From SEO to GEO: How Search Visibility Evolved by 2026

Updated in September 2026.

The SEO we work with in 2026 has relatively little to do with the SEO that was practised fifteen or twenty years ago. Many of its foundations are still valid, but Google has gradually added new layers of understanding that have transformed the way it interprets a search, analyses a page and decides which information deserves to be shown to the user.

During the early years of search engine optimisation, we worked mainly with keywords. Links then became one of the main signals of authority. Later came search intent, entities, machine learning, user experience, content usefulness and systems capable of understanding natural language.

We are now entering another stage. Search engines no longer simply rank pages and display a list of results. They can also consult different documents, connect information, cross-check it and build a direct answer using generative artificial intelligence.

This is where GEO, Generative Engine Optimization, comes in.

I do not think it is accurate to say that SEO has literally become GEO or that one is going to replace the other. The evolution is more interesting than that. GEO can be understood as a new layer within a transformation of search visibility that has been developing for more than two decades.

For me, 2013 is also a useful point from which to analyse this evolution. It was the year I began working professionally in SEO and digital marketing, and it also coincides with a period in which Google clearly began moving from a search engine based mainly on words towards one increasingly interested in understanding meanings, relationships and intent.

If we want to understand why we talk so much about GEO and AI visibility in 2026, we first need to understand how we got here.

The evolution from SEO to GEO at a glance

Before looking at each stage in detail, this table provides a useful summary of how the main focus of search visibility has changed over the years.

Period	What gained importance	Main change	Impact on SEO
Before 2011	Keywords and links	Matching queries, content and popularity	Highly aggressive on-page optimisation and link building
2011-2013	Quality, links and entities	Panda, Penguin and Knowledge Graph	Google tries to distinguish genuine signals from manipulation
2013-2019	Intent and semantics	Hummingbird, RankBrain and BERT	Less dependence on exact-match keywords and greater understanding of meaning
2020-2023	Usefulness, experience and trust	Core Web Vitals, Helpful Content and E-E-A-T	Greater attention to the actual quality of the source
2023-2025	Retrieval and answer generation	The emergence of GEO, AI Overviews and AI Mode	Visibility no longer depends exclusively on an organic ranking
2026	Entities, sources, mentions and citations	Expansion of generative search	SEO, brand, content, digital PR and GEO begin to converge

Seen this way, the arrival of GEO does not look like a sudden break with the past. It looks much more like the next step in an evolution that had already been under way for years.

Before 2013: when SEO revolved around keywords and links

During the early years of search engines, a major part of the work involved determining whether a page contained the words a user had entered into a search engine.

This is where many of the classic on-page SEO techniques that those of us who have worked in the industry for a long time have used at some point come from: including the target keyword in the title, using it in headings, repeating it within the content, creating descriptive URLs, optimising internal anchor text or even trying to calculate a specific keyword density.

The logic was simple. If someone searched for a particular phrase and our page contained that phrase in locations the search engine considered important, we increased the chances that it would understand that the document could answer the query.

The problem was equally simple: if a search engine depended too heavily on what we wrote on a page, it was relatively easy to try to manipulate it.

PageRank turns links into an external signal

One of Google’s major differences was that it also used the link structure of the web. Sergey Brin and Larry Page had already explained part of this approach in the 1998 academic paper The Anatomy of a Large-Scale Hypertextual Web Search Engine.

A link no longer served only as a way to navigate from one page to another. It could also provide information about the importance of a document within the structure of the web itself.

If different pages linked to the same piece of content, Google had an external signal that did not come directly from the owner of that page. And if some of those links came from relevant websites, that signal could become even more valuable.

This is how the two major areas we used for years to explain a search visibility strategy became established:

• On-page SEO: everything we do within our own website to help a search engine correctly interpret a page and assess its relevance.
• Off-page SEO: external signals mainly related to popularity, authority and links.

On-page SEO helped explain what a page was about. Off-page SEO gave Google external references about it.

The problem began when the entire industry understood just how valuable links were.

Directories, exchanges, backlink purchases, automated comments, footer links, private blog networks and articles created almost exclusively to place a link became a common part of SEO at the time.

Link building worked, and precisely because it worked, it became one of the main battlegrounds between Google and those of us trying to understand its algorithm.

Panda and Penguin begin to change the rules

Panda arrived in 2011, aimed at improving the quality of search results and reducing the visibility of certain websites with low-value content. Google stated at the time that its initial launch noticeably affected around 12% of English-language searches.

Penguin followed in 2012 and was much more closely related to webspam and different forms of link manipulation.

Google still includes both in its documentation on ranking systems. Over time, they stopped operating as isolated updates and their principles were integrated into broader systems.

This did not mean that content or links stopped mattering. It meant something much more significant: Google needed to learn how to distinguish a genuine signal from one created purely to manipulate its results.

That same tension still exists in 2026, although it is now also moving into GEO.

2012: from words to entities

Another fundamental change occurred in May 2012.

Google introduced Knowledge Graph with an expression that summed up its new direction perfectly: “things, not strings”.

The search engine wanted to stop interpreting the internet merely as an enormous collection of words and begin understanding people, companies, places, products, events and concepts as entities connected to one another.

The original presentation of Knowledge Graph already referred to more than 500 million objects and billions of facts and relationships.

Keywords still mattered, of course, but a far more interesting question was beginning to emerge:

What does the user actually mean by what they are searching for?

This shift from words towards entities is one of the keys to understanding GEO today.

2013: Hummingbird and the beginning of a much more semantic SEO

Google introduced Hummingbird in 2013. Today it appears as a historical system because many of its advances were eventually incorporated into later technologies, but conceptually it represents the change that was taking place very well.

Google was moving from a relatively literal interpretation of queries towards a broader understanding of meaning.

At the time, it was still common to find projects where several almost identical pages were created in an attempt to rank for small variations of the same keyword. As Google’s ability to understand semantic relationships improved, that way of working gradually became less useful.

It became more valuable to build a page capable of properly satisfying an entire search intent.

The evolution could be summarised like this:

keyword → intent → topic → entity.

It did not happen overnight. For several years, strategies based on exact-match keywords, extremely specific pages and fairly aggressive link campaigns continued to work. But the direction was becoming clear.

It was precisely in 2013 that I began working professionally in SEO. Since then, I have seen the same pattern repeat itself with almost every major change: many techniques do not disappear, but stop working effectively in isolation and become part of more complex systems.

We still use keywords. We still work on links. We still optimise titles, content, architecture and internal linking. What has changed is the amount of context Google can use around each of those signals.

From 2013 to 2019: intent, concepts and natural language

RankBrain arrived in 2015 and Google would later describe it as its first deep learning system used in Search.

One of its functions was to help connect words with concepts and understand queries that had previously been difficult to interpret.

The consequence for SEO was clear: Google could depend less and less on content using exactly the same words the user had entered.

If different expressions could represent similar concepts, it made far more sense to develop a topic properly than to chase every keyword variation separately.

Technical SEO also expands its scope

Another important transformation was happening at the same time: the internet was becoming increasingly mobile.

Google launched its mobile-friendly update in 2015 and began experimenting with mobile-first indexing in 2016. The full transition took years, but the way we worked on technical SEO changed.

It was no longer simply about making sure Google could access a URL. Architecture, crawling, indexing, performance, mobile optimisation, structured data, internal linking and user experience gradually became part of the same strategy.

All of this remains essential in 2026, even though the current discussion around artificial intelligence can sometimes make it seem as though traditional search optimisation no longer matters.

BERT improves language understanding

Another major step came in 2019 with BERT.

BERT allowed Google to interpret each word more effectively by taking into account the words around it and therefore understand nuances, relationships and context in a far more advanced way.

Google itself described BERT as one of the most important advances in the history of Search. It initially affected around one in ten English-language queries in the United States, and its use was later expanded considerably.

For those of us working in SEO, the conclusion was clear: writing for an exact keyword increasingly made less sense than properly answering what that search actually meant.

The better Google became at understanding language, the less necessary it was to write artificially for a machine.

From 2020 to 2024: usefulness, experience and trust

During the next stage, SEO expanded its scope once again.

It was no longer enough for a document to be relevant to a search. Google also had to try to determine whether the content was useful, trustworthy and satisfying.

Core Web Vitals became part of the systems related to page experience. They never meant that a perfect technical score guaranteed a number-one ranking, but they once again reflected the fact that Google could evaluate a page using more signals.

Helpful Content and the difference between text and information

Helpful Content arrived in 2022 and Google began drawing a much clearer distinction between content created primarily for people and content produced mainly to attract search traffic.

The issue is even more important in 2026.

For years, SEO created a very strong incentive to produce content. If we found a keyword, we created a page. If we found a hundred keywords, we could end up producing a hundred pages.

Generative artificial intelligence can now create thousands of words in seconds. This has made something that was always true much more obvious:

the amount of text and the amount of useful information are not the same thing.

A page can contain 4,000 words and contribute almost nothing that is not already available on a hundred other websites.

This is precisely why first-hand experience, original data, research, real-world cases and specialist knowledge are becoming even more valuable.

E-E-A-T and the importance of being able to trust a source

At the end of 2022, Google’s quality rater guidelines evolved from E-A-T to E-E-A-T, adding first-hand experience to the concepts of expertise, authoritativeness and trust.

Google makes clear in its documentation on helpful content that E-E-A-T should not be interpreted as a single ranking factor.

There is no button that allows us to “activate E-E-A-T”, nor a list of four elements that we can simply insert into a page.

What matters is the direction: Google tries to use enough signals to determine whether a source demonstrates experience, expertise and trust.

And that trust becomes even more important when an artificial intelligence system does not merely display a link, but can directly use that information to construct an answer.

2023: the GEO concept is formally introduced

In November 2023, the academic paper GEO: Generative Engine Optimization was published by researchers affiliated with Princeton University, Georgia Tech, the Allen Institute for AI and other institutions.

The study helped formalise the term Generative Engine Optimization to describe different strategies intended to improve content visibility within answers produced by generative engines.

The researchers developed GEO-bench and tested different content modifications. In certain experimental scenarios, they observed visibility improvements of up to approximately 40%.

It was a question that everyone working in search visibility was beginning to ask:

if we have learned how to optimise a website for Google, can we do something similar for ChatGPT, Gemini, Perplexity and other generative engines?

The answer is yes, but three years later we also know that the problem is considerably more complex than applying another checklist.

There is no new keyword density for GEO

A 2026 academic review covering dozens of studies published since 2023 introduced an important qualification.

The well-known improvements reported by the original study occurred in an experimental environment where certain sources had already entered the context available to the system.

That allows researchers to study which characteristics may make a retrieved source more prominent within an answer, but it does not prove that modifying an article in a particular way guarantees that it will be discovered, retrieved and cited from among millions of possible documents.

The 2026 academic review of GEO argues precisely that generative visibility should be understood as a process involving several stages: discovery, retrieval, selection, context, citation and answer presentation.

This is very similar to what happened for years in SEO.

Looking for the next trick is tempting. However, no isolated technique can replace a complete strategy.

2024 and 2025: generative search leaves the laboratory

In 2024, the change stopped being mainly academic. Google began rolling out AI Overviews and integrated Gemini models adapted to Search with its traditional search systems.

The classic model could be summarised approximately like this:

query → results → click → page.

Generative search introduces another possibility:

query → retrieval from different sources → synthesis → answer → possible clicks.

The difference is enormous.

Google is no longer only a machine that decides which pages we should visit. It can begin using information from those pages to directly construct part of the answer.

AI Mode followed in 2025, offering a much more conversational experience capable of breaking a question into different subqueries before producing an answer.

And this change is not limited to Google. ChatGPT, Perplexity, Gemini, Copilot and other systems combine model knowledge with information retrieved from the web in different ways.

I have explained this process in more detail in my article on how ChatGPT searches the web, because understanding the difference between model knowledge, source retrieval and answer generation is essential to understanding GEO.

2026: GEO is no longer an experimental issue

If 2023 was the year in which we began putting a name to GEO and 2024-2025 represented the expansion of generative search, by 2026 it is difficult to consider it a secondary issue.

Google has continued expanding AI Overviews and AI Mode, while millions of users now turn directly to assistants such as ChatGPT, Gemini and Perplexity for questions that only a few years ago would almost certainly have started in a traditional search engine.

The most important change is not only technological. It is behavioural.

A person no longer necessarily needs to run a search, open five tabs, compare several pages and mentally build a conclusion. They can ask directly for a comparison, request more detail on a specific point, introduce new conditions and continue a conversation around the same problem.

This changes the way information is discovered and, therefore, also changes the work of those of us whose job is to improve its visibility.

SEO and GEO in 2026: what is actually changing?

SEO and GEO are not separate disciplines. Nor are they exactly the same thing.

The main difference lies in what we are trying to achieve and at what level visibility occurs.

	Traditional SEO	GEO
Main objective	Achieve visibility for URLs in search results	Achieve visibility for information, brands and entities within generative answers
Main unit	Page or URL	Source, information, entity and brand
Important signals	Relevance, content, architecture, links and authority	All of the above, plus corroboration, mentions, citability and distributed presence
Visible result	Ranking, snippet and click	Mention, citation, recommendation, answer and possible click
Measurement	Rankings, impressions, CTR, traffic and conversions	Mentions, citations, share of voice, sources used, traffic and conversions

Google itself continues to insist that good SEO practices remain useful for its artificial intelligence experiences. There is no version of SEO that we can simply abandon and replace with GEO from scratch.

GEO relies on a significant part of the infrastructure we have spent years building through SEO.

On-page SEO remains the foundation

The emergence of generative systems can create the impression that optimising our own website has become less important. In reality, the opposite is true.

For a page to be retrieved, interpreted and potentially used as a source, its information needs to be accessible and understandable.

We still need:

• a clear website architecture;
• correct crawling and indexing;
• coherent internal linking;
• original content;
• strong topical coverage;
• clear and easy-to-interpret information;
• structured data where appropriate;
• images and video when they add useful information;
• a good user experience;
• and pages that genuinely satisfy the need behind a query.

Visual content is also becoming more important within this evolution. In September 2026, Google added a dedicated report to Search Console for multimodal web searches, including queries made through Lens, Circle to Search, images uploaded to Google Search and other forms of visual search. This information appears both in traditional performance reporting and in reports related to generative AI features, another indication that search visibility is becoming less exclusively text-based. The announcement can be reviewed in the documentation published by Google Search Central.

On-page SEO has not disappeared with GEO. It has become the infrastructure that helps our information be found, understood and retrieved by generative systems as well.

Off-page SEO has not disappeared either: the objective is changing

Links still have value. Google continues to analyse the relationships between pages and link building remains relevant in competitive markets.

What changes is that, in GEO, thinking only in terms of backlinks is too limited.

It makes more sense to talk about the external authority of an entity.

An artificial intelligence system can find information about a company on its own website, but also in media coverage, comparisons, interviews, videos, specialist directories, associations, studies, reviews or third-party content.

This introduces an important difference.

For many years, an off-page campaign could be framed around one question:

How do I get this website to link to my page?

Now we should add another:

How do I get the relevant sources in my industry to associate my brand with the topics in which I want to become a reference?

Brand mentions acquire a different kind of value

One of the studies that best illustrates this shift was carried out by Ahrefs across approximately 75,000 brands.

They found a correlation of 0.664 between brand mentions across other websites and presence in AI Overviews. The correlation with the number of backlinks was considerably lower, at approximately 0.218.

The full data can be reviewed in the analysis published by Ahrefs on brands and AI Overviews.

These figures need to be interpreted correctly: correlation does not imply causation.

We cannot claim that gaining twenty mentions will automatically improve twenty positions in ChatGPT or Google.

But an interesting pattern does emerge: a brand’s distributed presence across the web appears to have an important relationship with its visibility in generative systems.

This changes the way we think about off-page SEO.

A backlink is still highly valuable. But so can be an editorial mention, an interview, an original study cited by third parties, a review, a podcast, a comparison, a collaboration with another organisation or a YouTube video where our brand is naturally associated with a particular topic.

From link building to building entity authority

I believe this will be one of the major transformations in SEO work over the coming years.

It is not about abandoning links, but about expanding the objective.

A link is no longer necessarily the final outcome of the action. It becomes part of something larger: building enough external signals for a brand to become recognisable and correctly associated with a particular field of expertise or activity.

SEO, branding, digital public relations, reputation, content, video and communication therefore begin to have far less clearly defined boundaries.

This also changes the tools we need to use. In my analysis of the best APIs for SEO and GEO, I explain how measurement is beginning to move beyond rankings and backlinks to include mentions, generative engines, competitors and sources.

First-hand experience will become increasingly difficult to replace

There is one point that I think deserves particular attention because I have been able to observe it over more than a decade of working on real SEO projects.

In 2013, it was still relatively easy to find a keyword, review what the top results were doing, produce a similar page, optimise it properly and strengthen it with links.

That does not mean SEO was easy, but the distance between producing content and contributing genuinely useful information was smaller.

In 2026, any company can generate in a few hours an amount of content that would have required weeks of work several years ago.

As a result, repeating information that already exists has less and less value.

The difficult part is no longer writing 3,000 words. The difficult part is being able to contribute something that other sources cannot simply reproduce by asking an AI system.

That is why I increasingly focus on elements such as:

• first-party data;
• real experience working with clients;
• results achieved;
• internal methodologies;
• well-founded professional opinions;
• mistakes and lessons learned;
• specific examples;
• case studies;
• research;
• and original audiovisual content.

An AI system can summarise a hundred articles about SEO. What it cannot legitimately invent is the experience accumulated over years of working on real projects.

That first-hand knowledge is precisely one of the clearest ways for a source to differentiate itself.

The entity matters as much as the page

Another important change is that we should no longer think only in terms of URLs.

When someone searches for a company or professional, Google and generative systems try to determine who that entity is, what it does, where it appears, which sources talk about it and which topics it is associated with.

This is why maintaining a coherent presence matters.

In my case, for example, it is not enough for a single page to talk about SEO. It also matters that different sources consistently connect Javier Carmona Benítez with digital marketing, SEO, GEO, artificial intelligence and education.

I have been working professionally in digital marketing since 2013 and, since 2018, I have taught subjects related to marketing and search visibility at the Universitat Politècnica de València. That information has far more value when it does not appear in isolation on a single page but forms part of a recognisable and coherent digital entity.

This is why we need to take care of the information presented across our website, professional profiles, publications, media appearances, content and external platforms.

On my own website, I also group together different articles about artificial intelligence that help develop this topical specialisation beyond a single commercial page.

Ranking no longer means simply reaching a position

For many years, one of the easiest SEO metrics to understand was ranking position.

Being first was better than being second. Being second was better than being fifth.

That remains useful, but it no longer explains the full visibility of a brand.

We can now encounter very different situations:

• a page ranks very well organically but does not appear in a generative answer;
• another source may be used by an AI system even though it does not hold the top traditional organic position;
• a company may be mentioned even if the source being used is a third-party website;
• a brand can gain visibility within a comparison without immediately receiving a click;
• and a source may influence an answer even if the user never visits it.

This means our metrics need to expand.

In addition to rankings, impressions, traffic and conversions, we increasingly need to analyse brand mentions, citations, share of voice, sources used, branded search demand, the queries in which our entity appears and referral traffic from generative tools.

Not all of these metrics will have the same value for every business, but the concept of visibility has clearly expanded.

The click problem: ranking first no longer guarantees the same traffic

Generative search introduces another consequence.

An answer can resolve a significant part of the user’s need before they visit any website.

Ahrefs has published several analyses on the impact of AI Overviews on organic CTR. Its data has shown significant reductions in click-through rates for queries where this type of answer appears.

The figures vary depending on the sample, time period and methodology, so it makes little sense to turn a single percentage into a universal rule.

What really matters is something else:

ranking, visibility and traffic no longer necessarily have the direct relationship they had years ago.

A brand can create impact before receiving a click. It can be mentioned, considered within a comparison or become one of the sources used to construct an answer.

This does not remove the importance of traffic. In many cases, conversions still require the user to ultimately reach our website.

But the journey leading to that click is becoming considerably more complex.

How I would approach an SEO and GEO strategy in 2026

After analysing this evolution, I would currently divide the work into three broad layers.

1. A technically solid SEO foundation

Before talking about ChatGPT, Gemini or AI Overviews, we still need search engines to be able to work properly with our website.

Crawling, indexing, architecture, internal linking, content, semantics, structured data and user experience remain fundamental.

If a machine cannot access or correctly interpret our information, we are introducing problems before we even reach the generative stage.

2. Content that genuinely contributes information

The second layer is about producing content that has a real reason to exist.

It is not enough to rewrite what a hundred other pages have already explained.

We need to contribute data, experience, examples, research, real cases, methodologies, properly developed comparisons, original analysis or specialist knowledge.

The easier it becomes to produce generic content with AI, the greater the relative value of knowledge that only certain sources can provide.

3. Authority distributed beyond our own website

The third layer is the one that most clearly connects SEO, off-page work and GEO.

We cannot depend entirely on what our own website says about us.

Our brand also needs to exist within the information ecosystem of its industry.

That means media coverage, mentions, associations, videos, interviews, collaborations, studies, reviews and other sources capable of strengthening the relationship between our entity and particular topics.

A link is still excellent.

But the objective does not necessarily end with obtaining that link.

The goal is for the internet to contain enough consistent signals to understand who we are and why we may be relevant within a particular field.

The opinion of Javier Carmona Benítez

I began working professionally in digital marketing and SEO in 2013 and, after all these years, I believe one of the mistakes we repeat most often in this industry is turning every technological change into the supposed death of everything that came before it.

It happened with Panda, Penguin, Hummingbird, RankBrain, BERT, featured snippets and Helpful Content, and it is happening again with ChatGPT, AI Overviews and GEO.

The reality is usually much less dramatic.

Previous signals do not necessarily disappear. They become integrated into more complex systems and take on a different context.

Keywords still matter because people still use language to express what they need. On-page SEO still matters because a machine needs to understand the information we publish. Links still matter because they connect documents and continue to provide information about the structure and authority of the web.

Technical SEO still matters. Authority still matters. Content still matters.

But we now need to add new questions.

Does Google understand who our company is? Which topics do external sources associate with our brand? Is there enough information to corroborate what we claim? Do we publish content that genuinely deserves to be used as a source? Do we appear on the websites an artificial intelligence system consults when researching our industry? Is our entity coherent both on and beyond our own website?

This is why I consider GEO to be the next logical step in search visibility, even though it would not be accurate to say that SEO has simply turned into GEO.

At first, we tried to make sure a page used the right words. Then we worked on its authority through links. Later, we learned to satisfy intent, develop topics and build entities.

Now we need systems to be able to find our information, understand it, cross-check it, connect it with our entity and consider it useful and trustworthy enough to become part of an answer.

And this is, for me, the main conclusion of this entire evolution:

In 2026, on-page SEO continues to build relevance, context and retrievability; off-page SEO continues to provide authority and external signals; and GEO expands both concepts by requiring us to work on presence, mentions, corroboration and the authority of our entity across the entire digital ecosystem.

We do not have to choose between on-page SEO, link building and GEO. We need to understand the role each of them plays within a much more complex search environment.

Through my SEO agency, I work with precisely this combined approach: maintaining a strong technical and content foundation, building authority and gradually adapting the strategy to new forms of AI-powered search and recommendation.

You can also review my specific services as a GEO and AI visibility consultant or find me on Google Maps to see my profile, location and reviews.

Perhaps in a few years we will stop debating whether we should call it SEO, GEO, AEO, LLMO or use whatever other acronym appears over the coming months.

Because the underlying objective remains fairly similar to the one we had when I started in 2013, even though the technology is now far more sophisticated:

to ensure that when someone needs information, a product, a service or a professional related to what we do, our brand is one of the answers worth showing.

fusion_global=”17304″]