Google AI Mode Ads: 10 Markets, Indonesia Waits

Google AI Mode ads arrived unevenly

Google published a global expansion and a help page with the operational detail. Read together, they show two different products across the two markets.

The 10-country launch includes Singapore, Malaysia and the Philippines in Southeast Asia. Indonesia is absent from that list. Google Ads Help is more precise: Indonesia can serve ads inside AI Overviews, while Singapore can serve them inside AI Overviews and AI Mode.

Google launched AI Mode itself in Indonesia in Bahasa Indonesia on 9 September 2025. Availability of the search product does not mean availability of its newest ad placement. For an Indonesian SME, that is the catch hidden behind a regional launch headline.

The money stays in the same auction

Google Ads Help says there is no extra charge for entering AI Overviews or AI Mode. The same auction, cost-per-click or conversion-based bidding, and invalid-traffic protection apply. Existing eligible campaigns join automatically when the placement is relevant.

There is no separate AI Mode budget. An advertiser also cannot target only AI Mode or opt out of it separately from Search. Negative keywords, brand exclusions and asset controls still apply, but the new surface shares the existing campaign purse.

Our read: the launch changes where a search budget can be spent, not the basic invoice. That makes the first risk easy to miss. A business may see cost, clicks and conversions move without a clean line showing how much came from an AI answer.

“Ads will be clearly labeled as “Sponsored.””

The headline claim has no denominator

Google says click-through rates were similar for sponsored and organic product listings inside AI Mode. The evidence is its own US data from 20 August to 3 September 2026. It did not publish either rate, the number of searches, the advertiser mix or a Singapore result.

The longer-query claim has the same limit. The three-times figure comes from a May 2026 US report. Sundar Pichai added on 28 October that one in five AI Mode searches globally uses voice, images or documents, and that Google handles more than 50 billion visual searches a month. Those figures support a change in behaviour, but they do not prove a better return for a local SME.

Google also reports AI-surface impressions, clicks, conversions and cost inside standard campaign reports. Search terms are aggregated. That means an owner can watch the total account, but cannot isolate a clean AI Mode profit-and-loss line from the published controls.

Singapore gets the test; Indonesia gets a preview

Consider a Singapore renovation firm. A searcher might ask which contractor can renovate an occupied office after hours, handle permits and finish before a lease deadline. Google can match an ad to the query itself or to the context of the AI answer. A landing page that only says “quality renovation” gives the system and the buyer little useful evidence.

Now consider a Jakarta furniture seller. AI Mode works in Bahasa Indonesia, including follow-up questions and image input. Yet Google currently lists in-answer ad inventory there as English and AI Overviews only. The owner should not rebuild the campaign around a placement that has no published Indonesian-language launch date.

The practical difference is timing. Singapore advertisers should inspect results now. Indonesian advertisers should make product pages answer detailed buyer questions, keep normal Search measurement clean and wait for a dated local rollout before claiming an AI Mode strategy.

A sponsored answer does not finish the customer job

Google can introduce an ad group with an AI-generated explainer based on the query, the AI response and public website information. Advertisers cannot write, edit, approve or separately opt out of that explainer. Google says it remains distinct from the ad copy.

That raises the value of plain, accurate pages. Prices, service areas, delivery limits, return terms and booking rules should agree across the ad, landing page and customer conversation. If the website says one thing and the WhatsApp team says another, the extra context creates a faster route to disappointment.

The work after the click belongs to the business. A customer operations AI agent can collect the detailed request, check approved records and hand an exception to a person. The closest proof is Rosalie Cheese: WhatsApp Ordering, where staff approve each order before fulfilment continues. Buying the click cannot replace that operating handoff.

What should an SME do now?

• Write down the baseline. Record spend, qualified enquiries, completed orders and staff handling time for the four weeks before the rollout. Clicks alone do not show whether the new placement helped.
• Audit the questions behind the keywords. Add clear answers for constraints buyers actually mention, such as service area, timing, stock, price basis and approval steps. Keep every claim consistent with the customer team.
• Protect the auction. Review negative keywords, brand exclusions, landing pages and conversion definitions. Existing campaigns may enter the new inventory automatically.
• Read search terms every week. Look for longer questions that brought useful customers or waste. Google aggregates AI-surface reporting, so log the enquiries your team actually receives after the click.
• Fix the handoff before adding budget. Name who owns the lead, which record they check and where a price, booking or promise needs approval. Compare that flow with manual WhatsApp and spreadsheet work before hiring another admin.

What to watch next

• By 31 December 2026: Whether Google adds Indonesia to AI Mode ad inventory or adds Indonesian to the published language list.
• In the next Google Ads Help update: Whether advertisers receive placement-level reporting, a separate control or a way to test AI Mode inventory directly.
• At Alphabet Q4 2026 results: Whether Google publishes AI-search revenue, advertiser results or market-level evidence beyond internal US studies.
• After four operating weeks: Whether the SME sees more completed customer jobs per dollar and less handling time, not just longer queries or more clicks.