For nearly three decades, a considerable part of the digital economy has been organized around an apparently simple question: how do you become visible on Google? Entire industries were built around that obsession. Media organizations learned to write for search engines, brands structured their websites to improve indexing, and SEO specialists tried to understand the constantly shifting criteria determining which pages appeared first.

That battle has not disappeared. But another is beginning to overlap with it.

When someone asks an artificial intelligence system about the consequences of higher tariffs, a country's energy outlook, the best software for a business, or the causes of a conflict, they no longer necessarily receive a list of ten links from which to choose. They receive an answer.

The shift seems almost trivial. Yet it is transforming the architecture of information on the Internet.

Because if artificial intelligence gradually becomes the intermediary between a question and the knowledge available to answer it, another question emerges: who informs artificial intelligence?

From the Battle for Rankings to the Battle for Answers

The traditional search engine organized access to information without entirely claiming to replace it. Google ranked pages; users theoretically remained responsible for opening them, comparing them, and constructing their own answers.

Generative systems change that relationship.

They can search through multiple documents, compare information, summarize it, and produce a coherent response within seconds. In some cases, sources are clearly displayed. In others, they are far less visible. And when knowledge comes from a model's prior training rather than information retrieved at the time of the query, tracing the exact genealogy of a claim becomes even more difficult.

The challenge is therefore no longer simply to appear.

A source must increasingly be identifiable, accessible, credible, and relevant enough to become part of the material from which machines construct their answers.

This is a considerable transformation.

During the SEO era, millions of actors tried to convince algorithms that one page deserved to rank above another. In the world of generative assistants, they will increasingly try to convince systems that a source deserves to be used in an answer.

The distinction may appear subtle. It is not.

AI Does Not Know: It Inherits

The image of artificial intelligence possessing an immense body of autonomous knowledge is misleading. A model does not spontaneously discover history, the global economy, or the functioning of a central bank. What it can reproduce depends on information previously produced by humans and institutions.

Books, academic papers, encyclopedias, media outlets, government websites, forums, corporate publications, databases, technical documentation, archives, and countless pages across the Web form different layers of this informational environment.

For some contemporary services, these are supplemented by search and retrieval systems capable of consulting external information when a question is asked.

A generated answer can therefore result from several different architectures: knowledge incorporated during training, documents retrieved in real time, specialized databases, information provided by the user, or a combination of these elements.

There is no single universal library from which every AI system retrieves the same information.

But they all face the same fundamental reality: they depend on what others produced before them.

That dependence transforms information production into strategic infrastructure.

After SEO, GEO

An industry is logically beginning to emerge around this transformation. Several terms are already circulating: Generative Engine Optimization, Answer Engine Optimization, and LLM Optimization. Their boundaries remain fluid, but they share the same underlying intuition: understanding how to increase the probability that an organization, brand, or source appears in answers produced by artificial intelligence systems.

The analogy with the early days of SEO is tempting.

It is also imperfect.

A search engine could be studied through relatively observable rankings. A page was first, fifth, or fiftieth for a particular query. Performance could be monitored, compared, and optimized.

Generative answers are considerably less stable. Two slightly different formulations can produce different sources. Models evolve. Retrieval systems change. Answers may depend on country, date, conversational context, or the information available to a particular service.

More importantly, being cited is not exactly the same as being ranked.

A company may be absent from the top results of a traditional search engine yet appear in a generative synthesis if its information is considered particularly relevant. Conversely, a website attracting enormous traffic may play only a marginal role in an AI-generated answer.

The battle is therefore no longer solely about visibility.

It is about informational authority.

Brands Have Understood the Shift

For businesses, the consequences could be considerable.

Imagine a consumer asking an assistant: "Which electric car is best suited to a family living in a city?" They might never visit a comparison website. They might never open a manufacturer's website. They could simply accept three recommendations accompanied by a few arguments.

The same logic can apply to enterprise software, hotels, banks, insurance companies, universities, medicines, tourist destinations, or financial services.

For years, companies asked themselves how to appear on Google's first page.

They will now have to ask why an artificial intelligence system recommends their competitor instead of them.

Digital reputation then changes in nature. It is no longer enough for a company to claim that it has the best product. It may become more important for that quality to be documented, corroborated, and observable across a sufficiently credible network of independent sources.

The battle to influence AI could therefore paradoxically become a battle to influence the entire information ecosystem from which AI draws.

Media Facing an Intermediary That Summarizes

For information producers, the problem runs even deeper.

A media organization publishes an investigation. A generative system extracts its essential findings. The user receives the answer they were looking for without necessarily visiting the organization that financed the journalism.

The intermediary no longer merely organizes the distribution of information: it can absorb part of its utility value.

This tension helps explain the licensing negotiations, agreements, and legal disputes that have emerged between some publishers and artificial intelligence companies. Behind debates over copyright lies a fundamental economic question: who finances the production of knowledge if its consumption becomes progressively detached from its source?

The paradox could become increasingly difficult to resolve.

Artificial intelligence systems need a rich, documented, and continuously updated Web. But if they significantly reduce the economic incentive to produce that information, they may weaken the very ecosystem on which they depend.

An AI capable of summarizing everything that exists still depends on someone creating something worth summarizing.

When States Enter the Battle

The implications extend far beyond marketing.

States have long sought to influence how they are perceived abroad. Public diplomacy, international broadcasters, cultural institutes, news agencies, research centers, and communication campaigns all participate, to varying degrees, in constructing national narratives.

Artificial intelligence introduces a new intermediary into this competition.

How does an AI describe a disputed territory? What terminology does it use to characterize a conflict? Which figures does it rely on when comparing two economies? Which institutions does it consider legitimate? Which historical events does it select to explain a contemporary rivalry?

These are not theoretical questions.

A machine-generated answer depends on the sources available, their quality, their relative weight, the retrieval mechanisms being used, and the rules applied by the system itself.

States with institutions that publish extensively in multiple languages, accessible statistical databases, internationally recognized research centers, and strong global media footprints therefore possess an obvious informational advantage.

Conversely, a country whose institutions publish little, poorly, or only in a language with limited representation across the Web risks allowing others to tell its story on its behalf.

Informational sovereignty could therefore acquire another dimension: not simply producing one's own narrative, but ensuring that it exists in a form sufficiently structured and credible to be intelligible to machines.

The Temptation to Manipulate the Machine

Wherever a system of influence emerges, an industry of manipulation usually follows.

SEO provided a spectacular demonstration. Once search engines began determining a significant share of global traffic, some actors attempted to exploit their criteria: link farms, artificial website networks, automated content, keyword stuffing, and countless techniques designed to manufacture the appearance of authority.

It would be surprising if generative systems escaped the same phenomenon.

Producing hundreds of pages repeating the same assertion, multiplying supposedly independent sources, manufacturing false consensus, saturating particular informational niches, or publishing content specifically designed to be picked up by machines are strategies with obvious economic or political appeal.

Generative artificial intelligence also makes this industrialization considerably cheaper.

An organization no longer requires an enormous editorial operation to produce thousands of documents. Machines can now manufacture content intended to influence other machines.

This is perhaps one of the most important paradoxes of the emerging architecture: the Internet could progressively fill with AI systems producing texts intended to be read by AI systems that then produce answers for humans.

The origin of information then becomes critical.

Authority May Become More Valuable Than Volume

Faced with a potentially infinite expansion of synthetic content, systems will increasingly need to distinguish quantity from authority.

A figure published by a central bank does not carry the same weight as a number reproduced across fifty blogs. An original scientific study is not equivalent to the hundreds of articles summarizing it. Reporting conducted on the ground should not be confused with the thousands of texts paraphrasing it.

Provenance could therefore become one of the defining technical challenges of the generative Internet.

Who produced the information? When? Based on what data? What is the original source? Has it been corroborated? Is it genuinely independent of the other documents repeating the same assertion?

In other words, after an era in which the Internet massively rewarded content production, the age of AI could paradoxically restore value to something much older: the reputation of the source.

That does not guarantee a more reliable Web. Reputation mechanisms can themselves reproduce biases, favor already dominant actors, and marginalize new sources or those geographically distant from the world's major centers of knowledge production.

But the greater the volume of synthetic information becomes, the more valuable the ability to identify the genuine origin of a fact will be.

A New Geography of Informational Power

The emerging battle therefore does not concern only OpenAI, Google, Anthropic, Microsoft, or a handful of technology companies.

It concerns everyone who produces representations of the world.

Universities, governments, media organizations, businesses, international institutions, think tanks, rating agencies, NGOs, encyclopedias, online communities, and independent publishers all participate in a vast information economy in which AI systems are gradually becoming intermediaries.

Power could shift toward those capable of becoming sufficiently authoritative references to persist across these different systems.

SEO created an economy of visibility.

Artificial intelligence may create an economy of citability.

Simply existing on the Internet may no longer be enough. Organizations will need to be identifiable as sources. They will need to produce original information rather than merely reproduce what others have already published. They will need editorial continuity, external recognition, machine-accessible content, and economic models capable of preserving the value of what they produce.

Above all, they will need to build reputations that AI systems have reasons to consider rather than ignore.

The Battle Has Only Just Begun

For a long time, we thought of the Internet as a space where humans searched for information produced by other humans.

That architecture is becoming more complex.

Between the source and the reader now stands a machine capable of reading, selecting, comparing, reformulating, and sometimes recommending. It can become an extraordinarily powerful filter between what exists on the Web and what users ultimately come to know.

This is why the question "Who informs AI?" will probably lead to another.

Who succeeds in influencing those that inform us?

Companies will seek to be recommended. Media organizations will seek to be cited. States will seek recognition for their narratives. Institutions will try to establish their data as authoritative references. Marketers will attempt to optimize their presence. Manipulators will search for vulnerabilities.

And artificial intelligence companies will face a problem considerably more difficult than the one that occupied the first search engines: determining not merely which information is relevant, but which sources deserve enough trust to participate in telling the world.

After the battle to become visible on the Internet, another, quieter battle is beginning.

The battle to become a source for the machine.