A tourism marketing plan with AI: 5 steps and their prompts

A tourism marketing plan answers the same five questions today as twenty years ago: where are we, what do we want, who are we selling to, how, and on what budget. What has changed is the cost of answering them: what used to demand weeks of document writing now takes days with generative AI — if you know what to ask at each step and where not to let it decide. Here is the full walk-through, with a starter prompt per phase.

One rule that applies to all five steps: always give it your real data (numbers, customer comments, your specific situation) and never customers' personal data in public tools. Without the former, the AI invents; without the latter, you breach your own privacy policy.

Step 1 — Where are we? (analysis and diagnosis)

The job: look honestly at the market, the competition and your own digital state, and summarise it into weaknesses, threats, strengths and opportunities. Here AI is an excellent text analyst: give it raw material and ask for patterns. For the digital-state part there is one very concrete prior check: whether your website is readable by the machines that now read on the customer's behalf, something gone through point by point in the 7-point checklist.

Starter prompt: "Act as a tourism marketing consultant. I'm giving you: (1) a description of my business, (2) my last 50 reviews, (3) my 3 main competitors' websites. Extract: the 3 strengths my customers repeat most, the 3 recurring problems, what my competitors offer that I don't, and a draft SWOT as a table. Flag which statements are your deductions rather than data."

That last sentence is the most important in the prompt: it forces the AI to separate what it knows from what it assumes.

Step 2 — What do we want? (objectives you can measure)

A useful objective is one you can declare met or missed in December. "Improve visibility" is not one; "grow from 5,000 to 6,000 monthly organic visits in 6 months" or "25 direct bookings for the December long weekend on a 200-euro budget" are.

Starter prompt: "With this context [paste your step 1 diagnosis], propose 5 marketing objectives for the next 6 months. Each must include: current figure, target figure, deadline and how it will be measured. Reject yourself any that cannot be measured with the tools I told you I have."

Step 3 — For whom? (the customer with a name)

The classic mistake is confusing a segment ("families, 35-50") with a person ("Laura, 38, two kids, books the July trip in March, her number one fear is the children getting bored"). The humanised profile matters because it anticipates objections — and every anticipated objection is a page of your website that answers it before it is voiced. And it is worth building that profile on market data rather than intuition: in Spain, the segment that uses AI to plan the most is families with children, not Generation Z, and income discriminates more than age.

Starter prompt: "From these real reviews and enquiries from my customers [paste them], build 2 customer profiles: demographics, main trip motivation, the 5 doubts they will have before booking and which content on my website should answer each. Do not invent traits you cannot deduce from the material I give you."

Step 4 — How? (strategies and tactics)

Strategy is the direction ("become the reference stay for cycling tourists in the county"); tactics are the vehicles (content, partnerships, advertising). AI helps unfold each strategy into concrete actions — and spot which ones don't fit your resources.

Starter prompt: "My strategy is [write it in one sentence]. My resources: [weekly hours, monthly budget, who executes]. Unfold it into a 3-month action plan, week by week, prioritised. Mark which actions I can automate with AI and which require a person. If the plan doesn't fit my resources, cut it yourself and say what you sacrificed and why."

Step 5 — With what? (budget and measurement)

The budget derives from the tactics, never the other way round. And measurement is decided before starting: few indicators, one per objective, reviewed on a fixed calendar. Here the AI's role is sceptical controller.

Starter prompt: "Here are my plan and objectives [paste them]. Build my tracking table: for each objective, its indicator, where it is measured, how often it is reviewed and what threshold should make me change tactics. Ask me for whatever is missing instead of assuming it."

The red line

After this walk-through, the document will exist and read well. The danger is mistaking that for having a plan. The AI doesn't know whether your pool offsets the distance to the beach, whether your team can sustain two posts a week, or whether the bookings target is ambitious or laughable for your area. All of that comes from you, and its absence shows immediately: out comes a plan that would suit any business — that is, none.

A final workshop note: the quality of the answers depends directly on the structure of what you ask — role, task, context, format and constraints. It is the same principle behind the prompt improver I publish on this website, and the skill with the highest return per hour invested of anything in this article.