On July 21, 2026, Google announced three new Gemini models: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. The one that matters most for search visibility is Gemini 3.5 Flash-Lite, which Google confirmed is rolling out into Google Search, including agentic search and potentially AI Overviews and AI Mode. Google describes 3.5 Flash-Lite as its fastest and most cost-effective 3.5-class model, running at roughly 350 output tokens per second and priced at $0.30 per 1M input tokens and $2.50 per 1M output tokens. Google also said AI Mode now has more than 1 billion monthly users.
Read together, those numbers point in one direction. A model this fast and this cheap makes it practical for Google to generate AI answers for far more queries, for far more people, at far lower cost. So the real story is not the model itself. It is what a model like this lets Google do to the results page.
What Google actually announced
Three models, one search-relevant rollout. Full details are in Google's own post on The Keyword (Google's blog). The headline facts for marketers:
- Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber were announced together.
- Gemini 3.5 Flash-Lite is heading into Google Search, covering agentic search and possibly AI Overviews and AI Mode.
- Google positions 3.5 Flash-Lite as its fastest, cheapest 3.5-class model, around 350 output tokens per second.
- Pricing sits at $0.30 per 1M input tokens and $2.50 per 1M output tokens.
- AI Mode passed 1 billion monthly users.
Why speed and price change the results page
When the model behind an AI answer is expensive or slow, Google has a reason to hold back and show a plain list of links. When it is cheap and quick, that reason weakens. A faster, cheaper engine lets Google synthesize an answer on more searches, refresh it more often, and run the multi-step agentic queries that read several pages before replying. The output most users see shifts from ten blue links toward a written answer that cites a handful of sources.
Ranking gets you into the pool; citation gets you into the answer
Here is the shift worth internalizing. Ranking on page one still matters, because AI systems tend to pull from pages that already rank well. But ranking is now the entry ticket, not the finish line. The pages that get quoted, linked, and named inside an AI answer are the ones that earn attention, because the answer sits above the links and often satisfies the searcher before they scroll.
That is the core idea behind answer engine optimization. You are writing for two readers at once: the person, and the model that summarizes for the person. Practically, that means content structured so a machine can lift a clean, correct passage out of it. Clear question-shaped headings. A direct answer in the first sentence under each heading. Specific facts, figures, and named entities instead of vague phrasing. Our approach to that work sits under AI SEO and the wider practice of generative engine optimization.
What this means for Thai marketers
Google did not publish Thailand-specific figures with this announcement, so treat the following as reasoning rather than measured local data. Thai brands optimizing for Google AI Mode should expect AI answers to appear on a wider range of queries, including the longer, more conversational Thai-language searches people type when they are still deciding. If your category is one where buyers ask questions before they buy, more of those questions may get answered on the results page.
The response is not to panic about lost clicks. It is to make sure that when Google composes an answer in your space, your brand is inside it. Publish content that states facts plainly, keeps product and pricing details current, and answers the real questions customers ask. Build the kind of authority signals that make a model comfortable quoting you. And check how you already surface across AI assistants, since Google is not the only answer engine your buyers use. That visibility work runs alongside optimizing for ChatGPT and other AI assistants.
FAQ
Does this replace normal SEO?
No. AI systems still lean on pages that rank well, so technical health, relevance, and authority remain the foundation. AEO adds a layer on top: structuring content so it can be quoted accurately inside an AI answer.
Is this only about AI Overviews?
Google tied 3.5 Flash-Lite to Search broadly, including agentic search, with AI Overviews and AI Mode named as possibilities. The practical takeaway holds regardless: cheaper, faster models mean more AI-generated answers across more of the results page.
What should I do first?
Look at the queries that bring you traffic and ask which already trigger an AI answer. Rewrite those pages to give a clean, correct, quotable response near the top, then widen out from there.
If you want a plain read on where your brand stands in AI results and which pages to fix first, Relevant Audience can help you map it out. Start with a look at your current AI search visibility, then decide where the work pays back fastest.
