Here is the interview I had the opportunity to give on a French blog run by Olivier Sauvage.
Olivier:
Aurélien Bardon, a leading SEO figure in northern France
I have known Aurélien for a long time. He is one of the most recognized figures in digital entrepreneurship in northern France. Since we had some time and I have recently been very interested in GEO (Generative Engine Optimization), we thought it would be worthwhile to discuss the topic so that Aurélien could share his perspective as a nationally recognized SEO specialist.
What I appreciate about him is that he speaks frankly and does not try to sell himself at every opportunity. I knew in advance that this interview would be valuable and worth reading for anyone who genuinely wants to understand the current state of SEO and GEO in France in 2026.
Aurélien Bardon, a discreet and respected SEO expert
SEO expert: Aurélien Bardon is a leading figure in French SEO with more than 20 years of experience.
Entrepreneur: He founded the agency Aseox as well as the Oseox monitoring tool suite.
Educator: He is widely known for producing accessible guides and analytical tools for web professionals.
Speaker: He is an active member of the French-speaking SEO community and regularly shares his expertise through his blog and industry events.
What is GEO today?
Olivier: Hi Aurélien, thank you for agreeing to answer my questions. You know I’ve been very interested in GEO lately, but I’m far from being an expert. I thought the best approach would be to discuss it with you. I would like your opinion on AI and GEO, especially following Google’s recent statement that seems to position AI within the broader SEO landscape. Where do you stand on GEO today? What does it mean to you?
There are no GEO experts
Aurélien Bardon: It is essential to remain humble. There are no GEO experts because the field is so new and constantly evolving that what we claim today could be proven wrong tomorrow. It is a major and unavoidable trend, but there is also a great deal of misinformation. Service providers are adapting, often learning as they go, to meet the growing demand for visibility within LLMs.
Large organizations are justified in preparing for these changes because they have the resources to experiment without the same immediate ROI constraints faced by small e-commerce businesses.
However, we must be careful not to make unrealistic promises. In SEO, results are never guaranteed. In GEO, guarantees are even less possible because of the speed at which the landscape changes.
Believing GEO will deliver rapid ROI is currently an illusion in France
If the goal of GEO is to increase revenue or website traffic, people will be extremely disappointed. Despite spectacular month-over-month growth in AI usage, the contribution of LLMs remains insignificant compared to traditional search engines. The traffic generated today is negligible.
My position will change when traffic curves become meaningful. Some SEO professionals already acknowledge that traffic is no longer the relevant KPI. That is an honest assessment because LLMs do not generate enough users, provide very few links, and click-through rates on those links are extremely low.
An alternative approach is to focus on brand awareness by seeking citations.
When a brand is mentioned, it contributes to its visibility and long-term recognition. However, the question of scale remains. How much exposure does the brand actually receive? The ROI must still be compared against other marketing channels.

Measuring GEO performance in LLMs
Olivier: Is measurement reliable? My impression is that citation monitoring services, especially in the United States, are very uncertain and provide little real guarantee.
Aurélien Bardon: Measurement is indeed extremely complex. Several factors come into play.
- Personalization and variability: LLMs adapt their responses to user profiles and context. Even without personalization, responses are not always identical. Monitoring something so variable is challenging.
- Technical constraints: Available computing resources may vary, potentially affecting model performance at different times of the day. Monitoring tools must also reflect real-world usage. Using an API does not necessarily produce the same results as the consumer-facing interface.
- Black-box systems: Model updates, even when unannounced, can change outputs and citations in unpredictable ways.
The most honest providers attempt to address this issue through coverage KPIs, running large numbers of prompts throughout the day to estimate citation rates. It is a valid methodology, but it quickly becomes expensive.
Another challenge is deciding which prompts to monitor.
Unlike SEO keywords, we have no data on prompt popularity. We are left filtering existing SEO question databases, but LLM search volume remains very small anyway, roughly 300 times lower than Google’s.

Effective GEO optimizations
Olivier: Which GEO optimizations actually work?
Aurélien Bardon: GEO best practices are simply traditional SEO best practices. There is a lot of hype around techniques such as microformats. Although we have used structured data and schema.org in SEO for more than 15 years, their impact on LLM citations is often negligible.
The approaches that work are much simpler.
- Get mentioned: Identify the source pages that LLMs cite for a specific query and secure mentions on those pages.
- Fanout optimization: This is the most interesting aspect. An LLM operates in two modes:
- Internal knowledge base: direct responses based on training data up to a specific cutoff date.
- Active search mode: for recent or unknown information, the LLM breaks the query into key sub-expressions, known as fanouts, and queries a search engine.
Despite various commercial partnerships, ChatGPT quickly favored Google because of its relevance. The challenge is therefore twofold: rank on the SERPs generated by these fanouts and then be cited among the pages scraped by the LLM.
The new SEO market: buying citations
Olivier: Ultimately, this sounds like classic SEO, producing content or obtaining mentions on external websites. My conclusion is that we are moving from buying backlinks to buying unlinked citations.
Aurélien Bardon: Exactly. This is becoming the new market for link-building platforms, often at a slightly lower cost.
Studies show that the best-performing content is content that makes information extraction easier for algorithms: comparisons, precise data, and highly readable formatting such as tables and concise sentences. Once again, a good SEO article already includes clarity, UX, and accessibility.
However, it is important to acknowledge the major impact of LLMs on SEO. Google is frustrated by competitors such as ChatGPT scraping its results and has taken concrete action.
The company has hired specialists to combat scraping, making life more difficult for SEO professionals who rely on scraping to monitor rankings. Google has even adopted a strategy previously used by Amazon: returning false data to scrapers in order to contaminate their datasets, while indirectly favoring its own services such as YouTube.
It has become an open conflict.
Organic results are becoming less important to Google
Google appears increasingly less interested in SEO professionals and organic search results. The negative impact is already visible. Traffic for certain types of queries, such as simple tutorials or documentation pages, is disappearing permanently, as illustrated by the experience of Stack Overflow.

Olivier: Are you already seeing similar traffic declines in France, as some media companies have experienced in the United States?
Aurélien Bardon: The impact remains limited in France, primarily because features such as AI Overview have not yet been fully deployed. Regulatory and political factors currently provide some protection.
We must also differentiate between countries and website categories. News publishers are already affected and are increasingly turning toward direct partnerships, such as those between major newspapers and ChatGPT.
This represents a shift toward contractual relationships and trusted feeds rather than purely algorithmic distribution.
The greatest risk is not where most people expect it
Traffic decline is inevitable in the long term, but the biggest immediate threat for French SEO professionals comes from Google’s own interface changes. Some of my clients are losing traffic despite ranking first for their niche keywords.
The reason is simple: click-through rates are falling because Google pushes organic results further down the page with advertisements, maps, and videos in its pursuit of growth.
This erosion is no longer offset by increased Google usage. Our greatest risk is that Google fills the page with advertising blocks and users never scroll down to the organic results. Even so, SEO continues to deliver historically strong ROI.
Olivier: Regarding LLMs, what strategy should businesses adopt considering the growing importance of Reddit and LinkedIn?
Aurélien Bardon: LLMs explicitly target Reddit results within their fanouts through the site:reddit.com operator. This is a deliberate choice by developers and is reinforced by the financial relationship between Reddit and Google.
Ironically, the large-scale automatic translations of Reddit content for French audiences are often extremely poor, with English expressions translated literally.
SEO professionals are asked to produce high-quality content while Google simultaneously promotes poorly translated material.
Reddit is valued because of its spontaneous and in-depth content, preserving the spirit of early internet forums, unlike the standardized and overly polished content often found on LinkedIn. Reddit has also implemented scoring systems that reward quality contributions.
It is one way of bypassing traditional commercial marketing. Naturally, SEO professionals have already adapted by offering paid citation placements through established Reddit accounts.

Conclusion: how should companies approach digital marketing in 2026?
Olivier: To conclude, what overall recommendation would you give to an SME regarding its marketing strategy?
Aurélien Bardon: GEO is not yet the top priority in France, but companies should keep it on their radar.
Any organization with sufficient resources should engage in monitoring and exploratory investment. We are facing a major transformation that cannot be ignored.
If Google decided to promote Gemini by default, usage could immediately surge to billions of daily queries. Two factors currently limit that scenario.
- Costs: Processing an LLM query is significantly more expensive than processing a traditional search query. Companies are waiting for inference costs to decline.
- Advertising: Google must find a profitable way to integrate advertising into LLM-generated responses without cannibalizing clicks on existing ads.
This transformation is societal in nature and affects every industry.
We are living through a period of creative destruction reminiscent of the early days of the web, but even more disruptive. Many people feel paralyzed by uncertainty regarding the future of traditional skills and professions.
Olivier: Aurélien, thank you very much.
