Your destination is being recommended in AI. Can you see it?

A traveller can now go from "where should I go for a family beach holiday" to a full itinerary in a single chat. Your destination is being recommended in those answers, or skipped, and most brands have no way to see which. This is the biggest change in destination marketing right now, and this is how to see it and measure it.

Has travel discovery moved to AI?

900 million people use ChatGPT every week. That is more than the combined populations of the United States, the European Union, Australia and New Zealand, turning to one tool every week. Gemini sits near the same number monthly, and Google's AI Mode has passed 1 billion monthly users. The new AI trip planners are climbing fast too. Layla has passed a million people planning and booking trips inside it.

The behaviour has changed, and the clearest sign is personal. We are now using it ourselves, or we know someone who is. That is where the penny drops for tourism, because the shift stops being a statistic and becomes the way we plan our own trips.

For a destination this is a hard change, whether you market a whole country, a beach region like the Gold Coast or Sunshine Coast, a single luxury resort or an island. The old marketing ran in a line: make social content and a website, drive traffic and engagement, let that demand convert, then report on social, website, visitation and spend. That whole chain is condensing into AI answers, where discovery, comparison and a draft itinerary can happen in one conversation before anyone lands on your site.

Why don't falling website sessions mean falling demand?

Here is the pattern we keep seeing. Sessions soften, and revenue and visitation hold or climb. The demand has switched channels. The traveller still chose your destination, still spent, still stayed, but the journey ran through an AI answer you never saw. Nothing dropped except your ability to measure it.

That is the trap. If you report only on sessions, you will read a channel change as a decline and defend the wrong budget. The job is to make the invisible path visible before you make any call about it.

What should a destination measure?

You cannot open a search console for ChatGPT, which is why people ask whether GEO can be measured at all. It can. Bing has a console, Google is rolling out prompt impressions in the US, and it has not reached Australia or New Zealand yet. Responses also change by person, context and memory. In a live session Dave had a full room prompt "top things to do" for one beach destination at once, and every phone returned a different answer. What stayed consistent was the set of recommendations cited most often underneath.

So the foundation is a prompt library you build yourself, with enough variations to represent that personalisation. Structure it two ways at once. By journey stage: discovery, category research, itinerary, and the conversion questions people type before they commit. And by your strategic jobs as a destination.

Those jobs carry more than raw bookings. Get chosen. Increase length of stay. Increase spend. Spread demand across seasons and across your area, so the peak sites do not carry all of it. Booking confidence and governance, so you stay a trusted source of accurate information. Cultural representation, so the communities and heritage you draw on are portrayed correctly and travellers are not all funnelled to the same handful of places. Accessibility and inclusion. Responsible travel, whether that is a carbon reduction commitment or whatever your own mandate is. Repeat visitation, and whether you hold any content that says you are more than a place someone visits once.

What you measure is what you manage. If you never track how AI represents your cultural story or your sustainability commitment, you cannot optimise for it, and you cannot tell whether the answers travellers see match the destination you are trying to build.

Building that library well is the hard part. Pull the data that should shape it, GA4, Search Console, SEMrush and the people-also-ask sets, and generate a prompt model grouped by journey stage and strategic job, tagged so each cluster can be analysed on its own.

What does AI already say about your destination?

Run a set of entity diagnostic prompts and use the model as a mirror. Ask what it says about your destination: how it positions you, who it says you are best for, and whether it portrays you the way your brand and positioning work intended. Given everything a model has read, this is a direct read on how it perceives your destination, your brand and your experiences.

It doubles as a feedback loop on reputation and narrative accuracy. If the answers misrepresent you, you see it here first, then work on the content and signals that shape it.

How do I connect AI visibility to bookings and spend?

A share of voice number means little on its own. It becomes useful only placed next to the rest of the picture, so this is worth slowing down on rather than skimming past.

Do it by category, the way you already think about your destination. Say beach and surf experiences are a strength. Track your organic impressions and sessions for beach and surf experiences beside their prompt group, its share of voice and citations. That gives you a small funnel report per category, and side by side over time it shows where the journey is switching. You might watch accommodation tanking under aggressive OTA pressure while another category climbs, which tells you where to move budget.

Brand search is the underrated companion metric. A simple Search Console query shows how many people search your destination by name, and someone searching your brand is aware of the category, aware of the product, and remembers your marketing. As you work the top of the funnel in AI, a rise in brand search is the correlation you watch, because you cannot yet measure the AI exposure directly.

Why does generic destination content stop working?

Across more than 10,000 prompts, the pattern holds like early SEO: whoever answers the question most directly gets cited. Generic destination copy is the weak point. Google's own AI optimisation guidance is blunt about it, calling generic pieces commodity content with little unique insight for readers. The models are good enough at synthesising common information that writing your own version of it adds little.

What gets cited is opinion and proof. People care what other people think, which is close to what it means to be human, and it is why Reddit and YouTube surface so often in these answers. Prompts reward it too. The average head keyword runs about 3 words, and the average Google AI Mode prompt runs about 23, because people give the whole scenario, down to a toddler and a six-year-old and the exact kind of trip they want. That depth rewards content built for the real scenario and carrying a point of view of your own.

What do you change once you can see it?

The analysis produces a plan across three areas: technical, content, and the sources beyond your own site. Start with technical, because if AI cannot read your site nothing else matters. Some large brands block the crawlers, GPTbot, Perplexity's bot, Google's AI crawler, or run so much JavaScript that the page never renders for them. Check you are indexable first.

Content is the bigger opportunity for most destinations, and it is the shift above: answer the real questions directly, give your writing a named author and a point of view instead of generic destination copy, and lean into the social and video where people already talk. Beyond your own site, check who cites you, from TripAdvisor to Reddit to Wikipedia, and make sure your destination is represented accurately there.

How many prompts, and where do I start this week?

You can build a library too big to use. Dave has run one past 900 prompts that became unwieldy and stopped producing action. Build the full library, tag each prompt to a cluster, then choose one area for a proof of concept, like destination selection, and measure whether you can improve how you turn up there before you scale.

This week, do the simplest version yourself. Open ChatGPT and run the prompts your visitors run: "best things to do in [your destination]", "is [your destination] worth visiting", "where should I stay in [your destination]". Read whether you appear at all, and how you are described. That is your baseline, and the reason to build the rest.

When AI plans someone's trip, does your destination make the list?

Most destinations have never checked, so they cannot see how AI describes them or which rival it recommends first. Closing that gap is what our GEO work does. We run it for destinations and experiences as a measurable engagement that tracks how you show up across the AI engines and ties it to the visitor economy. See how it works.

Rather talk it through first? Book a call. Want to keep up as the ground shifts? Get our newsletter.