers and marketers.
Historically, media companies created value by producing or acquiring content. Netflix demonstrated that discovery itself could become a source of competitive advantage.
By combining collaborative filtering, behavioural analytics and increasingly sophisticated machine learning systems, Netflix transformed recommendation from a convenience feature into core business infrastructure. The platform's ability to connect viewers with relevant content reduces decision fatigue, increases session duration and strengthens customer retention.
The strategic insight extends far beyond streaming. In an environment of infinite content, the ability to help audiences discover relevance becomes more valuable than simply creating more content.
In an environment of infinite content, the ability to help audiences discover relevance becomes more valuable than simply creating more content.
The broader implication is that chronological distribution is becoming obsolete.
For much of the internet's history, content was organised by time. Websites published articles in sequence. Social feeds displayed posts in order. Audiences navigated information chronologically.
Today, relevance increasingly outweighs recency.
Algorithms determine which videos appear in a feed, which products appear in a marketplace, which articles appear in search results and which creators gain visibility. Distribution is no longer governed primarily by publishing schedules but by predictive models.
For publishers, agencies and brands, this represents one of the most consequential shifts in digital strategy. The challenge is no longer simply producing content that audiences value. It is producing content that recommendation systems recognise as valuable to audiences.
That distinction will define the next era of online competition. As artificial intelligence becomes the primary mechanism for content discovery, the organisations that thrive will be those that understand a simple reality: attention is no longer distributed. It is predicted.
Attention is no longer distributed. It is predicted.
Recommendation systems changed how content was distributed. Generative AI is changing how content is produced.
For most of the internet's history, digital platforms scaled by adding users, creators, products or publishers. Growth was constrained by human capacity. More content required more writers. More customer support required more staff. More product descriptions required more merchandising resources.
Generative AI changes that equation.
Today, platforms can generate product copy, customer service responses, marketing assets, code, summaries, translations and creative variations at a scale that would have been economically impossible only a few years ago. What was previously labour-intensive has become software-driven.
This is not simply a productivity story. It is a structural shift in platform economics.
The most significant impact of generative AI is not that individual employees work faster. It is that platforms can increase output without increasing operating costs at the same rate. The relationship between scale and headcount is beginning to change.
That shift is already visible across enterprise technology. Microsoft has positioned Copilot as an AI orchestration layer across Microsoft 365, Dynamics and Azure. Research conducted by Forrester and other industry analysts has found measurable productivity improvements among knowledge workers using AI assistants, including reductions in time spent on administrative work, content production and information retrieval.
Similarly, OpenAI's rapid enterprise adoption demonstrates how quickly AI capabilities are moving from experimentation into core business operations. Large organisations including Morgan Stanley, Cisco, T-Mobile, Booking.com and Amgen are deploying AI systems across customer service, research, software development and internal knowledge management functions. Increasingly, AI is becoming part of the operational fabric of digital businesses rather than an isolated technology initiative.
The most significant impact of generative AI is not that individual employees work faster. It is that platforms can increase output without increasing operating costs at the same rate.
Advertising has historically relied on human judgement, audience segmentation and creative development. Generative AI is automating each of those functions simultaneously.
Meta's Advantage+ platform provides a useful example. Rather than requiring marketers to manually optimise audiences, placements and creative combinations, machine learning systems increasingly handle campaign execution in real time. The platform continuously tests creative variations, reallocates budget and adjusts targeting based on performance signals.
The commercial implications are significant. When campaign optimisation becomes automated, marketing efficiency improves, but so does platform dependency. More advertising decisions move inside proprietary systems that advertisers do not directly control.
The same pattern is emerging in commerce.
Product discovery, merchandising, customer service and conversion optimisation are increasingly being managed through AI-powered workflows. Personalisation engines can generate product recommendations, promotional messaging and shopping experiences tailored to individual users at a scale that manual processes cannot replicate.
For e-commerce platforms, AI is becoming less of a feature and more of a commercial operating model.
For e-commerce platforms, AI is becoming less of a feature and more of a commercial operating model.
Few companies illustrate this transition more clearly than Shopify.
Historically, Shopify provided merchants with the infrastructure required to build and manage online stores. Increasingly, it is providing intelligence as well as infrastructure.
Products such as Sidekick, Shopify Magic and Shopify Flow embed AI throughout the merchant journey. Store owners can generate content, analyse performance, automate workflows and receive operational guidance without leaving the platform. Tasks that once required specialist expertise can increasingly be completed through conversational interfaces.
The longer-term vision extends beyond automation. Shopify's leadership has openly discussed the emergence of agentic commerce, where AI systems assist consumers with product discovery, purchasing decisions and transaction execution.
In this model, AI does not merely support commerce. It participates in it.
The distinction is important because it changes who the platform serves. Historically, platforms connected businesses and consumers. Increasingly, they will connect businesses, consumers and AI agents simultaneously.
Historically, platforms connected businesses and consumers. Increasingly, they will connect businesses, consumers and AI agents simultaneously.
The most important economic shift is still ahead.
Traditional software requires users to learn interfaces. AI-native platforms increasingly adapt to users instead. Search boxes become conversations. Dashboards become assistants. Workflows become recommendations. Navigation becomes intent-driven.
This reduces friction across the customer journey while increasing the platform's role as an intermediary between users and outcomes.
For publishers, marketers and platform operators, the strategic implication is clear. Generative AI is not simply helping organisations operate more efficiently. It is changing the architecture of digital business itself.
The companies creating the greatest value will not necessarily be those that deploy AI most aggressively. They will be those that redesign products, services and customer experiences around AI as a permanent layer of infrastructure.
That is why generative AI should be understood as an economic transformation rather than a technology trend. It is reshaping how platforms create value, how users interact with digital services and ultimately how revenue is generated across the modern web.
The benefits of AI-driven productivity are substantial. Platforms can create content faster, automate decisions more efficiently and personalise experiences at unprecedented scale.
Yet every economic advantage created by AI has a corresponding shift in power.
As platforms become better at generating, summarising and recommending information, they also become more capable of keeping users within their own ecosystems. The same technologies improving user experience are reshaping how value flows across the web.
Every economic advantage created by AI has a corresponding shift in power.
Nowhere is that shift more visible than in publishing.
For more than twenty years, the commercial relationship between publishers and search engines was relatively straightforward.
Publishers created content. Search engines indexed it. Users clicked through to source websites. Traffic was exchanged for information.
The arrangement was never entirely balanced, but it created a functioning ecosystem. Publishers invested in journalism, expertise and content creation because search engines delivered audiences in return.
Artificial intelligence is changing that relationship.
Google's AI Mode and AI-generated search experiences represent the most significant shift in web discovery since the introduction of PageRank. Instead of directing users to websites, search increasingly seeks to answer questions directly. Information is extracted, summarised and presented within Google's own interface.
The strategic consequence is profound. Discovery is moving from link-based navigation to answer-based consumption.
Discovery is moving from link-based navigation to answer-based consumption.
For users, this often creates a better experience. Questions are answered faster. Research takes less effort. Information becomes easier to access.
For publishers, however, the economics are fundamentally different.
The value of publishing has historically depended on attracting audiences to owned properties where advertising, subscriptions, memberships and first-party data could be monetised. When answers remain inside the search interface, much of that value remains with the platform rather than the content creator.
The challenge is not that Google is sending less traffic. The challenge is that search itself is being redefined.
The effects are already becoming visible across the publishing industry.
Multiple studies tracking AI Overviews and AI-powered search experiences have reported significant reductions in click-through rates when answers are displayed directly within search results. Research cited across publishing and SEO sectors has pointed to traffic declines approaching 43 per cent for some publisher categories, particularly for informational content where user intent can be satisfied without leaving the search page.
This trend reflects the continued growth of what has become known as zero-click search.
In a traditional search journey, users discovered information through a sequence of clicks. In an AI-mediated journey, discovery and consumption increasingly happen simultaneously. The answer appears before the visit.
The implications extend beyond search traffic alone.
Research from the Reuters Institute for the Study of Journalism has repeatedly shown changing audience behaviours, including growing news avoidance, increasing platform dependency and fragmentation of attention across multiple digital channels. As Nic Newman has observed through successive Digital News Reports, audiences are becoming less loyal to specific publishers and more reliant on platform-mediated discovery.
Artificial intelligence accelerates this pattern.
Users no longer need to know which publication produced a piece of reporting. They need only ask a question. The platform handles the rest.
In an AI-mediated journey, discovery and consumption increasingly happen simultaneously. The answer appears before the visit.
For publishers whose business models depend heavily on search referrals, the result is a growing disconnect between content production and audience acquisition.
This is where the strategic challenge becomes existential.
Publishing organisations bear the cost of producing original reporting, expert analysis and authoritative content. AI systems increasingly capture and synthesise that information without necessarily sending equivalent value back to the source.
A.G. Sulzberger, publisher of The New York Times, has repeatedly argued that quality journalism depends on sustainable economic models. The concern is not simply copyright. It is economic sustainability.
If the organisations that create trusted information lose the commercial incentives to produce it, the quality of the information ecosystem itself begins to deteriorate.
The question facing publishers is therefore larger than search rankings or SEO strategy.
It is whether the open web can continue to fund the creation of original knowledge when discovery increasingly occurs through intermediaries that do not require users to visit the source.
The question is whether the open web can continue to fund the creation of original knowledge when discovery increasingly occurs through intermediaries.
The most resilient publishers are already adapting.
The New York Times has spent years reducing dependence on search traffic by investing in subscriptions, reader relationships, product diversification and direct engagement. Games, audio, newsletters and memberships all strengthen the connection between audience and publisher without relying exclusively on platform referrals.
Across the industry, membership-driven publishers are pursuing similar strategies. Reader revenue, community-building and first-party data programmes are becoming increasingly important as referral traffic becomes less predictable.
The common characteristic is not their business model. It is their ownership of the audience relationship.
The common characteristic is not their business model. It is their ownership of the audience relationship.
It is tempting to frame the current situation as an AI problem.
In reality, it is a dependency problem.
The publishing industry has encountered this pattern before.
Between 2012 and 2018, many publishers became dependent on Facebook for audience growth. When the platform changed its algorithm, traffic disappeared almost overnight.
Between 2010 and 2025, many businesses became heavily dependent on organic search. As search behaviour evolved, so did the risks associated with that dependency.
The emerging AI era presents a similar challenge. Organisations that become entirely dependent on AI-mediated discovery risk surrendering control over audience acquisition, distribution and commercial performance to systems they do not own.
The lesson is not that publishers should abandon search, social media or AI platforms. It is that no external distribution channel should become the foundation of a business model.
The fundamental risk is not AI itself. It is over-dependence on any external discovery mechanism.
The publishers most likely to succeed in an AI-mediated internet will be those that treat discovery as rented attention and audience relationships as owned assets.
Email subscribers, memberships, communities, events, first-party data and direct engagement may lack the scale of platform distribution, but they offer something increasingly valuable: resilience.
As artificial intelligence becomes the primary interface between audiences and information, ownership of the audience relationship becomes a strategic advantage rather than a marketing objective.
The publishers most likely to succeed will treat discovery as rented attention and audience relationships as owned assets.
The previous sections focused on the economic benefits of artificial intelligence: greater productivity, improved personalisation and new platform operating models.
However, the same forces creating value are also creating new forms of risk.
The fundamental challenge is not technological capability. It is trust.
As AI dramatically reduces the cost of generating content, information becomes more abundant than at any point in the history of the web. At the same time, it becomes increasingly difficult to verify where that information originated, whether it is accurate and who should be accountable for it.
This creates a new strategic tension for publishers, platforms and marketers. The competitive challenge is no longer simply attracting attention. It is establishing credibility within environments where synthetic content, algorithmic decision-making and AI-generated outputs are becoming normalised.
In the next phase of the platform economy, trust may prove more valuable than scale.
In the next phase of the platform economy, trust may prove more valuable than scale.
This challenge is closely connected to the broader transformation explored in our Future of Digital Publishing in the Age of AI analysis.
The ability to generate convincing text, images, audio and video has dramatically lowered the cost of producing synthetic content.
This creates obvious commercial opportunities. It also creates unprecedented opportunities for deception.
Deepfake incidents have increased sharply in recent years, driven by improvements in generative models and wider public access to AI tools. What once required specialist expertise can now be produced with consumer-grade software and minimal technical knowledge.
For publishers and platforms, the implications extend beyond isolated cases of fraud or misinformation. The larger risk is a gradual erosion of confidence in digital content itself.
When audiences can no longer easily distinguish between authentic and synthetic media, trust becomes harder to establish and easier to lose.
Research cited by the OECD and multiple academic institutions has highlighted growing concerns about AI-generated misinformation, particularly during elections, public health events and breaking news situations. The speed at which synthetic content can be created and distributed often exceeds the speed at which it can be verified.
For media organisations, this creates a paradox. AI can accelerate content production, but it simultaneously increases the importance of editorial verification. As synthetic content becomes abundant, authenticated content becomes more valuable.
As synthetic content becomes abundant, authenticated content becomes more valuable.
The risks associated with AI extend beyond misinformation.
AI systems learn from historical data, and historical data frequently reflects existing social, cultural and economic biases. As a result, algorithmic systems can unintentionally amplify inequalities rather than eliminate them.
Research published through the IEEE and numerous academic studies has documented concerns relating to bias in recommendation systems, content moderation tools and automated decision-making processes. These issues can affect visibility, representation and access to information at scale.
For online platforms, this raises difficult governance questions.
Who is accountable when an algorithm systematically favours certain viewpoints, creators or publishers? How should organisations explain decisions generated by increasingly complex AI systems? What level of transparency should users expect?
The UK Science, Innovation and Technology Committee has repeatedly highlighted the challenges associated with harmful algorithmic amplification and platform accountability. These concerns become more significant as AI systems gain greater influence over what people see, read and believe.
For business leaders, algorithmic bias is not merely an ethical issue. It is a reputational and commercial risk. Trust can be damaged as easily by flawed recommendations as by inaccurate content.
Trust can be damaged as easily by flawed recommendations as by inaccurate content.
The third major challenge concerns ownership.
Generative AI systems are trained on vast quantities of existing content, much of it created by publishers, authors, artists and rights holders. This has triggered a growing wave of legal disputes over copyright, licensing and fair use.
The debate is not simply about intellectual property law. It is about economic incentives.
If AI companies can extract value from creative and journalistic work without providing sustainable compensation mechanisms, questions emerge about who will fund future content creation.
Organisations such as Cambridge University Press and publishers across multiple sectors have argued that transparency around training data is essential for maintaining a healthy information ecosystem. Legal experts including Catie Sheret have also highlighted the importance of balancing technological innovation with the rights of creators and publishers.
The outcome of these disputes will help determine how value is distributed across the AI economy.
For publishers, the stakes are particularly high. Their content increasingly powers AI systems, even as those same systems reduce direct traffic and audience engagement opportunities.
The long-term challenge is therefore larger than copyright enforcement. It is the sustainability of the information economy itself.
The long-term challenge is larger than copyright enforcement. It is the sustainability of the information economy itself.
As artificial intelligence becomes embedded across online platforms, capability will continue to improve. The harder question is whether institutions, platforms and creators can build systems that remain transparent, accountable and worthy of public trust.
The next phase of the AI era will not be defined solely by what machines can generate. It will be defined by whether users believe what they see.
The next phase of the AI era will not be defined solely by what machines can generate. It will be defined by whether users believe what they see.
The next competitive advantage in digital publishing may not come from producing more content. It may come from proving where content originated, who created it and whether it can be trusted.
As AI systems increasingly mediate discovery, regulators, platforms and users are all demanding greater transparency. What began as a compliance discussion is rapidly becoming a visibility discussion. In an AI-mediated internet, trust signals are becoming operational infrastructure.
In an AI-mediated internet, trust signals are becoming operational infrastructure.
The European Union's AI Act represents the first major attempt to create a comprehensive regulatory framework for artificial intelligence.
While much attention has focused on high-risk AI systems, the legislation also contains provisions with significant implications for publishers, marketers and platform operators. Article 50 introduces transparency obligations for certain AI-generated and AI-manipulated content, requiring organisations to disclose when users are interacting with AI systems or consuming synthetic media in defined circumstances.
For businesses operating internationally, these requirements extend beyond legal compliance. They establish a precedent. As regulators seek to improve accountability and reduce misinformation, transparency is becoming a baseline expectation rather than a differentiator.
The strategic implication is clear: organisations that build transparency into their content operations now will be better positioned than those that treat regulation as a future problem.
Transparency is becoming a baseline expectation rather than a differentiator.
Regulation alone cannot solve the trust challenge. Technical standards are also emerging to help verify the origin and history of digital content.
One of the most significant developments is the growing adoption of the C2PA (Coalition for Content Provenance and Authenticity) standard. Supported by major technology, media and software organisations, C2PA enables cryptographically verifiable metadata to travel with digital assets, recording information about creation, editing and publication.
At the same time, industry initiatives are expanding beyond simple content labelling towards richer systems of attribution and ownership. Research from the ORA Framework, developed through the DECaDE Research Centre, explores how machine-readable provenance information can help establish ownership, rights and accountability across increasingly complex content ecosystems.
The UK Digital Regulation Cooperation Forum (DRCF) has also highlighted provenance and transparency as critical areas of future policy development, particularly as generative AI systems become more widely deployed.
For publishers, the significance is substantial. The question is no longer whether content was generated with AI assistance. The question is whether its origin can be verified.
The question is no longer whether content was generated with AI assistance. The question is whether its origin can be verified.
The long-term importance of provenance extends far beyond regulatory compliance.
Research led by Prof. John Collomosse at the DECaDE Research Centre demonstrates how provenance systems can create persistent records of content origin, modification and ownership. In practice, this means a publisher could provide verifiable evidence showing who created a piece of content, when it was produced and how it has been edited over time.
Today, this capability is emerging. Tomorrow, it may become mandatory infrastructure.
Digital Faction believes provenance will evolve into something akin to SEO in the early search era.
Publishers once invested heavily in technical optimisation to help search engines understand and rank content. The next phase of the web may require organisations to invest similarly in provenance infrastructure so AI systems can assess authenticity, authority and trustworthiness.
Provenance becomes SEO for trust.
In that sense, provenance becomes "SEO for trust".
The organisations that implement verifiable authorship, transparent editorial processes and machine-readable trust signals early will gain advantages that extend beyond compliance. They will be easier for AI systems to trust, easier for users to verify and better positioned in an ecosystem where credibility increasingly influences visibility.
As AI becomes the primary intermediary between audiences and information, trust signals may become as important as ranking signals once were. For a deeper exploration of information quality, trust signals and digital infrastructure, see our Building Better Information Systems for the Modern Web .
As AI becomes the primary intermediary between audiences and information, trust signals may become as important as ranking signals once were.
The economic stakes are difficult to overstate. PwC estimates that AI could contribute $15.7 trillion to the global economy by 2030, while the World Economic Forum projects significant changes to how work, information and commerce are organised. Yet the most important shift may not be productivity gains or automation. It may be the emergence of AI agents as active participants in digital ecosystems.
The internet was built for humans. The next version may increasingly be navigated by machines acting on behalf of humans.
The internet was built for humans. The next version may increasingly be navigated by machines acting on behalf of humans.
For two decades, publishers and marketers optimised for search engines and social algorithms. The next challenge may be optimising for autonomous agents.
Research from the OECD, Gartner and BISI increasingly points towards agentic AI systems capable of independently researching options, evaluating sources, comparing products and completing multi-step tasks. Platforms including OpenAI, Google, Microsoft and Amazon are all investing heavily in this direction.
In practical terms, this means a growing proportion of online interactions may begin with an AI assistant rather than a browser search. Instead of typing a query into Google, a user may simply instruct an agent to find the most trustworthy source, book a service or recommend a product.
The intermediary is no longer a search engine. It is an autonomous decision-making layer.
The intermediary is no longer a search engine. It is an autonomous decision-making layer.
The implications extend beyond information discovery.
Shopify's vision of agentic commerce illustrates how AI is becoming embedded throughout the purchasing journey. Rather than consumers manually researching products, comparing options and completing transactions, intelligent agents may increasingly handle these processes on their behalf.
The result is the emergence of agent-to-agent commerce: AI systems negotiating with other AI systems to identify products, evaluate suppliers, optimise pricing and execute transactions.
Singapore IMDA and OECD future-scenario research suggest that this model could become a meaningful component of digital commerce within the next decade. The competitive landscape would change fundamentally. Visibility alone becomes insufficient. Businesses must also become machine-readable, trustworthy and easily evaluated by autonomous systems.
Visibility alone becomes insufficient. Businesses must also become machine-readable, trustworthy and easily evaluated by autonomous systems.
This brings the article's central question into focus.
The discovery dependency problem does not disappear with AI. It evolves.
Many publishers became dependent on Facebook distribution between 2012 and 2018. Many then became dependent on search visibility between 2010 and 2025. A new generation of businesses now risks becoming dependent on AI intermediaries.
This brings the article's central argument full circle.
Every major era of the internet has been defined by a dominant discovery mechanism. First came portals. Then search. Then social platforms and recommendation engines. Agentic AI represents the next transition.
The risk for businesses is repeating a familiar mistake: becoming overly dependent on distribution systems they do not control.
The organisations most likely to succeed will not optimise exclusively for algorithms, search rankings or AI assistants. They will invest in assets that retain value regardless of how discovery evolves: recognised expertise, trusted brands, direct audience relationships and verifiable provenance.
In a world where AI agents increasingly evaluate information, compare providers and make recommendations on behalf of users, visibility becomes only the starting point. Credibility becomes the deciding factor.
Visibility becomes only the starting point. Credibility becomes the deciding factor.
The next competitive battle may not be for human attention. It may be for AI agent preference.
The next competitive battle may not be for human attention. It may be for AI agent preference.
Artificial intelligence is no longer a feature layer sitting on top of online platforms. It is becoming part of the infrastructure through which information is organised, decisions are made and digital experiences are delivered.
That shift explains why the current AI wave differs from previous technology cycles. The most important change is not the automation of individual tasks. It is the emergence of AI as an intermediary between audiences and the information, products and services they consume.
For publishers, marketers and platform operators, this reframes the challenge entirely. Success is no longer determined solely by creating content, attracting traffic or improving efficiency. Increasingly, it depends on remaining visible and trusted within systems that are making more decisions on behalf of users.
Success increasingly depends on remaining visible and trusted within systems that are making more decisions on behalf of users.
The organisations best positioned for this environment will combine technological capability with institutional credibility. They will use AI to improve productivity, but they will also invest in expertise, editorial standards, direct audience relationships and transparent provenance systems. They will understand that authority becomes more valuable as content becomes more abundant.
The lesson extends beyond publishing. Every major platform transition has rewarded organisations that owned their relationship with users rather than relying exclusively on external distribution channels. The rise of AI does not change that principle. If anything, it makes it more important.
Throughout this article, the same pattern has emerged repeatedly. Recommendation systems determine visibility. Generative AI changes platform economics. Search is evolving into answer generation. Trust is becoming machine-readable. Provenance is becoming infrastructure. Agentic systems are beginning to participate directly in discovery and commerce.
In each case, the organisations most likely to succeed are those that combine technological adaptation with strategic independence.
The organisations most likely to succeed are those that combine technological adaptation with strategic independence.
The next generation of online platforms will not be defined by who creates the most content, but by who earns the most trust.
As AI increasingly determines what people see, read, buy and believe, trust becomes more than a brand attribute. It becomes infrastructure.
And in an AI-mediated internet, infrastructure is where competitive advantage is built.
As AI increasingly determines what people see, read, buy and believe, trust becomes more than a brand attribute. It becomes infrastructure.
Digital Faction Limited is a UK digital products company focused on digital publishing, artificial intelligence, search technologies, analytics, automation, research and online platforms.
Learn more about our work, topics and research initiatives.