Google 搜尋變天！AI Mode 衝破 10 億用戶、Gemini 3.5 Flash 成預設，傳統搜尋框迎來 25 年最大改版

If you've recently searched for "Taipei weekend family attractions" on Google, you may have noticed that the results page looks different from what you're used to. In the past, you would get a list of blue links that you had to click on one by one to compare; now, a segment of AI-organized answers pops up at the top of the search results, conveniently listing several locations, opening hours, and suitable ages for you. You don't even need to click on any websites, and your problem is already half-solved.

This perceptual change is backed by a series of heavy-hitting announcements made by Google at the I/O 2026 developer conference on May 19. Two numbers are particularly worth remembering for Taiwanese users: Google's AI Mode has surpassed 1 billion monthly active users, and Google has upgraded its latest Gemini 3.5 Flash model to be the global default engine for AI Mode. In other words, over 1 billion people worldwide are now using a search entrance that directly answers questions with generative AI every day - and this entrance has just been equipped with a faster and smarter heart.

The way we search is being redefined before our eyes.

Event Background

The Google search box has been the iconic entrance to the internet for over two decades. Its core logic hasn't changed much over the years: you enter keywords, it returns a list of webpage links, and you're left to make your own judgments. This model has supported the entire content industry, the SEO industry, and even Google's own advertising empire.

However, generative AI has disrupted this logic. When AI can directly understand your questions, aggregate information from the entire web, and answer you in a single segment, the old "give you a list of links" approach seems clumsy. Over the past year, Google has been testing the waters with its AI Mode feature, gradually introducing a conversational, AI-organized search experience to users. Now, with the official release of the 1 billion monthly active user figure and the announcement that query volume has doubled every quarter since its launch, it's clear that this experiment has crossed the threshold of user acceptance.

Upgrading to Gemini 3.5 Flash is the next step in this experience. The Flash series emphasizes being "fast and affordable," which perfectly suits the high-frequency, low-latency, and high-volume request scenario of search. By making Flash the default model, Google is sending a clear message: AI search is no longer a feature for the minority, but a basic infrastructure for global scale.

Key Points

• AI Mode monthly active users surpass 1 billion: Google officials state that query volume has doubled every quarter since its launch, marking a key milestone in the transformation of search.
• Gemini 3.5 Flash becomes the global default: AI Mode is equipped with the latest Flash model, which boasts stronger encoding, agency, and multimodal performance while maintaining the Flash series' speed and cost advantages.
• The largest search box overhaul in 25 years: The new "smart search box" will dynamically expand as you input, providing AI suggestions that go beyond traditional auto-completion and supporting multimodal input such as text, images, files, and videos.
• Personal Intelligence expands significantly: Personalized intelligence features are now available in nearly 200 countries and regions, supporting 98 languages, and can be used without a subscription.
• Agent functionality preview: Features like information agents (automatic monitoring of webpage updates), agent bookings, and agent calls will be launched in the summer of 2026 in the US and other regions, with some functions initially limited to AI Pro and Ultra subscribers.

Market Impact Analysis

Taiwanese users: The good news is that Personal Intelligence has expanded to nearly 200 countries and 98 languages, and can be used without a subscription, making it more accessible to Traditional Chinese users. However, a deeper change lies in user habits - when search changes from "giving you links" to "answering for you," we need to relearn how to ask questions and judge the credibility of AI-provided answers. My suggestion is to use Google AI Mode as a quick aggregation station, but for important decisions (medical, legal, financial), cross-check with original sources and use tools like Perplexity or directly open Gemini and ChatGPT for more in-depth inquiries.

Enterprise applications: For Taiwanese brands, e-commerce, and media, this is a wake-up call. When answers are directly displayed on the search results page, users' motivation to click on websites is weakened, and traditional models that rely on search traffic for sales and advertising revenue will face pressure. Enterprises should not panic but adjust their strategies: content needs to be authoritative and structured to be cited by AI as a source; meanwhile, they should diversify their channels to reduce dependence on a single search traffic source. This is also a good time to review AI introduction tasks and incorporate "being seen by AI" into content strategies.

Developers: Gemini 3.5 Flash emphasizes stronger encoding and agency capabilities, and with the simultaneous opening of the Gemini API, it's a direct benefit for developers. The Flash series' "fast and affordable" positioning is particularly suitable for applications that require large-scale calls and are cost-sensitive, such as customer service robots, instant summaries, and batch content processing. For engineers who are used to integrating multiple models with Cursor, incorporating Flash into their toolbox and doing model diversion (using large models for high-difficulty reasoning and Flash for high-frequency simple tasks) will be a practical cost optimization method. Those who want to practice can start by testing Flash's response quality with a few prompt words.

Future Development Trends

The real signal from this I/O is that search is moving from "information retrieval" to "task agency." Google's preview of agent bookings, agent calls, and information agents all point to the same thing: the end goal of search is no longer an answer, but a completed task. You're not just asking "which restaurant is good nearby," but directly letting it book a table for you.

This will lead to two chain reactions. First, "zero-click search" will accelerate, and more queries will be resolved on the search page, reshaping website traffic structures and forcing the content industry to find new ways to be discovered and monetized. Second, model layering will become the norm - using Flash for high-frequency daily tasks and flagship models for complex reasoning. This "choose a model based on the task" thinking will spread from developers to general tool users.

Of course, the risks are real. AI directly providing answers means that misinformation can spread faster and be harder to trace; when 1 billion people rely on the same AI entrance, its biases and blind spots will also be amplified. This is the cost of convenience, and it's a topic we'll need to keep an eye on in the coming years.

TheAI Academy Summary and Evaluation

The 1 billion monthly active user figure says more about the problem than any model's running points: AI search is no longer "something that will happen in the future," but "something that is happening and can't be turned back." Google's switch to Gemini 3.5 Flash as the default model is equivalent to officially moving the search entrance, the largest traffic entrance, to the foundation of generative AI.

Evaluation: The search box has changed its heart, and what's changed is not just the technology, but the way we find answers and make decisions; learning to ask questions with AI will be more valuable than learning any single tool.

Practical suggestions for Taiwanese readers: on the user side, practice "asking more specific questions and checking more cross-references," and don't take AI's first answer as the final verdict; for content workers and enterprises, take stock of your content's structure and authority, and start diversifying your traffic sources. Don't put all your business on a single search traffic source. The dividends of AI search will go to those who know how to use it well and maintain their judgment.

Data Sources

• Google Official Blog (Search at I/O 2026): https://blog.google/products-and-platforms/products/search/search-io-2026/
• Google I/O 2026 All Announcements: https://blog.google/innovation-and-ai/technology/ai/google-io-2026-all-our-announcements/
• 9to5Google Report: https://9to5google.com/2026/05/19/google-io-2026-news/

This article is compiled based on publicly available information, with official announcements as the standard. The launch time, region, and subscription conditions for each feature may differ from the release date, and the actual availability range for Traditional Chinese and Taiwan regions may vary.