A reader asks a question, receives a summary and closes the tab. Sometimes, the article that informed the answer was never opened. For news publishers, this is the major shift in search in the age of artificial intelligence: being useful no longer means being visited. Looking ahead to September 2026, this disconnect threatens the funding of journalism as much as it reshapes its distribution. This analysis draws on publicly documented developments in 2024 and 2025; the potential developments discussed for 2026 are forward-looking scenarios, not established outcomes.
From search engine to answer desk
Zero-click search did not begin with ChatGPT. Weather, sports results, definitions: Google has long displayed certain information directly on its pages. But generative answers change the scale of the phenomenon. They can combine several sources, explain a news story and answer a complex question within an interface that holds the user’s attention.
Google launched its AI Overviews in the United States in May 2024 before expanding their availability. In 2025, its AI Mode took conversational search further. OpenAI, meanwhile, launched search in ChatGPT in October 2024. Perplexity has made answers accompanied by sources the core of its product. These interfaces differ, but their promise converges: to deliver an answer rather than a list of leads.
For a user asking why their electricity bill is rising, the convenience is obvious. For the newsroom that interviewed experts and dissected the rates, the equation is less clear. Its work can help produce a satisfactory explanation without generating the visit that funds that work.
Being cited is not the same as being read
Search visibility rested on an imperfect but understandable exchange: publishers made their pages accessible to search engines, which sent them visitors. Generative summaries introduce an additional intermediary. The engine selects sources, rephrases them, ranks them and decides how prominently to display links.
Three forms of value are now diverging: contributing to an answer, being visible within it and earning a click. A publisher may inform a summary without receiving clear attribution. It may also appear among the sources without attracting a reader. And an unobtrusive citation does not carry the same weight as a link placed in the main text.
This distinction calls for a more critical assessment of claims about the “visibility offered” to publishers. A mention can strengthen a brand, but it does not automatically replace an advertising impression, a newsletter signup or a subscription. Conversely, lower traffic volumes could bring more engaged readers: someone who clicks through after reading a summary may be looking for greater depth. That remains something to measure, not assume.
Not all content faces the same exposure
Pages answering a simple question appear particularly vulnerable. Definitions, how-to guides, basic comparisons and summaries of announcements lend themselves readily to rephrasing. Yet this service-oriented content has long been a profitable entry point, sometimes helping support reporting that attracts fewer searches but costs more to produce.
Exclusive investigations, original local reporting and bylined analysis have other strengths. Their value lies not just in a factual answer, but in discovery, trust or a distinctive perspective. That does not protect them from reuse, but it gives audiences more reasons to consult the original.
The danger would be to conclude that newsrooms should abandon all practical information. Explaining things is also part of their mission. The challenge is instead to enrich these formats: verifiable data, interactive tools, expertise from identifiable specialists and regular updates. All of these elements make a visit worthwhile beyond the summary.
The licensing battle is reshaping the balance of power
Faced with this redistribution, several groups have opted to negotiate. Associated Press reached an agreement with OpenAI as early as 2023. Axel Springer, followed notably by the Financial Times, Le Monde and News Corp, also announced partnerships. Their scope and terms vary; financial details often remain confidential.
These agreements should not be confused with a universal pricing framework for news. Accessing archives to train a model, consulting recent content and displaying it in an answer are distinct uses. A contract may cover some without addressing the others. The existence of a licence therefore guarantees neither a volume of visits nor satisfactory attribution.
Other publishers favour litigation. The lawsuit filed by the New York Times against OpenAI and Microsoft in December 2023 illustrates the disagreements over the use of copyrighted articles. In France, neighbouring rights have already shown how securing payment from platforms for content can require regulatory intervention. However, applying those rights to generative uses does not provide an automatic solution.
A market that favours must-have catalogues
An international group has archives, brands and legal resources that carry weight in negotiations. An independent publisher may produce indispensable reporting without having the same leverage. Looking ahead to 2026, the risk is therefore a highly unequal distribution of licensing revenue, unless genuinely accessible collective arrangements emerge.
Blocking bots, but at what cost?
Publishers can restrict certain forms of automated access, notably through the robots.txt file. But this tool provides instructions to cooperative bots: it is neither an absolute technical barrier nor a payment mechanism. Crucially, platforms do not all draw the same distinction between crawling for search and uses related to their models.
The dilemma is tangible: restricting access to protect content while remaining discoverable. When visibility depends on a dominant player, the theoretical ability to refuse is not enough to balance the relationship. Genuine negotiating power also depends on being able to leave without disappearing.
Regaining control of the reader relationship
For newsrooms, the response cannot be limited to a new search optimisation formula. Structured pages, identified authors and explicit sources make content easier to understand, without guaranteeing citations. Optimising solely for generative search engines would risk recreating the dependency publishers are seeking to reduce.
The priority is to track citations, visits and conversions separately, then strengthen direct channels: newsletters, apps, events and subscriptions. Platforms, meanwhile, should provide usage data that is more readily auditable. Without visibility into what is being reused and the visits actually referred, negotiating the value of journalism amounts to discussing it almost blind.
What next? Looking ahead to September 2026, two paths remain possible: answers that direct users towards a publishing ecosystem that is paid for its work, or interfaces that gradually absorb its audience. The outcome will depend on contracts, the law and design choices. For news publishers, the goal is not to save every click, but to retain a recognisable and financially sustainable place between the facts they establish and the public they serve.
