Seeing your brand appear in ChatGPT, Gemini or another AI system can feel like part of the battle is already won.

It is reassuring. It means the business exists in the information environment the engine uses and is identifiable enough to be included in a response. But it does not yet tell you whether it will be chosen.

This is an important distinction, because AI does not just answer the question: “Which businesses exist?” It must also answer another, far more demanding question: “Which ones seem best suited to this specific situation?” And that is where visibility and recommendation begin to diverge.

Presence is only the first level

Take a hotel. It may rank well in search, appear across many sources, have a comprehensive website and be regularly cited in generated responses. All of this is positive. But if a traveller asks: “I am looking for a quiet hotel for a three-night stay as a couple, with a beautiful beach, little entertainment and a budget below €250”, the AI is no longer just looking for visible hotels. It must weigh up several properties and determine which seem to best match the request as a whole.

A hotel can therefore be clearly visible yet rarely appear among the top choices. There is no contradiction. It simply means that presence and preference are two different things.

Presence, relevance, preference

To analyse this phenomenon, we find it useful to distinguish three levels.

Presence

Does the AI know enough about the business to include it in its responses?

Relevance

Does it understand the situations in which this business could genuinely be a good fit?

Preference

The AI weighs up several options and decides that one offers a better answer to the specified context.

This distinction completely changes how GEO performance is measured. A business can have a strong presence but generate little preference. Conversely, a less visible property may be selected very regularly within a specific area because its positioning, experiences and available evidence align more closely with the request.

“Visibility measures presence. Recommendation measures preference.”

And the two do not necessarily improve together.

We explore the broader mechanisms in our reference article: what is GEO for tourism businesses?

Why might AI prefer a competitor?

When several businesses seem to meet the same request, AI must find reasons to distinguish between them. These reasons may come from positioning, the precision of the available information, the experiences described, the strength of the evidence, customer reviews or external sources that confirm certain characteristics.

The important point is that the business does not control this narrative alone. It may state on its website that it is particularly well suited to couples, families or travellers seeking peace and quiet. But if reviews, partners, media outlets or other sources mostly tell a different story, the signal becomes less clear. Conversely, when several independent sources converge, AI has more evidence to associate the business with a specific situation. This is often where preference is formed.

What we observe in audits

During a recent audit, we observed a property that appeared regularly in generated responses but was selected among the top choices much less often in certain scenarios. The problem was not a lack of visibility. The property was known, accessible and sufficiently documented. But for several specific requests, other businesses presented signals that were more consistent with the specified context. In other words, the property was present, sometimes relevant, but not sufficiently differentiated or validated to become the preferred choice.

It would be an overstatement to attribute this result to a single cause. The mechanisms behind generative engines remain complex and continue to evolve. But this type of gap reveals something essential: increasing presence is not always enough to increase recommendations.

What moves a business from relevance to preference

Preference is rarely built on a single signal. It tends to emerge when several elements reinforce one another. The business must first be clearly associated with certain situations. The vaguer its positioning, the harder it becomes for AI to know when to suggest it.

That positioning must then be supported by evidence. An experience described on the website, confirmed in reviews and echoed by external sources creates a much stronger signal than a simple marketing claim. Finally, these elements must remain consistent with one another. AI may encounter the official website, platforms, reviews, partner content or media coverage. If all these sources describe very different realities, its understanding becomes less reliable. Conversely, when they tell a compatible story, the business becomes easier to interpret and recommend.

What this changes for an executive

Measuring only a brand’s presence in AI therefore risks providing an incomplete picture. Knowing that your hotel or tour operator appears in 50% of responses may seem encouraging. But the real question lies elsewhere: “In how many relevant situations is it actually selected among the best options?” And, above all: “Why does the AI choose it — or choose someone else?”

This second perspective makes it possible to move from a simple visibility indicator to substantive positioning work. It also helps focus efforts on areas where the business has a genuine chance of being preferred, rather than trying to appear everywhere.

The real challenge of GEO

GEO is therefore not simply about multiplying opportunities to be cited. Visibility remains necessary. But it is only the beginning. The real challenge is to help generative engines understand the situations in which a business offers a particularly credible answer, and to give them enough evidence to decide in its favour.

“The real challenge of GEO is not just to be present in responses. It is to be chosen when the request genuinely matches what you do better than others.”