SEO and GEO: Be found by Google, cited by AI

Search engines still send the bulk of the traffic, while AI assistants increasingly decide what gets cited. They work together, but in different ways.

Since AI-generated responses now interject between a question and a site, a concern arises: should we abandon traditional SEO? No. But we must stop treating it alone.

The SEO aims for a position in a list of results. The GEO, to Generative Engine Optimisation, aims at something else: being the source a model chooses to cite when it writes its response. The levers overlap significantly, but not entirely.

The technical foundation has not changed

Before any content strategy, a website must be machine-readable. This foundation is not new, and it remains the first blocking factor we encounter.

  • Pages that load quickly, even on a 4G phone
  • Content present in the HTML, not solely built by JavaScript
  • A stable, explicit address per topic with a canonical tag.
  • A coherent hierarchy of titles: a single H1, H2s that truly break it down
  • Structured accurate data: organisation, article, frequently asked questions, Ariadne's thread
  • An up-to-date site plan and robots.txt that does not block anything useful

A website that fails on these points is not poorly ranked: it is partially invisible to both search engines and models.

What search engine optimisation demands today

Logic has not changed its nature: responding better than others to a real intention. However, three evolutions deserve to be taken seriously.

First, intent trumps keyword : A single page that thoroughly addresses a topic is more effective than three pages that merely repeat the same expression. Then, experience matters : speed, display stability, and mobile readability have become requirements, not bonuses. Finally, the credibility is evaluated : A credible author, a clear date, a reliable source, and a real company behind the website carry more weight than they did five years ago.

What the engines really want

Google states that its AI features are based on the fundamentals of SEO: an accessible, indexed, useful, and admissible-to-extraction page. It requires no special AI file or particular structured data. The file llms.txt can serve as a complementary directory; its presence does not guarantee any citation.

FAQ and structured data should describe visible content. Multiple H2 headings on a page are normal when they organise distinct sections. Decorative images should keep empty alternative text; only informative images require a description.

Sources: Google Search Central: AI features and Google SEO launch guide.

GEO: be cited by a model

An assistant who writes a response does not gather ten links: it selects a few sources and extracts passages from them. Optimising for this behaviour requires writing differently.

  • Respond early and explicitly. The answer should appear in the first lines, not after an introduction of three paragraphs.
  • Break down into extractable units. A title = a question. A paragraph = a complete idea, understandable out of context.
  • Provide verifiable elements. Numbers, definitions, numbered procedures, comparisons: these are the snippets that are cited.
  • Become identifiable. Legal notice, contact page, structured data entity, consistent presence elsewhere on the web.
  • Stay up to date. A visible update date and genuinely revised content significantly increase the chances of being selected.
  SEO GEO
Objectives To be ranked To be cited
Useful unit The Page The transition
What is evaluated Relevance and authority Clarity, verifiability, freshness
Measurement Positions, clicks, impressions Mentions and citations in responses
Pace Several months Variable, sometimes rapid

What they have in common

Precise, structured, dated, and attributable content is better indexed by a search engine and better cited by a model. The essential work is shared.

This is good news: there is only one editorial requirement to fund, served by a correct technical foundation, followed by some format adjustments for GEO.

What to start with, in order

  1. Improve the technical foundation: speed, structure, structured data, site plan.
  2. Clarity: what you do, for whom, in what words, on the home page as well as on the offer page.
  3. Write five to ten pages that answer the real questions your clients ask before making a purchase.
  4. Make the company identifiable: Legal notice, contact information, author, consistent presence elsewhere.
  5. Measure: Search Console for SEO; regular manual searches on assistants for GEO.
  6. Maintain a realistic pace: a monthly publication kept up is better than six abandoned.

Order of Magnitude

The technical foundation can be corrected in a few days. Positioning and the first pages take several weeks. Effects on positions can be seen from three to six months: any faster promise deserves caution.

To keep in mind

  • The technical foundation conditions everything else: start there.
  • Write for an intention, not for a keyword.
  • To be cited by an AI: respond early, break into autonomous segments, provide verifiable information.
  • Make the company identifiable: it's a criterion, not a formality.
  • An editorial strategy, two output formats: the work is largely collaborative.

Your next step

Hidden despite real expertise?

Show us your current site. We will identify what is blocking you and where to start.

Book a call

To go further

01
Decision

Automate, buy, or build

A simple grid to compare the frequency of the need, its differentiation, the risk, and the operational cost.

Read the note
02
Useful AI

From AI testing to controlled usage

A useful pilot links a task, an authorised data point, human validation, and an observable business metric.

Read the note
03
Product

Five questions before a first SaaS

Primary user, recurring problem, critical data, responsibility, and future cost: the foundation of a reasonable V1.

Read the note