No single factor explains why AI recommends one hotel, tour operator, agency or destination over another.

A recommendation depends on a range of signals: what the company says, what other sources confirm, what customers report, how the offering is documented, the quality of the information available and, above all, its relevance to the traveller’s specific request. It would therefore be misleading to speak of a universal formula or a single recommendation algorithm. However, it is possible to organise the main factors to monitor.

Based on our audits and observation of generated responses, we group them into four levels: Eligibility → Relevance → Trust → Preference. This framework helps explain why a company can be technically visible without actually being recommended, a distinction we explore in being visible in AI does not mean being recommended.

1. Eligibility: can AI find and understand the company?

Before even discussing recommendations, we need to check that useful information is accessible. A company can have an excellent product yet remain difficult to interpret if its website is poorly structured, hard to access or incomplete. The initial factors are therefore relatively conventional.

These factors alone do not generate a recommendation. They establish a prerequisite: being accessible and understandable enough to enter the comparison. That is why a good technical score never guarantees strong GEO performance.

2. Relevance: does AI understand when the company is a good answer?

Being understood is not enough. The company must also be associated with the right situations, and this is where positioning becomes important. AI must be able to understand whether a property is better suited to a couple, a family, a solo traveller or business travellers. It must also be able to identify a price range, a location, a type of experience, a specialism or a travel occasion.

The key point is simple: the vaguer the positioning, the harder it is for AI to know when to suggest the company.

3. Depth: is there enough information to support this positioning?

Positioning can be perfectly articulated yet remain difficult to use. Saying that a hotel is “ideal for couples” ultimately provides little information. Why is it ideal? Because it is quiet? Because some rooms have a particular view? Because it offers experiences for two? Because its layout limits foot traffic? Because its guests confirm it? AI needs more precise information to associate a company with a situation.

The issue is therefore not simply having “a lot of content”. The content must make it possible to understand why the offering meets a particular request.

4. Trust: are the company’s claims confirmed?

At this level, GEO clearly extends beyond the website alone. A company can make many claims about itself, but a claim does not carry the same weight as an attribute confirmed by several sources. We therefore need to examine the available evidence.

“A commercial promise is a claim. A confirmed promise becomes evidence.”

This distinction is important. And when AI has to choose between several alternatives, that difference can matter.

5. Consistency: do the different sources tell the same story?

The number of sources is not enough either: they must not contradict one another. A property may present one positioning on its website, be described differently on a platform and be primarily associated with a third experience in customer reviews. This kind of discrepancy does not necessarily mean AI will reject the property, but it makes understanding it more complex. Several forms of consistency therefore need to be examined.

This is one of the points we examine most closely in our audits, because the issue is not just whether signals are present, but whether they can reinforce one another.

6. Context: does the company genuinely match the request?

Even an exceptionally well-documented company should not be recommended everywhere. A recommendation only makes sense if it addresses the specific context of the request. This is why some factors must never be analysed in isolation: the same company can be excellent in one scenario and of little relevance in another.

This is why we avoid speaking of an absolute “GEO ranking”. A company is not good or bad for AI: it is more or less relevant in a given context.

The 20 factors should not be treated as a checklist

Taken separately, these factors may give the impression that optimising them one by one is all that is needed. That would be a mistake, because they play different roles. Some simply enable a company to enter the comparison, others allow it to be recognised as relevant, others strengthen trust, and some can ultimately make the difference between two similar options. We therefore interpret them across four levels.

Eligibility

Is the information accessible, understandable and sufficiently structured?

Relevance

Does the company genuinely fit the requested situation?

Trust

Does it have sufficiently consistent evidence and sources to make that relevance credible?

Preference

Are there strong enough reasons to choose it over an alternative?

This perspective avoids a common trap: believing that a technical improvement will automatically improve recommendations. Technical foundations allow a company to enter the comparison. They are not enough to win it.

The BloomingPilot analytical framework

In practice, we therefore bring the analysis back to six key questions.

This framework then makes it possible to move from findings to an action plan. An accessibility problem is not fixed in the same way as a positioning problem, a lack of evidence is not addressed in the same way as inconsistent sources, and a poor recommendation in a given context does not necessarily mean overall visibility is poor. This is precisely why a GEO audit must go beyond a single score. We describe how we conduct this work in how to approach GEO: See, Align, Build, Amplify.

What executives should take away

Do not try to optimise twenty factors simultaneously. First, understand where the real barrier lies. Is the company poorly understood? Poorly positioned? Insufficiently documented? Inadequately corroborated by its customers and external sources? Or simply not very relevant to the scenarios analysed? Only after this assessment is it possible to prioritise actions.

And this is where the difference between AI search optimisation and GEO strategy becomes tangible. Being technically visible allows a company to enter the comparison. Relevance, evidence and consistency play a greater role in determining whether it emerges as the winner.