Google no longer holds a monopoly on online search. Queries are increasingly being made on AI engines like ChatGPT, giving rise to a new discipline: GEO. Or how to optimize your presence on these new tools. SEO expert and Isarta trainer Myriam Jessier breaks it all down.
Can you remind us what GEO is?
Myriam Jessier: Generative Engine Optimization (GEO) is the discipline of optimizing an organization’s presence in the synthesized responses generated by AI engines: ChatGPT, Gemini, Perplexity, Claude, and Google’s AI Overviews.
It’s not SEO in the traditional sense. To use an analogy: SEO is the librarian who hands you a book, while GEO is the detective who reads everything for you and delivers a summary — without you ever having to open the book.
The fundamental mechanism is Retrieval-Augmented Generation, or RAG. The AI engine works in two stages:
- it retrieves the most relevant documents for a given query,
- then synthesizes them to produce a response, selecting between two and seven sources to cite.
GEO optimizes precisely that second stage.
How is GEO similar to — and different from — SEO?
Myriam Jessier: To extend the analogy: the detective needs the books the librarian has on hand. It’s a team effort. SEO and GEO are complementary layers of the same visibility strategy.
They don’t conflict: 48% of AI citations come from the top 100 organic Google results. Without a solid SEO foundation, a site never enters the RAG pipeline. But where SEO stops at the door of a list of links, GEO goes inside the very structure of language.
| Dimension | SEO | GEO |
|---|---|---|
| Objective | Rank in the SERP, clicks to the site | Being cited as an authoritative source in the AI response |
| Target interface | Results pages with link lists | Conversational interfaces and text summaries |
| Query nature | Short keywords, ~4 words on average | Long conversational queries, 23+ words on average |
| Success metric | Position, CTR, organic sessions | Citation rate, AI share of voice, response sentiment |
| User behavior | External browsing and evaluation | Immediate consumption, often without outbound clicks |
| Content format | Keyword-optimized pages | Modular, dense, programmatically extractable content |
The most important divergence is conceptual: in SEO, authority is built through inbound links and domain authority.
In GEO, authority is earned through factual accuracy, structural clarity, and the content’s ability to serve as irrefutable evidence for a language model. It’s a shift from a culture of the click to a culture of the citation.
Is an organization’s website indexed by AI tools by default?
Myriam Jessier: No, not by default. And the most common mistake is to assume that Google indexing automatically guarantees a presence in AI tools.
LLMs were trained on a vast amount of data. Your site may have existed for ten years and be completely absent from the picture a model has of your organization. Worse: it may be present with incorrect, outdated, or competitor-attributed information. Search engines rank what is said. LLMs evaluate what you mean to say.
Real-time visibility in answer engines depends on active optimization. In 2026, management starts with AI crawlers, which each play a distinct role:
| Crawler | Organization | Role |
|---|---|---|
| GPTBot | OpenAI | Training future ChatGPT models |
| OAI-SearchBot | OpenAI | Real-time browsing for SearchGPT |
| PerplexityBot | Perplexity AI | Indexing for direct citations |
| Google-Extended | Controls access to Gemini (separate from Google Search) | |
| ClaudeBot | Anthropic | Claude’s knowledge and research |
What’s the difference between a Google search and a ChatGPT search?
Myriam Jessier: Google excels at quick factual lookups, local navigation, and accessing specific sources. ChatGPT has become the go-to tool for synthesis, complex problem analysis, and information processing.
The two coexist: in 2025, Google maintains a traffic volume ratio of 14 to 1 compared to ChatGPT. 95% of ChatGPT users also use Google, according to intent-segmented usage data. Google is used to find a source. AI is used to process the information from that source.
Should I expect an inevitable drop in traffic?
Myriam Jessier: Yes, it’s real and it’s accelerating. This erosion is not uniform: it hits hardest the sites whose value rests on simple factual information that AI can summarize without losing nuance. Articles written purely for SEO with no real value added for users are the most affected.
2025 data indicates that visitors referred by AI platforms convert at a rate roughly four times higher than visitors from traditional search. AI acts as a qualification filter: when a user clicks a source link after reading a summary, their purchase or conversion intent is already far more mature.
The real challenge for an organization is therefore not to defend its traffic volume. It’s to shift from a volume strategy to an influence strategy within the generative conversion funnel. Visibility in AI responses is a metric in its own right, distinct from organic traffic.
A user who learns your brand name through ChatGPT and then searches directly for your site generates a direct visit that is invisible in your standard analytics. That’s real value that current measurement tools systematically underestimate.
Is GEO more technical than SEO?
Myriam Jessier: GEO shifts the technical complexity: away from JavaScript rendering and Core Web Vitals, toward semantics, knowledge structure, and RAG mechanics. It’s not necessarily harder. It’s different. AI doesn’t read words — it maps concepts.
GEO requires understanding how a language model extracts, decontextualizes, and cites content. These are two distinct areas of expertise that coexist within a complete visibility strategy.
Is content still as important as ever?
Myriam Jessier: It has never been more important — but its ideal form has changed. To perform well in GEO, text must be designed for machine readability. This means adopting semantic triplet structures: Subject → Predicate → Object.
For example, instead of “Our solution improves your efficiency,” write: “Software X reduces operational costs by 15% for small businesses managing more than 50 employees.”
The five criteria that define content quality in the age of LLMs:
- Structural adequacy. Descriptive, non-metaphorical H1–H3 headings build a hierarchy that models can read. “Our innovative solutions” is invisible. “Operational cost reduction for SMBs” is extractable.
- Information density. Given the grounding budget of ~380 words per page, every sentence should name an entity, state a relationship, or preserve a condition. Filler dilutes coverage without increasing selection.
- Extractability. Every important claim must be self-contained and understandable without surrounding context. Vague pronouns (“this,” “that,” “our solution”) are invisible to a model that retrieves in fragments. The right question to ask: can this sentence stand on its own in a ChatGPT response?
- Entity completeness. Name subjects and relationships explicitly. “He founded the company” creates no usable triplet. “Marie-Ève Tremblay founded Atelier Numérique Montréal in 2019” creates three usable knowledge nodes.
- Natural language quality. AI models penalize excessively repetitive or robotic structures. Precision and voice are not opposites. Precision is a voice.
Content matters more than ever. But the quality criteria have changed: it’s no longer “is this content optimized for keyword X?” It’s “can this claim be cited directly by an LLM, accurately and without distortion?”
How do you measure your GEO today (tools, audits)?
Myriam Jessier: GEO measurement is still a discipline taking shape. AI platforms don’t share their query logs with site owners, creating a feedback loop that is data-poor. But the tool ecosystem is consolidating quickly.
Key indicators to track:
| Indicator | Description |
|---|---|
| Citation rate | Percentage of times your site is cited for a set of strategic queries |
| Brand mention frequency | Number of times your name appears in generated responses, with or without a link |
| AI sentiment | The tone used by AI to describe your organization (positive, neutral, negative) |
| Share of voice | Your brand’s visibility vs. direct competitors in AI responses |
| AI Overviews visibility | Presence in the generative modules integrated into Google Search |
training.isarta.com 
