For much of the commercial web’s history, publishing followed a relatively stable exchange. Publishers created information, search engines indexed it, and users clicked through to the original website. Those visits generated advertising impressions, subscriptions, leads, and other forms of commercial value.
Generative search is changing the middle of that exchange. Instead of presenting a list of sources and asking users to choose, an AI system can read multiple pages, synthesize their contents and deliver a finished response. The publisher may have contributed the reporting, data or explanation, but the audience can receive most of the value without visiting the source.
Early usage data suggests this is already changing behavior. Pew Research Center found that users clicked a traditional Google result during 8 percent of visits when an AI summary appeared, compared with 15 percent when one did not. Links cited inside the AI summary received clicks during only 1 percent of visits. Users were also more likely to end their browsing session after receiving an AI-generated answer.
This raises a question that reaches beyond search engine optimization. If machines increasingly read, interpret and redistribute online information before humans encounter it, are they becoming the web’s primary audience? The answer will shape how content is produced, how advertising is bought and how publishing can remain economically viable.
From search rankings to generated answers
Generative engine optimization, commonly shortened to GEO, has emerged as a way of describing content optimization for AI-generated responses. The term gained wider recognition following research presented at ACM SIGKDD, which found that certain content changes could improve visibility in generative engine responses by as much as 40 percent. Results varied considerably by subject and optimization method, suggesting that GEO is less predictable than a universal checklist.
Traditional SEO aims to secure a prominent position among links. GEO attempts to make a source understandable, credible, and useful enough to be selected during answer generation. The machine is no longer simply deciding where a page should rank. It may decide which claims to extract, which sources to cite, and which details to leave out.
Google maintains that publishers do not need special AI markup to appear in AI Overviews or AI Mode. Its guidance for AI features in search continues to emphasize established practices such as accessible text, accurate structured data, internal linking, crawlability and helpful, reliable content. Google also says AI search can expose users to a wider range of websites and that visitors who do click from AI Overviews may spend more time on the destination site.
These claims are compatible with GEO when it is treated as an extension of sound publishing rather than a formula for manipulating models. Pages with clear explanations, identifiable authors, original evidence, current information and well-supported claims are easier for both people and machines to assess.
The risk appears when publishers start writing primarily for extraction. Articles can become collections of self-contained statements, repeated definitions and predictable question-and-answer sections designed to secure citations. Such content may perform well as machine input while becoming less rewarding for a human reader.
The web now has a layered audience
Machines are not an audience in the human sense. They do not subscribe because they value a publication, purchase products after seeing an advertisement or develop loyalty to an editorial voice. They increasingly operate as an intermediary audience, reading content first and determining how much of it reaches people.
The imbalance between machine consumption and human referral is considerable. Cloudflare measured approximately 70,900 crawler requests from Anthropic for every recorded referral from its platform during one week in June 2025. Cloudflare acknowledged that native applications do not always pass referral information, so the ratio may overstate the gap, but the broader pattern remains significant. AI systems can consume large volumes of publisher content while returning relatively little measurable traffic.
The Reuters Institute’s 2026 industry report notes that some observers expect bots to make more requests to publisher websites than people. Three-quarters of the media executives surveyed also expected agentic AI tools to have a large or very large impact on the news industry.
The web’s audience is therefore becoming layered. Publishers still write for people, but their first reader may be a search model, chatbot or autonomous agent. That intermediary may summarize the content, combine it with competing sources or use it to complete a task without showing the original page.
The click is disappearing from publishing economics
The open web’s advertising model depends heavily on page visits. A reader loads an article, an advertisement is served, and the publisher receives a share of the resulting revenue. AI answers can break that sequence by separating information consumption from the page where it was created.
A 2026 study comparing English-language Wikipedia pages exposed to Google AI Overviews with equivalent pages in other languages estimated an average traffic reduction of approximately 15 percent. The decline was larger for cultural topics and smaller for science and technology subjects, suggesting that concise informational content is especially vulnerable when a generated summary can satisfy the user’s immediate need.
Media executives are preparing for a broader contraction. Respondents to the Reuters Institute’s 2026 survey expected search traffic to fall by an average of 43 percent over three years. Chartbeat data included in the report showed that Google search referrals to more than 2,500 news sites had already fallen by 33 percent globally between November 2024 and November 2025, although the report cautioned that AI Overviews were unlikely to be the sole cause.
The economic tension becomes sharper when the answer platform can continue selling advertisements. A recent academic analysis found that more than half of the pages cited by Google AI Overviews carried display advertising. When the overview satisfies the query, the source publisher can lose an impression even while Google displays sponsored placements alongside the generated answer. The same research found that 11 percent of the claims examined were not supported by the cited pages, showing that citation does not always guarantee faithful representation.
Advertising moves into the answer layer
For advertisers, AI-generated search may create a more concentrated and intent-rich environment. A conversational query can reveal what a consumer is comparing, which constraints they have, and how close they may be to making a decision. Google is already testing new advertising formats in AI Mode and positioning them as a way for brands to participate directly in generated recommendations and commercial journeys.
This could reduce friction between discovery and purchase. It may also transfer more control to the platform operating the answer engine. Advertisers that once reached audiences across hundreds of independent publications may increasingly purchase access through a smaller number of AI interfaces. The platform can control placement, attribution, reporting and the boundary between an independent recommendation and a paid result.
There is also a contextual cost. Publisher environments give advertisers access to identifiable communities, editorial subjects and trusted brands. A placement beside specialist reporting carries meaning that may be difficult to reproduce inside a general-purpose response. As traffic declines, advertisers could lose some of the independent media environments that once helped them reach niche and professionally relevant audiences.
The Interactive Advertising Bureau expects generative and agentic AI to affect the entire campaign lifecycle, including media planning, segmentation, partner selection and attribution. Its 2025 State of Data report found that only 30 percent of agencies, brands and publishers had fully integrated AI across that lifecycle, while half of the industry still lacked a strategic roadmap.
Advertisers should therefore measure AI visibility, but they should also examine where their spending supports original information production. Efficient targeting offers diminishing returns when the underlying supply of credible, useful content begins to shrink.
Publishers need content machines cannot commoditize
GEO can improve discoverability, but citation alone will rarely replace revenue at scale. Publishers need to distinguish between information that can be summarized easily and work that creates a reason to visit, subscribe, or return.
Original reporting, proprietary data, expert access, local knowledge, practical tools and first-hand analysis are harder to reproduce from a short answer. They also give AI systems a clearer reason to cite the publisher as a primary source rather than one of many sites repeating the same information.
This direction is already visible in publisher strategy. Reuters Institute respondents said they planned to place greater emphasis on original investigations, contextual analysis and human stories, while reducing investment in generic news, evergreen articles and service content that chatbots can readily summarize. They also expected to invest more heavily in video and audio.
Publishers should still make this work machine-readable. Important claims should be supported by evidence. Authors and experts should be clearly identified. Dates, sources and update histories should be visible. Structured data should match what readers can see, and valuable information should not be hidden entirely inside graphics or inaccessible formats.
The objective is to make content easy to verify without making it easy to replace.
A new value exchange is taking shape
The publishing economy may eventually develop mechanisms that compensate machine consumption directly. Cloudflare has introduced tools allowing website owners to block AI crawlers or experiment with charging for access. The IAB Tech Lab has also proposed a Content Monetization Protocol intended to help AI systems and publishers establish commercial agreements before content is crawled or used. These initiatives remain developing standards rather than settled business models.
Larger publishers may negotiate licensing agreements directly. Smaller publications will need collective standards, technical controls or intermediaries capable of aggregating their rights and negotiating power. Otherwise, machine-readable content may become another market in which scale determines who gets paid.
Direct audience relationships will remain equally important. Newsletters, memberships, events, applications, communities and specialist services give publishers access to readers without depending entirely on algorithmic referrals. They also produce first-party insight that can help advertisers understand the value of an audience beyond raw page views.
Machines are unlikely to become the web’s main audience in terms of purpose. The web still exists to inform, entertain, and connect people. Machines may, however, become its main gatekeepers and some of its most intensive consumers.
That distinction will define the next publishing economy. Publishers will need to create material that machines can find while retaining experiences that people still choose to visit. Advertisers will need to decide how much value lies inside the answer layer and how much depends on the independent publications that supply it. AI platforms, meanwhile, will face pressure to provide attribution, traffic or compensation when their products depend on work created elsewhere.
The open web can survive a machine-mediated attention economy, but the exchange can no longer be based on crawling alone. Content consumption must produce enough value for the people and organizations that continue to create it.
TNGlobal INSIDER publishes contributions relevant to entrepreneurship and innovation. You may submit your own original or published contributions subject to editorial discretion.
Featured image: Growtika on Unsplash
Why grid readiness is becoming Southeast Asia’s next tech constraint