StartseiteArtikel

Bei AI-Browsern hat Google im Vergleich zu OpenAI nur den Mangel, dass der Chef nicht live Waren verkauft.

字母榜2025-10-23 20:04
Nach Chrome suchen, Chrome konkurrieren, die Nutzer stellen fest, dass dies Chrome ist.

In 2025, every time a new product is launched in the AI browser industry, Google's Chrome browser seems to face a new threat.

The most recent one, of course, is Atlas newly released by OpenAI. In the news flash titles, phrases like "Google's biggest fear" and "Chrome assassin" often appear. Coupled with the news that Google's stock price fluctuated accordingly and dropped by up to 5% during intraday trading, the future of Chrome browser really seems to be in danger.

However, similar evaluations could also be seen after Comet, an AI browser developed by Perplexity, was launched in July 2025, but the voices were much quieter.

For AI browsers, whether emphasizing the AI performance of the product or its browser positioning, Google Chrome cannot be bypassed. Google will not be absent from the wave of AI browsers.

01

In the current upsurge of AI browsers, giant enterprises like Google and Microsoft, which have both self - developed large - model bases and browser products, are accustomed to gradually introducing large - model functions into their existing browsers.

The advantage of this path - dependence is that the cost of integrating resources is low, and the product has relatively strong and predictable robustness. The drawback is that they can't, like OpenAI and other AI startups, make a new product similar to Gemini in Chrome a highlight of their marketing campaign.

Now, more players are entering the AI browser market, but the functions of their products are gradually converging. Since the functions of the base large models are gradually moving from the "revolution" stage to the "evolution" stage, the scores of large models in various benchmark competitions can't directly bring the intuitive impact of the ChatGPT miracle to users. As a result, users will find AI browsers based on these models quite similar.

Yes, most of the amazing functions of OpenAI's Atlas browser can be found in Google's Project Mariner announced at the end of May and the finished product Gemini in Chrome delivered in mid - September.

There was no live press conference for Gemini in Chrome. Instead, Parisa Tabriz, the vice - president of the Chrome department and the product lead, posted a blog article on Google's official website on September 18, showing the transformation of Chrome after integrating the Gemini model.

When OpenAI's Atlas browser was launched, it emphasized the integration of all pages into AI and the memory function of users' usage records and browsing history. Google's Gemini in Chrome had already updated this function more than a month ago.

In the example on Google's official blog, when a user clicks on the Gemini star icon in the Chrome browser, a side dialog box pops up. Without opening a new tab, the updated built - in AI function in Chrome can directly retrieve a webpage about a team - building activity that the user viewed last week from the browsing history. When the user asks further about "team - building activities for groups of less than 10 people that last less than half an hour", the AI automatically extracts the correct content from the webpage to answer.

When OpenAI's Atlas browser was launched, it emphasized the "agent mode". GPT can directly operate web pages and complete work tasks, evolving from a "window for browsing" to a "hand for doing things".

Chrome also updated a similar function more than a month ago. In the example, after a user enters a sentence in the sidebar, the AI agent reads a recipe in an email and automatically orders all the ingredients in an online store and adds them to the shopping cart according to the recipe. Except for the initial instruction and the final confirmation of payment, the user doesn't need to intervene in the workflow of the AI agent at all.

OpenAI's Atlas browser emphasizes "Chat comes with you anywhere", meaning that all pages and functions are seamlessly integrated with all the functions of GPT in real - time. When users are reading web pages, listening to podcasts, or watching videos in the browser, they can call out GPT at any time to summarize the content they are browsing.

Coincidentally, Chrome also demonstrated a similar function more than a month ago. When a user is watching a YouTube video and clicks on the Gemini star icon to ask "Summarize this video", the AI immediately provides an outline summary and the corresponding timestamps of the key points in the video.

Now, looking back, the essential differences between Gemini in Chrome and Atlas browser may lie in:

1. Different base models: one is GPT, and the other is Gemini;

2. Different release methods: Google's product lead quietly updated the official website blog, while OpenAI's celebrity boss led the R & D team in a live - streaming product promotion.

That is to say, there is no obvious performance difference between Gemini in Chrome and Atlas browser, but there is an obvious generational difference in the marketing methods of Google and OpenAI. As the saying goes, great minds think alike. Having the boss appear in person in a live - streaming sales event is the spirit of the times.

02

Most companies aspiring to become leaders in the AI browser industry have gone through the process of "seeking to buy Chrome, benchmarking against Chrome, and having users realize it's just like Chrome" in the past year.

In 2024, a US court ruled that Chrome browser "strengthened" Google's dominance in the online search market, violating antitrust laws.

On November 21, the US Department of Justice and antitrust enforcement agencies recommended that Google be forced to divest and sell Chrome browser, "permanently ending Google's control over this key search entry point and enabling competing search engines to enter this browser, which is the gateway to the Internet for many users".

The first piece of information that OpenAI was planning to develop an AI browser was released on November 22. According to a report by The Information on the same day, "according to people who have seen the product prototype or design", OpenAI was considering developing a web browser integrated with its AI chatbot. OpenAI was laying the public - opinion groundwork for a possible future where Chrome would no longer be owned by Google.

OpenAI was not the only one eyeing Chrome at that time.

During the three - week supplementary trial, on April 22, Nick Terry, the head of ChatGPT, was questioned in court as a witness: if the federal court ruled that Google must divest Chrome browser, would OpenAI seek to acquire it? Nick Terry replied, "Yes, we would, and many other companies would too."

By the end of the supplementary trial phase of Google's monopoly case in mid - May 2025, more than 20 industry witnesses had appeared in court, and almost everyone expressed their company's desire to acquire Chrome.

Google also first revealed that it was developing an AI browser during a court hearing around the same time. On April 25, during a several - hour hearing, Parisa Tabriz, the Chrome product lead, said that the Chrome team was working on deeply integrating AI into the browser to enhance Chrome's "agent ability" to automatically perform tasks on behalf of users.

Combined with the previous public - opinion situation, this was Google directly stating in court that it would be a competitor to OpenAI. The "Project Mariner" announced at Google I/O Developer Conference at the end of May 2025 was an additional update on the development progress.

As late as August 2025, some companies were still trying to buy Chrome. On August 12, Perplexity AI announced that it had sent an acquisition offer for Chrome browser to Google's parent company Alphabet, offering $34.5 billion, far exceeding the startup's own valuation.

In hindsight, the suggestion to force the sale of Chrome was a legacy of Lina Khan's political career before its end. Lina Khan, the chair of the US Federal Trade Commission during the Biden administration, and her mentor and colleague, Professor Tim Wu, had been obsessed with breaking up industry giants like Facebook and Google both before and after becoming government officials.

However, the current US government has different priorities. The revenge - seeking attitude of the core decision - making layer to "reverse the policies of the previous administration" and various strange and unrestrained measures have made it impossible for the task of splitting Google's main business to be prioritized on the to - do list of the executors.

It was not until the prospect of buying Chrome became slim that OpenAI and companies like Perplexity became determined to develop their own AI browsers. On July 9, Reuters reported that OpenAI was preparing to launch an AI - driven web browser. On the same day, Perplexity AI announced the launch of its AI browser Comet, exclusively for registered users of the $200 - per - month Max package. On October 2, Comet browser was made free for global users.

Even when OpenAI launched its Atlas browser on October 22, the timing was deliberately set to compete with Google. After the head of Google AI Studio announced "a major update" on October 20, a screen recording circulated in the industry, saying that Gemini 3 would be announced on October 22 and launched on November 12. The launch time of Atlas browser clearly aimed to "pre - empt" Google.

03

Similar to other AI product segments, the AI browser industry rapidly evolved from a wild frontier where startups and large companies competed fiercely in 2025 to a stable market dominated by large companies.

The most typical example is Browser Company and its AI browser Dia.

In 2024, Browser Company was valued at $550 million, and it had stopped developing the Arc browser, which had made it popular two years ago. As its CEO publicly admitted, the usage rates of various functions of Arc were extremely low. Back - end data showed that only 5.52% of daily active users regularly used multiple Spaces, and only 4.17% of users used Live Folders.

Browser Company's solution was to abandon Arc and recreate its functions using AI tools, resulting in the new browser Dia. After Dia was launched in June 2025, it was sold to Australian software company Atlassian for $610 million in cash in September.

Dia is less of a product for ordinary users and more of a financing tool to save the company's future. Browser Company has no ambition to compete with Google and others to seize the leading position in the AI browser wave. It is just glad to get out of trouble.

In the AI browser segment, the Matthew effect of "the rich get richer" is so significant that small - scale companies like Perplexity AI are forced to pay users to use their products. On October 23, Perplexity's homepage announced that users who successfully invite new people to register for the Pro package of Comet browser will receive a $20 rebate for each new user. Newly registered users of Comet can use the Perplexity Pro package for free for one month.

The determination of Google and OpenAI to develop AI browsers will not be shaken.

Google needs to use AI to consolidate its throne in the browser market. According to data from market research firm Statcounter as of August 2025, Chrome accounts for 70.25% of the global PC desktop browser market and 69.15% of the mobile browser market. A situation where "the combined scores of the top four runners - up, even when doubled, are still lower than the first - place finisher" disappearing because of its not - lagging - behind AI products will not be tolerated by any company's decision - making layer.

And OpenAI needs the browser as an entry point to refine its AI model. What users say in an AI chatbot may not be true, while the web browser, a basic digital tool since 1993, can capture users' purest intentions and thought - chain data.

Every click, every search, the duration and result of each micro - decision of users reveal the most honest choices that users themselves may not even be aware of or express, and at the same time, they are also used to calibrate and label the model's generated answers against real - world benchmarks.

The web browser is a pipeline to contemporary humans' basic thinking and desires. Whoever controls the browser - like product market in the AI era will have the most valuable and high - quality training and alignment datasets.

The large AI model and the web browser are now like the roots and leaves of a tree. They nourish each other when they cooperate, and it will be difficult for either to survive if they are separated. This is an unchangeable and unnecessary - to - change future trend.

This article is from the WeChat official account “Zimu Bang” (ID: wujicaijing), author: Li Xi, published by 36Kr with authorization.