Google search for "AI mode" is here, is Perplexity panic?

This is Google's direct response to recent controversy. On the one hand, Google search has been criticized. Apple executives claim that Google search has been challenged by competitors such as Perplexity and ChatGPT, and the number of searches on the Safari browser (the default engine is Google Search) has declined for the first time since 2022.

On the other hand, AI startups have begun to eye browsers, and Chrome is also being re-examined. Perplexity is about to release browsers, and all the names have been given. OpenAI also has similar rumors and bluntly stated that it is interested in acquiring Chrome.

Google's further upgrade of search and browser is showing its determination to "continue to greatness."

Of course, the highlights of this keynote speech at Google I/O conference are not only that, but also the latest progress of the Android XR platform, the upgrade of 3D video chat projects, the upgrade of Wensheng image tools Imagen and Wensheng video tools Veo, the launch of AI film production applications, Project Astra is more proactive, etc.

Google has almost done everything with AI.

01

Google searches for "AI Mode" and Chrome's Gemini

At last year's I/O conference, Google launched the "AI Summary" feature in search.

In short, when searching, an AI-generated summary will be formed at the top to help users summarize (Baidu has similar functions). After the launch of the AI ​​summary, some troubles also arose, such as the old hallucination problem.

Now, Google has taken a step further and announced that it will add "AI mode" to search.

The entrance of "AI Mode" will be paralleled with "All", "Image", "Consultation", etc., and will be at the first position, displayed in the lower part of the search box.

In AI mode, the user gives a search request in natural language, and the model directly generates a summary answer, displays it in graphic form, and gives the relevant web page to the right.

The AI ​​model uses Google's cutting-edge model and leverages what the company calls "query fan-out" technology. Google says the method breaks down user queries into smaller subtopics, running multiple separate searches simultaneously. This allows the AI ​​model to perform more in-depth searches than traditional Google searches, Google explained.

It is not difficult to see that the user experience is no different from the search mode of any mainstream AI application. But this is a big step for Google, both a response to challengers such as Perplexity and ChatGPT, and an internal subversion of its long-standing search logic for keyword-web page results.

Another major change is that the Gemini AI assistant will be connected to the Chrome browser.

In the Chrome browser, you will see a small sparkling icon in the upper-right corner. Click on it and the Gemini chatbot window will open – it is a floating UI that you can move and resize. There, you can ask questions about the website.

Initially, users could use Gemini to communicate while browsing various tabs. But "later this year", Gemini in Chrome will allow users to select multiple tabs at once and ask questions for all tabs.

Furthermore, Google will allow Gemini to not only summarize and answer questions for you, but also directly help you browse the web. In a demo, open Gemini Live in Chrome to help browse a recipe website. The user asks Gemini to scroll to the ingredients section and the AI ​​will quickly jump to that section. Gemini responded as the user requested Gemini to help convert the amount of sugar needed from cups to grams.

Yes, the imagination space of the future is still left to the AI ​​Agent.

02

Google Assistant walks away, Gemini Assistant takes over

Agent, which has been very popular recently, will naturally not be absent. You can see the clues from Google's ambitions for Gemini Assistant.

First, sort out the relationship: Google has had an assistant application called Google Assistant for a long time, but nowthe  Gemini application is taking on the expectations of a true "super assistant" in the future.

Gemini Live is an excellent manifestation of the role of "Assistant".

Under this function, users do not have to trouble describing what they see, or sending pictures and other materials to let the AI ​​analyze it, but can directly let the AI ​​"see" the content on the screen or let the AI ​​"see" things around through the camera. Then, AI can solve your problems like a real little helper.

Gemini Live has camera and screen sharing capabilities and is now available for everyone on Android and iOS for free, so users can point their phones at any object and talk via voice.

At the I/O developer conference, Google announced that it would expand its Frontier Gemini 2.5 Pro model to a "world model", which means it will be able to understand what it sees and develop plans according to Google. In terms of artificial intelligence, it is becoming more proxy.

Demis Hassabis, CEO of Google DeepMind, said the updates are "critical steps" to build a "general AI assistant" to better understand users and act on their behalf.

What new progress will be there in the future? Gemini Live is an extension of Google’s earlier project, Project Astra, an AI agent that leverages visual perception of the world around you. It is worth mentioning that at this I/O conference, Project Astra became more "active".

On the basis that you can use your phone camera to "observe" the objects around you, it can let it complete the task on your behalf, even if you do not explicitly ask it to do so. For example, it can choose to speak based on what it sees, such as pointing out errors in your homework.

Google's "universal AI assistant" is quite attractive. An assistant that can accompany you everywhere—whether in your phone or in a pair of augmented reality glasses—can sense the world, answer questions and convey information to you in seconds.

03

Gemini is actually an artist

Creative professionals and programmers, please note: Google's enhancements to its creative tools will either make your work easier and more efficient, or make you obsolete.

There are two iterations and two new products that are worth paying attention to.

First, talk about iteration.

At this I/O conference, Google iterated the image generation model and video generation model, and launched Imagen 4 and Veo 3, respectively.

Image 4 can render "fine details" such as fabrics, water droplets, and animal hair. The model can handle both photorealistic and abstract styles, and can create images with various aspect ratios and resolutions up to 2 K. "We have also put in a lot of effort and improved on the way it generates text and terrain, so it's perfect for making slides, invitations, or anything else that needs to blend images and text," Josh Woodward, head of Google Labs, said at a press conference. "

From the popular functions built into ChatGPT to Midjourney V7, there is a wide range of AI image tools on the market. What is the competitiveness of Imagen 4? In addition to the aforementioned expertise, Google also pointed out that Imagen 4 is very fast, faster than Imagen 3. And it will soon get faster. In the near future, Google plans to release a variant of Imagen 4 that will be 10 times faster than Imagen 3.

Veo 3, without doubt, is a competitor of OpenAI Sora. This time, Google has also put a lot of effort into differentiating its advantages.

That is—Veo 3's video + audio output, which can generate video with sound. For example, it can create bird videos with bird singing audio, or city street videos with traffic noise, or incorporate character conversations into the video.

This is not a tool that can automatically "dupe" the generated videos, but Google is very confident about how useful it is. Google says Veo 3 also performs well in real-world physics and lip sync.

Let's talk about the new product.

Google launches a new product, Flow, defined as an "AI movie production tool".

Flow integrates Veo, Imagen and Gemini to create cinematic editing and scenes. Users can describe the final output they want in natural language, and Flow will immediately make it for them.

Specifically, using Flow, users can use text-to-video prompts and material to video prompts (basically, share a few pictures, and Flow can use them in combination with prompts to help the model understand what you want) to create an AI-generated short film with 8 seconds. Users can then use Flow's scene-building tool to splice multiple short films together.

Google provides several "movies" cases made using Flow in its official blog post, which last for several minutes, and the characters, backgrounds and textures are quite stable.

The next new product has nothing to do with audio and video. Google has launched an "asynchronous coding agent," Jules, which aims to enable you to turn the rough design scribble on napkins into complete code or graphic design, while showing you the work it does in the process.

04

How can XR glasses be missing from my Google?

Another highlight that cannot be ignored in the keynote speech of I/O conference is the latest progress of Android XR (Android XR).

The system was released in December last year and was designed for extended reality devices such as XR headsets and smart glasses. It realizes the "contactless assistance" function through deep integration with AI.

It is not surprising that Google is eyeing the XR ecosystem. Google's Android system is one of the most important underlying operating systems in the smartphone era, and XR devices are regarded as a potential stock of the next generation of smart terminals. Previously, Meta wanted to develop an XR operating system and make "Android" in the XR era (but failed).

Google hopes to achieve what Android has achieved in the smartphone world in augmented reality, mixed reality and virtual reality.

This time, Google announced that it will work with glasses brands such as Gentle Monster and Warby Parker to launch more smart glasses products with a sense of fashion design. At the same time, Google will also link with more partners such as Kering Glasses. To further promote the technology ecosystem, Google will also deepen its cooperation with Samsung to jointly build a software and hardware reference platform designed specifically for glasses devices, and developers are expected to receive adaptation support later this year.

It is worth mentioning that at the conference, Chinese technology company Xreal released Project Aura, which is the second official device equipped with the Android XR platform. Google and Xreal plan to announce more about Project Aura at the Augmented Reality World Expo (AWE) in June 2025.