How much should you spend on AI marketing in 2026?

What does the data say marketing budgets look like in 2026?

They are flat, and the money inside them is being reshuffled rather than grown.

Gartner’s 2026 CMO Spend Survey puts marketing budgets at about 7.7% of company revenue — essentially unchanged for a third straight year. Paid media now takes 31.4% of the average budget, up again and funded largely by cuts to agencies. Digital is more than two-thirds of all media spend.

The AI line is the one that grew. CMOs report allocating 15.3% of the marketing budget to AI — a real commitment, sitting next to a quieter admission: by Gartner’s own count, 70% of CMOs want to lead on AI, yet only about 30% feel ready to scale it. The budget is there; the plan for it often is not.

That is the gap this guide is about. Not whether to fund AI marketing — the survey says most teams already do — but how much, and against what, so the 15.3% does something.

How much of the budget should go to AI marketing?

Start from the share, not a round number. AI marketing is a slice of a marketing budget that is itself a slice of revenue.

The chain is simple. Marketing tends to run near 7.7% of revenue. Of that, the benchmark figure for AI is about 15%. A business doing Rp2 billion a year in revenue, then, is in the rough vicinity of Rp150 million a year — a little over Rp12 million a month — before any decision about where it goes. The point of the benchmark is not precision. It is a sanity check: a budget far below it is probably under-investing relative to peers, and one far above it usually has a readiness problem, not a spending problem.

For most Indonesian SMEs and growth-stage brands, that lands somewhere between Rp10 and Rp100 million a month. The rest of this guide treats those as the working tiers, because the more useful question is rarely the total. It is the order.

One caveat for Indonesian teams in particular: the readiness gap tends to be wider here than the global average, not narrower. The appetite is rarely the problem — PwC found 69% of Indonesian workers have used AI for their role in the past year — but the workflow to spend an AI budget well usually lags the willingness to fund one. Which is the argument for starting small and in order, rather than committing the full benchmark before there is anything to point it at.

What should the budget fund first?

Four things, in sequence. The order matters more than the amount, and it does not change as the budget grows.

1. Monitoring. Find out how often AI names you when buyers ask about your category. This comes first because every other line depends on knowing the baseline. It is also the cheapest move — at the low end, it is your own time.

2. A clear source-of-truth page. One canonical, owned page that states your facts plainly and answers the question in its first lines — the clean source an engine can retrieve and quote.

3. Content that answers real buyer questions. Pages built around what buyers actually ask, one question to a page, written to be useful rather than promotional.

4. Independent coverage. The third-party mentions an engine trusts more than your own site. This is the slowest line to move and the one most worth funding once the first three are solid.

The order is not arbitrary; it follows how an answer gets built. An engine has to retrieve a source before it can quote one, and it leans on independent sources over a brand’s own pages — Semrush’s look at more than 150,000 AI citations found forums, reference sites and the press cited far more often than marketing pages. So monitoring tells you where you stand, the source page gives the engine something clean to read, content widens what you can be retrieved for, and coverage is what tends to get you quoted. Skip a step and the later spend has less to work with.

Where do you put Rp10, 25, 50 and 100 million a month?

Treat each row as a monthly budget. Monitoring is the constant; every tier above the first simply adds the next layer in the sequence.

Where an AI-marketing budget goes, by monthly tier (≈ Rp16,300 / US$1, June 2026)

The common mistake is buying the bottom of the table before the top is in place. A hundred million rupiah of content and ads, pointed at a brand with no clear source page and nothing independent confirming it, still leaves you absent from the answer. Fund the foundation, then scale what the monitoring tells you is working.

What if the AI-marketing budget is zero?

The first tier is mostly time, so the foundation is doable at no cash cost. Budget buys speed and scale, not permission.

• Monitor by hand, monthly. Ask ChatGPT, Perplexity and Google the five questions your buyers really ask about your category — phrased without your brand name — and note whether you appear, and who does.
• Make one page answer-first. Take your most important page and lead with the clear, quotable answer in the first sentence.
• Keep your facts consistent. Same name, same numbers, same core claims everywhere you appear. A fact a model has to guess at is one it tends not to repeat.
• Claim every listing. Google Business, the marketplaces, the local listicles in your category — these are exactly the pages AI cites.
• Earn one honest mention. One credible third-party article, or a genuine contribution to a relevant community thread, can outweigh a month of your own posts.

Where do AI-marketing budgets get wasted?

Mostly on old reflexes: spending on the last layer while skipping the first.

The classic leak is paying to amplify a brand that is still invisible in the answer — running ads, or churning thin, near-identical posts, before there is a clear source page or any independent coverage to retrieve. With paid media already at 31.4% of the average budget, the pull toward more of it is strong; the discipline is to make sure the foundation is funded first. A budget spent strictly in order — monitor, source-of-truth, content, coverage — rarely runs into this. (If the question is the marginal rupiah rather than the standing line, that is a separate decision about where the surplus compounds.)

Frequently asked questions