On 5 August 2026, Tourism Australia released Tourism 2035, its strategy for the next decade. It sets a target of $61 billion to $69 billion in overnight spending by high-yield travellers by 2035, and four priorities for the next three years. The first is called “Marketing to humans and machines”. In Tourism Australia’s words, the aim is to “ensure our marketing engages with machines, algorithms and AI systems to recommend Australia” (Tourism Australia, August 2026⁠ (opens in a new tab)).

Most coverage since has treated it as Tourism Australia’s job. Our argument is that most of the work lands somewhere else: on the regional and local destination websites, and the Operator data behind them. We should be upfront that this is also Roam’s business, since we build those websites. Judge the argument on the reasoning below, and on the share-of-model method at the end, which works whoever runs your site.

What the strategy says, and what was reported

The release names four three-year priorities: marketing to humans and machines; showing the breadth of Australian experiences, beyond beaches and wildlife; using the “major events runway” ahead of Brisbane 2032; and a distinctive luxury positioning built on nature and rare experiences. It also names the pressure behind the first priority: “content overwhelm and digital disruption reduces the time and focus available for destinations”, and planning and booking now follow “multiple paths to purchase”.

KarryOn’s report of the strategy adds the detail. Tourism Australia plans to optimise its content through “better tagging, machine-readable information and connected APIs”, to track Australia’s “share of model” (how often Australia appears in AI travel recommendations), and to give Operators “practical tools and training” for AI-ready content. KarryOn also reports the warning that businesses with “incomplete, outdated or poorly structured digital information risk being excluded before a traveller even begins comparing products” (KarryOn, August 2026⁠ (opens in a new tab)).

Why the work lands on regional sites

An assistant rarely recommends a country. It recommends a town for a long weekend, a walk with an early start, a winery that does lunch on Mondays, a festival on a particular date. Tourism Australia can make Australia legible as a choice. It can’t, from australia.com, make legible the hundreds of places and thousands of Operators that turn the choice into a trip.

Those facts already have a home. The Australian Tourism Data Warehouse (ATDW) exists to distribute Operator profiles to hundreds of sites (Tourism Australia, ATDW guide⁠ (opens in a new tab)), and the regional and local destination websites are where those profiles meet the editorial that explains a place: which road to take in winter, which town to base yourself in, what changes after Easter. When an assistant tries to answer “where should we go for four days in the Grampians in May”, the best material on the web is usually on the regional site, if the regional site lets it be read.

That makes “marketing to machines” a supply chain: Tourism Australia sets the demand, and state bodies, regional tourism organisations, councils and Operators hold the supply of facts.

What the priority asks of a regional website

Structured Operator data, kept current. The strategy’s warning about incomplete or outdated information is aimed at Operators, and a regional site inherits whatever their profiles say. Feed Products from ATDW, chase the stale ones, and fill the fields assistants can use: accessibility, accreditations, opening days, booking links. Our ATDW in 2026 roundup lists this year’s new fields.

Pages that answer planning questions in sentences. The strongest pages for a machine reader are the ones a good visitor information centre volunteer would give you: where to base yourself, how long things take, what is closed when. Write them as plain statements with dates and distances.

Machine-readable by default. Readable HTML, structured data, an llms.txt file and crawler permissions that let assistants in. They are plumbing: they let a model read what you already publish (see How AI search is changing destination discovery).

Breadth, findable. The second priority, showing travellers “multiple experiences”, is a regional site’s whole job. A region with ten towns and one famous one should make the other nine as easy to read as the first.

Events with dates. The third priority leans on the events calendar through to 2032, and the release talks about a national platform for destination marketing around the Olympics “to spread the benefits across all states and regions”. An events page with dates, venues and booking links is how a region joins that, which is why we wrote Event pages that send Visitors on to Operators.

Measuring your own share of model

“Share of model” is a good idea with a hard measurement problem. Assistants answer differently each time, by user and by wording. If you want a regional version, keep it modest and consistent.

  1. Write 20 questions a Visitor might ask about your region, in their words: “best winery lunch near X”, “where to stay for a walking weekend in Y”.
  2. Each quarter, put the same questions to the same three or four assistants, logged out, and record which towns, Operators and sources appear.
  3. Track the trend and the sources cited, not a single score. When the cited source is a stale third-party blog, you know which page to improve.

Treat it as a listening exercise. Nobody can promise you a place in an AI answer, and a vendor who offers to is selling a guess.

What is still unknown

A strategy is a statement of intent, and the operational detail of this one, the tools, the tagging standards, the APIs, has been reported but not yet published in full. Some of it will change. What a regional site can do now holds either way: make its facts current, specific and readable, so it is ready for whatever form Tourism Australia’s tools take.

On Roam

Roam syncs 40,000+ ATDW Products automatically, and on every Roam site AI Discovery is live at no extra cost: an llms.txt file the platform generates and keeps current, and clean, structured markdown for any AI agent that visits. See Roam’s approach to AI.