Periscopy
Articles
Methodology7 min read

SEO and GEO: the base matters, and it is no longer enough

SEO is what lets an engine find and read your site. It is not what makes a model cite your brand. The order matters: base first, then the layer on top.


When the talk turns to visibility in ChatGPT, Claude or Google's AI Mode, the question that comes up most is whether SEO still counts. It does, a great deal. But the right question is a different one: is SEO on its own enough to appear in the answers AI gives?

The short answer is no. The long answer is what follows, and it starts with what SEO still does well.

What SEO does well, and goes on doing

Technical SEO, indexing, schema, performance, extractable HTML, link architecture: that is the engineering that lets a search engine find, read and understand your site. None of it changed. Quite the opposite: in June 2026 Google itself formalised, in the Search Central documentation, that its AI features rest on the same core ranking and quality systems. Whoever has reasonable SEO has the foundation Google uses to choose what to cite.

In classic SEO that work serves to climb the ten blue links. It is the game most marketing teams already know how to play.

Where SEO alone falls short

Being at the top of Google is one thing. Being cited by the AI is another, and the two are less and less the same.

Recent public research puts the overlap between Google's top organic results and the sources AI cites in its answers as having fallen from around 70% to under 20% in a few years. In other words: in most cases, whoever the AI cites is no longer whoever sits at the top of SEO.

The reason is structural. In SEO the engine optimises to show a list of options and the user chooses, so your job is to draw attention. In GEO the engine optimises to give one answer with two or three brands inside it, so your job is to be one of those brands. They are different games played on the same board, and winning one does not guarantee winning the other.

The missing layer: GEO

GEO is the layer built on top of SEO. Where SEO works on the site, GEO works on the brand's citability as an entity:

  • Entity authority: the brand appears consistently on Wikipedia, Wikidata, Crunchbase, G2, sector directories, with sameAs linking the profiles. The models recognise who it is.
  • Mentions in sources the models trust: sector publications, comparisons, transcribed podcasts, qualified Reddit. This is where the AI takes most of what it knows about your category.
  • Extractable, citable content: clean definitions at the top, sourced data, snippable paragraphs. Different from content optimised for the click.
  • Measurement of your own: citation rate and share of voice per engine, over real questions from the sector. Not SERP position.

None of this competes with SEO; all of it rests on it. Without reasonable indexing, the best external authority never gets read. With reasonable indexing but no entity work, the site exists but the brand does not exist for the model.

Scope: the AI is not only Google

One important note that many articles skip: the reading above, SEO fundamentals as the solid base of GEO, is particularly true in the Google ecosystem (Search, AI Mode, Gemini). Other generative engines work differently.

  • ChatGPT: combines training knowledge with web search through Bing. Technical SEO helps on the web-search side; entity authority weighs more on the training side.
  • Perplexity: real-time retrieval, with direct citation. Close to classic SEO in some mechanics, but with less weight on ranking and more on source quality.
  • Claude: uses live web sources and has a strong editorial-quality signal. Less sensitive to on-page tactics; more sensitive to entity coherence.

Generalising “the AI” as if it were all Google is an expensive mistake. Honest GEO work is done engine by engine: what serves each one, what each one rewards, what each one ignores.

What this means in Portugal

In our study of AI visibility among Portuguese B2B SaaS we measured 45 companies across 4 engines. Several of them have reasonable SEO, some of them good, and even so:

  • 56% are recommended by none of the four engines when you ask about their category.
  • 31% are completely invisible: the models do not even name the brand.
  • 96% have no defined policy for AI crawlers, the mark of a decision nobody ever took.

What these companies are missing is not, in most cases, more SEO. It is the layer above, the GEO layer. And whoever invests now occupies thinly populated ground; eighteen months from now that window is not the same one.

To close

It is not SEO or GEO. It is SEO as the base, GEO on top. The two games are played on different boards but with shared pieces. Whoever only does SEO is building foundations with no house; whoever tries GEO without an SEO foundation is putting a house on sand.

For a Portuguese company today the most defensible path is that one: make sure the technical SEO is where it should be, and build the model-citability layer on top of it. The order matters.

See also: Google made GEO official (what Google says, in June 2026, about what works and what does not) and GEO vs SEO in 2026 (a wider reading, with a comparison table). Sources: Google Search Central documentation (May and June 2026); recent public research on the overlap between the organic SERP and the sources AI cites; destaque.ai 2026 study.