Generative AI vs machine learning: Why travel retail media leaders need to understand the difference

Artificial intelligence has become one of the most talked-about topics in marketing. In many conversations today, however, “AI” has effectively become shorthand for generative tools, systems capable of writing copy, creating imagery, summarising reports, or responding conversationally to prompts.

These capabilities are impressive and have brought AI into the everyday workflow of marketers, making it possible to generate creative assets or campaign ideas in seconds.

But generative AI only represents one part of the AI landscape. Behind many of the systems already shaping travel commerce and digital advertising sits a quieter form of artificial intelligence, machine learning.

Understanding the difference between the two is important, particularly for leaders responsible for travel retail media networks. While generative AI excels at producing language and creative outputs, machine learning remains the engine behind many of the predictive systems that drive commercial decisions.

In short, one type of AI generates content. The other predicts outcomes. Both matter, but they solve very different problems.

The AI everyone is talking about: Generative AI

Generative AI models, such as large language models (LLMs) like ChatGPT, Perplexity and Claude, are trained on enormous amounts of text, images, and other unstructured data. Their purpose is to generate plausible new outputs based on patterns learned during training.

Ask a generative AI model to draft a hotel description, generate a travel itinerary, or write ten variations of a destination advert and it will produce coherent responses almost instantly.

However, these systems operate probabilistically, they are designed to produce answers that are plausible rather than deterministic. That’s why asking the same question twice can produce slightly different answers each time.

It is also why generative AI can sometimes struggle with tasks that require absolute consistency, for example producing an exact character count or spelling a word repeatedly in the same way.

The well-known “strawberry test”, where AI models occasionally miscount the letters in the word, has become a humorous illustration of this behaviour. The point is not that generative AI is unreliable, but that it was designed to generate language rather than produce deterministic calculations.

The AI that has been quietly powering the internet: Machine learning

Machine learning models work differently. They learn patterns from structured data, using clearly defined inputs and outcomes to generate predictions. Given enough historical data, they learn how certain combinations of signals relate to specific results. The goal is not creativity, but consistency. If the same inputs are provided repeatedly, the model should produce the same prediction each time, updating only as new data improves the model over time.

Machine learning already underpins many systems across the travel industry. Airline revenue management systems use predictive models to forecast demand and adjust seat pricing. OTAs rely on recommendation engines to surface the most relevant hotels or destinations based on a traveller’s search behaviour. Airlines and booking platforms also use machine learning to detect payment fraud or unusual booking patterns.

In digital advertising and retail media, machine learning plays a similar role. These models determine which ads appear in front of which audiences, ranking millions of potential impressions based on the likelihood that a traveller will engage or convert. Signals such as search behaviour, travel dates, destination interest and past booking activity can all help inform those predictions.

While generative AI dominates headlines, machine learning has been optimising digital ecosystems for years.

For example, an airline that lacks robust demand forecasting might over promote a route where seat inventory is already constrained, depressing yield even as media spend increases. With a stronger predictive backbone, that same budget can be redirected toward routes with spare capacity, raising both load factors and media performance.

Why the distinction matters in travel retail media

Travel retail media operates in a highly data-rich environment. Every search, click and booking generates signals that can help indicate intent.

A traveller searching for flights from London to Barcelona two weeks before departure is sending a very different signal from someone browsing “best beaches in Thailand” six months ahead of a possible trip. Add additional context such as party size, seasonality, price sensitivity or past behaviour, and the picture becomes more complex still.

The challenge for a retail media network is to interpret these signals and determine which opportunities are most valuable. Which travellers are likely to book? Which destinations are about to see demand spikes? Which advertising placements will drive measurable revenue? These questions require prediction.

Machine learning models are designed to analyse structured signals such as travel dates, destinations, engagement history and booking behaviour to estimate probabilities. They can forecast demand on specific routes, identify high-value audience segments or determine which ads should be shown in which contexts.

In other words, machine learning helps determine what is most likely to happen.

Generative AI, meanwhile, helps shape how marketing teams respond to those insights.

If predictive models identify a high-value audience segment planning late summer trips to the Mediterranean, generative tools can assist in producing tailored ad copy, destination descriptions or campaign messaging for that cohort. They can summarise performance data for commercial teams or generate variations of creative assets for testing.

One technology predicts the opportunity, the other helps operationalise it.

Where the confusion comes from

Part of the confusion between generative AI and machine learning comes from the way both technologies are grouped under the umbrella of artificial intelligence.

In reality, they are designed to solve different types of problems. Machine learning is typically used for structured prediction tasks where consistency and measurable outcomes are essential. Generative AI is used where flexibility, interpretation and language generation are valuable.

Modern platforms increasingly combine the two approaches. A travel company might use machine learning to determine which users are most likely to book a particular route, while generative AI produces the creative assets used to reach them.

When viewed from the outside, this layered architecture can appear as though a single “AI system” is responsible for everything.

In practice, the underlying technologies are performing distinct roles.

Why travel retail media leaders need to understand both

Retail media networks are expanding rapidly across the travel sector, as airlines, OTAs and travel platforms recognise the commercial value of monetising their audiences.

As these ecosystems grow, artificial intelligence will increasingly sit at the centre of how advertising inventory is valued, how audiences are segmented and how campaigns are optimised.

For leaders overseeing these platforms, understanding how different forms of AI contribute to this process is critical.

Machine learning models form the predictive backbone, interpreting signals and identifying opportunities across vast datasets. Generative tools can then help translate those insights into creative output, operational efficiency and more accessible reporting.

If you’re overseeing a travel retail media network, your first question shouldn’t be “Where can we use generative AI for content?” but “Where do our predictive models most directly affect yield and ad revenue?”

Generative AI may be the most visible face of artificial intelligence today. But behind the scenes, machine learning continues to play the central role in predicting demand, identifying value and guiding investment decisions.

In the next article in this series, we will explore why predictive intelligence remains the backbone of travel retail media networks, and how machine learning models built on travel-specific signals can unlock more effective media optimisation.

‍