Google has published the rules for AI search. Most Chiang Mai businesses have not read them.

The short version, for owners who do not work in marketing

• Google now answers the question itself, in its own words, above the list of links. Most searches end there without a click.
• Search traffic to small sites has collapsed. Chartbeat data puts the drop at roughly 60% for small publishers over two years.
• In May 2026, Google published its own written instructions on what makes a business visible inside AI answers. The document is free, short, and written for non-specialists.
• There is no trick, no special code, and no agency shortcut. Google rewards material a competitor cannot copy: your own numbers, your own cases, your own experience.
• An AI answer names two or three businesses. Coming eighth in your category produces nothing, where it once produced enquiries.
• AI referrals remain small in volume. This is a shift in how customers decide, not yet a replacement traffic source.

Three things to do. Read Google’s own guide to generative AI features before you pay anyone to interpret it. Publish content only your business is able to produce. Measure how often your name appears in AI answers instead of counting visitors.

A search for “best boutique hotel Chiang Mai” now returns a paragraph before it returns a hotel. The paragraph names two or three properties, explains in a sentence each why they are worth booking, and places small citation links beside the text it drew from. Below the paragraph sits the older version of search, the ten ranked links that built an entire local marketing industry, and most people never scroll that far.

Two things happened in Google Search over the past three months, and the second matters more than the first. Google redesigned how its AI layers present the web, which was expected. Google also published official documentation explaining how to be visible inside those layers, which was not. Simon Moser set out the commercial consequences in Entrepreneur on 6 August. The documentation removes the last excuse available to any business owner, which is not knowing what the rules are.

In early May, Google updated AI Mode and AI Overviews, the AI-generated layers that sit above conventional results. The updates added suggested follow-up angles at the end of an answer, site previews when a user hovers over a link, prominence for news publications a user already subscribes to, and citations positioned next to the specific claims they support.

These read as interface adjustments, and they are not, because together they confirm that the synthesised answer rather than the ranked list is now the thing Google delivers. The user reads a response and then decides whether any onward click is necessary at all.

The traffic consequences are recorded. Chartbeat data covering thousands of sites shows that search referral traffic fell by roughly 60% for small publishers and 47% for mid-sized ones across two years, with large publishers losing 22%. The gap between those figures is the story. Businesses with a recognised name and a direct audience absorbed the change, and businesses that depended on being found absorbed the full force of it. Our quarterly digital audits find the same pattern across Chiang Mai, most visibly for operators whose enquiries once arrived through blog content and category pages.

In mid-May, Google released its first official guide to optimising for generative AI features. The central position is that optimising for AI search remains SEO, performed properly, rather than a separate discipline requiring separate tooling.

The guide is most useful for what it rules out. There is no special schema markup that unlocks AI visibility, and Google names llms.txt files, content chunking, and AI-specific rewriting as things site owners can ignore. Manufacturing brand mentions across unrelated websites is far weaker in effect than the agencies selling it suggest. What the documentation rewards consistently is content an AI system cannot assemble from anywhere else: original research, first-hand experience, subtopic depth, material that exists because a specific business did the work.

On 5 June, Google extended its long-standing advice on hiring SEO help with a new document on assessing third-party tools, services, and advice, naming generative engine optimisation and answer engine optimisation as service categories for the first time. Google states plainly that it does not evaluate or endorse third-party services, and it supplies vetting questions for owners to use directly. Does the provider cite official Google documentation. Is the AI optimisation advice consistent with what Google has published. Anyone in this city who has sat through a pitch promising guaranteed placement in AI answers now has a free instrument for testing it.

Two claims in the current coverage deserve more scepticism than they receive, and both cut against the urgency being sold around this subject.

The first is the idea that AI referrals are quietly replacing lost search traffic. They are not, at least not yet at any meaningful scale. Chatbot referrals grew sharply through 2025 and still account for less than 1% of publisher page views. The visitors who arrive from an AI recommendation do convert at a higher rate, because the comparison stage happened before they landed, and a higher conversion rate on a very small number is still a very small number.

The second is Google’s own position, which is that AI Overviews increase search usage overall and produce higher-quality clicks. Google has published data supporting this and has an obvious commercial interest in the conclusion. Both things are consistent with the publisher figures, because total search activity grows while the share of that activity reaching any individual small site falls.

The honest reading sits between the two. What has changed decisively is where the buying decision gets made. What has not changed is that traffic volume is no longer the number worth defending.

Dozens of Chiang Mai businesses compete for the same handful of commercial queries in every sector. Spa treatments, international schools, dental work, coworking, wedding venues, long-stay accommodation. In the ranked-list era, placing eighth in that field still produced enquiries, because visitors browsed the field and built their own shortlists.

That behaviour is ending, and the market structure changes with it. When an assistant returns three recommendations in conversational form, most users accept the three. There is no page two to venture onto. The recommendation is the shortlist, and the shortlist is the market.

The distance between being cited and not being cited is therefore not a matter of degree. A business that appears in AI answers accumulates advantage with every query in its category, and a business that does not appear is invisible to a growing share of the people actively looking to buy from it. Google’s new features create a few additional slots inside the answer through suggested angles and community perspectives, and every competitor in the category is contesting those same few slots.

Chiang Mai carries a specific disadvantage here. A large share of local business content is commodity content, produced cheaply, describing what any competent operator in the sector also offers. AI systems synthesise that material from a thousand interchangeable sources and cite none of them. The operators with genuine first-hand material, the vineyard that logs its own harvest data, the clinic that publishes treatment outcomes, the hotel that documents what guests actually ask for, hold an advantage they have not yet priced.

Run this test before you read any further

This takes 20 minutes and costs nothing.

1. Write down the five questions a customer asks immediately before buying from you. Use their words, not your marketing words.
2. Put each one into Google AI Mode, ChatGPT, and Gemini. Add “in Chiang Mai” where it applies, and run the Thai-language version as well if you serve Thai customers.
3. Record every business named in each answer. That list is your real competitive set, and it will differ from the one in your head.
4. Check what is cited beside the answer. Where the citation goes to a booking platform or a listings site rather than to any business website, the businesses in that category have handed the customer relationship to a third party.
5. Repeat the same five queries in 90 days. The direction of travel matters more than any single result.

Read Google’s optimisation guide before commissioning anyone. It is short and written for non-specialists, and it now functions as the benchmark against which every AI visibility proposal is measured. A vendor whose recommendations contradict it is selling something else.

Move content strategy from volume to depth. One piece of original research on your own sector outperforms 20 posts assembled from the same sources your competitors used. This is the single change with the clearest return, and it is also the change most local businesses resist, because it requires the business to have something to say.

Change what the dashboard measures. Raw traffic describes a shrinking game. Track how often the business is named in AI answers for the queries that produce revenue, and track the conversion rate of visitors arriving from AI referrals. Fewer clicks are doing considerably more work, and a dashboard built for 2023 hides that entirely.

Budgets are already moving. PMG, one of the largest independent marketing agencies in the world, advises clients to pilot generative engine optimisation at 1.5 to two times their existing search spend, according to Matt Allfrey, its head of SEO for EMEA. A Semrush study published in June 2026, covering 481 marketing professionals, found more respondents planning to invest in AI search optimisation than in conventional SEO, at 38% against 36%. In the same survey, 85% said AI has already changed how they approach search.

The businesses that will hold the recommendations in Chiang Mai 18 months from now are making one decision this quarter, which is to publish something a competitor cannot copy and to stop measuring the wrong number. Everyone else is reading traffic reports and waiting for figures that are not coming back.

---

If you run the 20-minute test above and your business does not appear, the next question is why, and that usually comes down to whether an AI system can reach your pages at all. CMBN Media Services runs a digital audit that checks exactly that: crawlability, structure, and whether your information is readable and attributable to you. Most Chiang Mai sites we audit fail before content is even part of the conversation. Email gee@chiangmaibusiness.net with your URL and we will run it and send you the findings.