Article sponsored by Uplix

This is the question on everyone's lips, how we become visible in LLMs (ChatGPT, Gemini, etc.)

We will try to give very concrete advice and an example of results.

SEO is undergoing a profound transformation, this is not the first time.

John Mueller emphasizes, in the media seoclaims.com (the bible of Google's statements), that companies must rely on their own data analyzes rather than on trends like “GEO” to assess the real, still minimal, impact of AI on their traffic. Indeed, these remain new acquisition sources that require expertise.

For years, the goal was clear: appear in the first Google results. Be well positioned, capture clicks, generate traffic.

But Google is not the only channel, ChatGPT is on everyone's lips: we want Generative Engine Optimization (GEO) or put differently “I want my brand to be mentioned in ChatGPT”.

“I want my brand to be mentioned in ChatGPT”.

It's possible, but remember that ChatGPT says that as of July 2025, there were over 700 million weekly active users, sending around 2.5 billion requests per day. Accurate GPT even though approximately 70% of uses are made outside of a professional context.

In summary, this confirms the importance of AI in daily life, but also the need to maintain a critical mind.

With the emergence of LLMs like ChatGPT, Perplexity or Google's AI Overviews, the user is no longer looking for a list of results. He expects a direct, synthetic, ready-to-use response.

At the house of Uplixwe are lucky to have perspective on the techniques that really work.

From search engine to response engine?

Historically, Google worked like an intermediary. He proposed:

  • Multiple links
  • Several sources
  • Several interpretations

The user remained the decision-maker.

With LLMs, this logic gradually evolves. The interface changes:

  • A question
  • A unique (or almost unique) response
  • Sometimes some secondary sources

But above all, an implicit selection is made by the AI.

It is no longer the user who chooses, but it is the model which arbitrates.

This means a very simple thing: if you are not in the answer, you don't exist.

To understand how to appear in ChatGPT, you need to understand how it “reasons”.

An LLM does not work exactly like a traditional search engine, it will not “search for the best page”.

It will:

  • Predict a sequence of words
  • Rely on probabilities
  • Mobilize learned knowledge
  • Possibly enrich with external sources (RAG- Retrieval Augmented Generation)

Its goal is not absolute truth. Its objective is to produce the most plausible answer in a given context.

And that changes everything, because to be cited, you must be:

  • Likely
  • Credible
  • Consistent with the question

Not just well referenced.

What are the 3 levers to appear in ChatGPT?

With hindsight and concrete tests, we can structure things around three main levers. These are not theories, they are field observations.

1- Be present in the training data

It’s the most intuitive lever, and often the first we think of.

LLMs are trained on massive volumes of data: web pages, articles, documentation, public content.

Logically, the most historically visible brands are more likely to be known to the model.

This is why certain companies naturally stand out:

  • Big brands
  • Historical actors
  • Market leaders

But you have to be lucid. This lever has two major limits:

First, it is very difficult to operate, you cannot decide to be integrated into a Dataset training.

Then it is slow to operate. Building sufficient awareness can take a long time.

Conclusion: being “known” helps, but it absolutely does not guarantee being cited!

As proof, today, many well-known brands are rarely present in the responses. And conversely, some smaller ones manage to emerge.

2- Become a (citable) source!

It is the most operational lever today.

LLMs rely on existing content to structure their answers, particularly when using information retrieval systems (IRA).

But be careful: not all content is equal. Here is what we observe very clearly:

  • Structured content performs better
  • Direct answers are favored
  • Educational content is over-represented
  • FAQ or “guide” formats are very effective

In other words, it is not enough to produce content, you must produce “quotable” content.

Quotable content is content:

  • Clear
  • Well structured
  • Directly exploitable by an AI
  • Aligned with specific intention

This is exactly the logic behind an internal tool:

The goal is no longer just to please Google. The objective is to become a key answer for LLMs.

The tool is based on a simple logic: reproduce what LLMs do… but better.

From a strategic query (QFO- Query Fan Out), it analyzes the Google SERP in depth. Not only the organic TOP 10, but also its entire structure (Featured SnippetsAAP –People Also Ask, Knowledge Graph) to understand what Google considers the “best answer”.

Then, it goes further: it scrapes content from the best sources, identifies dominant angles, semantic gaps and opportunities for differentiation.

Based on this, it generates a ready-to-use GEO prompt, designed to force quoting: Query Fan Out to cover all intentions, citation hooks to encourage LLMs to return to the source, structured data enrichment (JSON-LD), and clear action plan.

Result : we don’t just produce content, we “manufacture” a source calibrated to be used and cited by AI.

3. Semantic alignment

This is probably the most powerful lever today. And paradoxically, the least well understood.

An LLM works with vector representations. It associates concepts, entities, contexts. When a user asks a question, the model activates a network of associations.

And in this network, certain brands are present. Others don't. This is where everything comes into play.

If your brand is not associated with the right concepts, you will never appear, even with good content.

From theory to implementation: concrete tests

In our example “best SEO consultants” works, Uplix is top 3 in Google.

I tested my recipe with my personal site edv.fr on prompts “Give the 5 best SEO consultants” or “Top 5 SEO consultants” and the results are clear (external tracking tool: meteoria here)

Let's take another simple example with a new prompt:

“What is the best SEO agency?”

The model will generate a response based on:

  • Semantic associations
  • Brands already linked to this notion
  • Signals of credibility

If your brand is not in this semantic space: it does not exist for the model.

This is precisely what working allows waikay.io : this tool no longer works only on keywords, but on relationships.

Objective : connect your brand to the right queries in the model's mental space.

And it works because for this example prompt, Uplix is top 3 in Google.

The often forgotten lever: authority

There is one point that many overlook. Even in a world dominated by LLMs, the notion of authority remains central.

A brand is much more likely to be included in the responses if it is:

  • Quoted
  • Mentioned
  • Relayed
  • Recommended

For what ? Because these signals reinforce its perceived credibility.

How ? This involves levers that we already know very well in SEO:

  • Networking
  • Press relations
  • Media presence
  • External citations

Classic methods are not disappearing. On the contrary, they serve to build a solid foundation on which to lean, because LLMs go through search engines!

So the SEO of your site will have a significant impact on your presence in LLMs, particularly via Query Fan Out Who are searches made by LLMs to sort the relevance of the results.

What we actually observe about authority

On the ground, the differences are impressive. Certain brands are systematically mentioned on dozens of strategic prompts and others are absent. The difference is not just about authority.

It is played on:

  • The structure of the contents
  • The ability to be taken back
  • Semantic alignment
  • Presence in the ecosystem

Many companies still approach LLMs with classic SEO logic. This is a mistake.

The main errors that we observe:

  • Produce generic content without distinctive value
  • Write for Google and not for AI answers
  • Ignoring actual user prompts
  • Do not work entities and associations

Result : no impact in the generated responses

The Bigfoot effect refers to the recent trend of artificial intelligence (such as ChatGPT Search) to massively favor a handful of very high authority websitesthus overwhelming the diversity of sources.

The visibility space generated by AI has been drastically reduced. Web giants are grabbing all the attention from LLMs, making it much more difficult for other sites trying to gain visibility.

We have therefore launched a specific offer to work on your present on Reddit, Wikipedia and other web giants!

The final word?

The brands visible today will capture a growing share of demand. This is a train to take quickly for a share of voice that is still low but in the making!

As in SEO, we no longer just seek to be visible. We seek to be selected, to convert. This requires:

  • A new understanding of models
  • A new way of producing content
  • A more global strategy

We have implemented complex processes to obtain objective results, but the logic remains the same as for SEO: technical, semantic and authority.More information on our site: https://www.uplix.fr/agence-geo-ia/ (Analysis of your needs without obligation)