ChatGPT Atlas Unleashes Web Browser War - What Are the AI Alternatives?

"Experienced" internet users may still have nostalgic memories of it: how we once scoured the internet via the Mosaic web browser, then Netscape Navigator, followed by Internet Explorer, and eventually Google Chrome. For years, Chrome was the undisputed king of the web. But just like in previous technological revolutions, movement is stirring once again. The new battle is no longer about speed or extensions it is about intelligence.

We are at the beginning of a new browser war, waged by a generation of agentic AI browsers: browsers that don't just show what you're looking for, but also understand what you mean and can act on your behalf. OpenAI’s Atlas, Perplexity’s Comet, Arc by The Browser Company, and even Anthropic’s new extension for Chrome show that the lines between 'searching', 'asking', and 'acting' are blurring.

Yet there is more at play than just innovation. Under the hood, a battle is raging for data, platform power, and user trust. Because whoever controls the browser controls not only access to the web but also the massive data stream that feeds AI models.

Atlas by OpenAI – The smart assistant that reads along with you

OpenAI launched its own browser in October 2025: Atlas. At first glance, it looks like just another variant of Chrome: based on Chromium, compatible with extensions, and featuring a familiar interface. But beneath that familiar layer lies something new.

Atlas combines a classic browser with the intelligence of ChatGPT. In the sidebar, your personal AI agent appears, contextually reading along and making suggestions based on what you are doing at that moment. Writing an email? Atlas offers to rewrite or analyze it. Searching for something? It summarizes the key results and suggests follow-up questions.

A striking feature is Browser Memory a kind of long-term memory that allows Atlas to remember which sites you've visited, what you searched for previously, and what questions you asked. This allows you to ask weeks later: "What was that article about cement that could store energy?" and Atlas knows exactly what you mean.

Additionally, there is Agent Mode, in which the browser can perform tasks: such as booking a hotel, comparing products, or preparing an order.

It feels like a logical next step: the AI assistant inhabits your browser and helps you everywhere, instead of you having to open a separate chat. Simultaneously, it raises questions about privacy and control especially if the browser knows more than you do.

Comet by Perplexity – The first true agentic browser

While Atlas shows some caution, Perplexity’s Comet takes a different approach. Comet proudly calls itself the first agentic browser: a browser that doesn't just react but acts independently.

Built on Chromium but completely redesigned around AI interaction, Comet offers an innovative browsing experience. The browser can understand context, automatically summarize information, fill out forms, and even prepare purchases all via natural language. This is possible through two buttons next to the "search bar": summarize all information from the current webpage, or start a conversation with the AI assistant by simply talking to it.

However, the browser also has limitations. You cannot customize the homepage everything remains linked to Perplexity. And with that centralization comes a growing dependence on a single party that records your search behavior, preferences, and commands.

Arc – Design, productivity, and AI as a tool

Arc, developed by The Browser Company, takes a different route. Where Atlas and Comet aim for full AI integration, Arc focuses on design and productivity. The browser was built on the premise: “Why does the web still look like it’s 2008?”

Arc breaks with conventions. Tabs are no longer at the top but in a vertical sidebar. You work in Spaces, separate environments for projects or themes. Via Boosts, you can visually customize websites, and with Easels or Notes, you can capture ideas without external apps.

Since 2024, Arc has expanded with Arc Max, an optional AI layer. This allows you to summarize pages, automatically generate titles, or ask questions about the site's content (“Ask on page”). It is AI as a tool, not as the driver.

On mobile, Arc has a specific asset with Arc Search: the “Browse for Me” function searches on its own, reads multiple pages, and directly displays a summary. This gives you an answer in seconds without opening dozens of tabs.

Arc therefore does not offer the full autonomy of Comet or Atlas, but rather a smarter, calmer way of working. It is especially embraced by creatives, writers, and researchers not because it takes everything over, but because it alleviates the workload.

Claude for Chrome – AI as an extension

Not every innovation requires a new browser. Anthropic, known for the Claude models, chooses a pragmatic approach: adding AI to the browser you already use.

With Claude for Chrome, an experimental extension, you can already benefit from agentic features within your existing Chrome environment. The AI can read webpages, understand context, perform tasks, and, for example, automatically fill out forms. Claude for Chrome demonstrates how AI can be incrementally integrated into the browser without overhauling everything.

Ladybird – An alternative without AI

Amidst the AI craze, an outlier appears: Ladybird. This open-source browser is not based on Chromium, WebKit, or Gecko, but was built entirely from scratch. That alone makes it unique in an era where almost all browsers are derivatives of Chrome.

Ladybird originated from the SerenityOS community, with the goal of creating a browser that is truly independent of Big Tech. No data collection, no tracking, no integration with AI services. Just a clean, efficient browser that respects open standards.

While Ladybird is still in development, it shows that there is room for alternatives that do not lean on the AI trend. In a way, it is a counter-reaction to the agentic wave: a reminder that there is also value in simplicity and transparency.

Why do all these companies want a browser?

The reason so many parties are entering the browser market simultaneously is twofold.

Firstly, it’s about data. A browser is the central entry point to an individual's digital life. Whoever knows where you surf, what you search for, and how you navigate possesses invaluable behavioral information. For AI companies, that is pure fuel: with more context, their models can become more relevant, personalized, and powerful.

Secondly, there is an economic incentive: the rise of the so-called Agent Commerce Protocol (ACP). This new paradigm should make it possible to buy products and services directly through AI agents, without a traditional webshop. A kind of e-commerce 3.0, where the browser is not just a window to the web, but a marketplace itself.

In other words: whoever controls the browser controls both the user and the transaction. And that makes this battle strategically more important than ever.

The risks of agentic browsers

The promise of agentic browsers is immense: a world where software doesn't just listen, but also acts. Yet there is a dark side to this progress. Because agentic browsers like OpenAI Atlas and Perplexity Comet can perform tasks independently, they also introduce new vulnerabilities.

One of the primary risks is the rise of prompt injection attacks a relatively new but fast-growing security problem. In these attacks, malicious actors hide instructions on webpages, often in invisible text or in the source code. You see a normal website, but the AI reads everything including the hidden rules. Those instructions can prompt the model to perform actions you never authorized.

Because agentic models can have direct access to files, sessions, or even payment accounts, the damage can be substantial. A hidden prompt could, for instance, instruct the agent to forward sensitive data, open files, or initiate transactions. In practice, this amounts to digital social engineering but at a machine level.

The problem is not so much the AI itself, but the autonomy it is granted. The more rights a browser agent has, the greater the risk that someone will abuse that power. And as long as those agents are not properly sandboxed, small leaks can have major consequences.

Fortunately, developers are already working on countermeasures. OpenAI, Anthropic, and Perplexity are experimenting with sandboxing, where agentic functions are separated from regular browsing. Explicit consent requests are also being added more frequently ("May the agent read this email?" or "Do you want it to fill out this form?"). Such safeguards are crucial to prevent the AI from acting unseen.

For users, simple advice applies for now: be cautious. Use agentic functions preferably with limited permissions, do not give the AI direct access to sensitive documents or passwords, and keep business and personal accounts strictly separate. For those who want to experiment safely, it is better to use ChatGPT or Claude in a normal browser without extra system access. This allows you to benefit from the advantages while limiting the risks.

In summary: the technology is promising, but the security is still in its infancy. Agentic browsers are the future but for now, primarily for those who understand the risks and how to manage them.

Is this the end of the Chrome glory years?

The arrival of this new generation of browsers raises a provocative question: are Chrome's days numbered?

On one hand, the answer seems to be "yes." Chrome's market dominance is finally being challenged, not by faster or lighter alternatives, but by browsers that are genuinely smarter. Atlas, Comet, and Arc show that the browser of the future is not passive but actively thinking and participating.

On the other hand, there is the human factor. Humans are creatures of habit. Chrome is deeply embedded in our daily lives: in our bookmarks, extensions, password managers, and work processes. Even if new browsers are better, switching will take time and trust.

Therefore, there will likely not be a sudden shift in power, but a gradual transition. First in niches: tech enthusiasts, productivity fanatics, and companies experimenting with AI. Then, as the technology matures and becomes safer, the masses will follow.

Arc could very well be the ideal intermediate step: a browser that uses AI as a tool rather than a driver, making it feel less intimidating than a fully agentic system.

The future of the web: From clicking to co-creating

What once started as a simple window to information is now evolving into an intelligent interface that thinks, writes, and participates with you. Browsers are changing from passive tools into active partners and with that, our relationship with the internet is fundamentally changing.

In the coming years, the dividing line between app, browser, and agent will blur further. Perhaps soon you won't say "I'm opening Chrome," but rather "I'm asking my browser to handle it."

Whether that is good or bad depends on how we handle it. Technology itself is neutral it is our choices that determine whether it empowers us or makes us dependent.

Conclusion

The new browser war is more than a battle for market share; it is a battle for the future of our interaction with the web. Atlas, Comet, Arc, and Claude each bring a vision of that future: from total autonomy to subtle assistance.

But with that innovation comes responsibility. Agentic browsers can make our lives easier, but also riskier if we trust them blindly. Those who want to utilize the power of AI must also learn to think like a developer: critically, consciously, and prepared.

One thing is certain: the coming years will be anything but boring. The browser, once merely a window to the internet, is now becoming the architect of our digital world.

// About the author

Remy Gieling

Mede-oprichter, AI-expert & bestseller-auteur

Tech-expert (1988) gespecialiseerd in kunstmatige intelligentie en mede-oprichter van ai.nl, The Automation Group, Proxies en eBrain.ai. Oud-hoofdredacteur van diverse zakenmerken en daardoor een geoefend verteller op het podium en in de media. Verzorgt jaarlijks 150+ AI-keynotes in binnen- en buitenland en is gastdocent aan Nyenrode. Co-auteur van zeven boeken, waaronder 'Handboek AI Strategie' en 'AI Agents', en bekend als presentator op radio en RTL Z. Reist langs de labs van OpenAI, Nvidia en Tencent en vertaalt de nieuwste doorbraken naar inzichten die leiders direct kunnen toepassen.

LinkedIn