AI browsers are here, and you need to learn how to use the web properly – Digital Trends
About a month ago, I gave a tech demo to a bunch of freshman students on how to create a custom skill in an AI browser and automate the research work on assignments. Instead of bogging them down with the wild goose chase on Google Search, the “AI agent” limited its search to only a handful of academic and learning sources to provide the answer.
I did it all by simply typing “/course,” followed by “Faraday’s Law of Induction.” The summary and answers offered by the browser were strictly from the school syllabus, and nothing too deep or shallow. The whole approach is fast, efficient, and removes the unpredictability of an AI just spitting out jargon from poor sources or simply hallucinating.
The group, which included my brother, was astonished. It was a whole new way of finding relevant information on the web. Instead of having a generic AI chatbot do the job using its own understanding of which source is good or worth skipping, they can now dictate exactly where they want their answers sourced from.
It’s a whole new world and a dramatic reimagination of how we interact with web browsers. OpenAI’s ChatGPT Atlas, Dia, Perplexity’s Comet— and to some extent, Google Chrome and Microsoft Edge — are firmly pushing into that direction. Here’s a quick primer on the most notable ways AI browsers are changing the game.
Legacy web browsers limit the information you can absorb to the same tab. For background research, contextual questions, or just verifying a basic fact, you must launch another tab, or open a new window
In a nutshell, tasks quickly get cluttered. AI browsers solve that problem with a persistent side panel where you can talk with a built-in assistant about what’s appearing on the screen, go off-topic for background research, collectively pull information from other tabs, and do a lot more in the same place.
I was recently reading an article about a breakthrough that essentially turns concrete into an energy-storage device. The paper discussed all the ways it could be deployed, and one of those proposals was carbon sequestration. I was not entirely sure about it, so I simply opened the built-in ChatGPT assistant in the side panel and asked, “What is carbon sequestration?”
I got my answer in the side panel and asked further questions, as well, including a visual explanation using a diagram created by the built-in image generator. All this happened in the side panel, without having to juggle between multiple tabs and windows.
Going a step further, I can launch a full-fledged web search in the same side panel, create a comprehensive Deep Research report, and add my own sources (from connected services such as Gmail, Drive, Notion, etc). I can also enter the Agent Mode and get work done.
In this context, I enabled the autonomous browsing mode and asked it to pull all the articles published by MIT on the topic of smart concretes that can also store energy. In roughly two minutes, I got a long list of such articles neatly arranged in a table, with a summary and the MIT subdomain that published them.
All this happened in the background, while I continued working on another tab. The sidebar in AI browsers — whether it’s ChatGPT in Atlas, Gemini in Chrome, or Copilot in Edge — is a dramatically advanced tool that dilutes the drudgery of web-based tasks and saves a lot of time, as well.
One of the most impressive tools that is now available in AI browsers is the system for creating custom skills and shortcuts. Think of them as one-shot commands that perform a specific job. These tasks can be anything you want, and work just the way you describe them into existence.
For example, I created a custom GPT called “Research Assistant.” When I summon it using the “/” command, it performs the described task in a new tab, or the assistant sidebar in the same window. A similar system called shortcuts and skills is available in Comet and Dia, as well. The image above shows a sample of creating one in the Comet browser.
Here’s how it works. You pick a random name, describe what you want to do in a browser, and hit save. For example, I have created one for shopping, called, well, “shopping.” And here’s the best part. You simply describe what you want the skill to do, and it will be configured as such.
I was recently checking out a pair of wireless earbuds on Amazon, but wanted to compare the price on other websites, as well. Instead of opening each shopping site and manually searching for the product, I simply typed “/shopping” in the assistant sidebar in the same tab, and it gave me a list of all sites selling the same pair of earbuds, alongside the price and link.
On a similar note, the browser agent can not only expand user queries across multiple tabs, but it can also combine information from all the active tabs. For example, if you have hotel listings open across half a dozen tabs, you can simply use an “@” shortcut in the sidebar and ask the assistant to summarize all the crucial information, such as price, amenities, date-wise availability, distance from the airport, and more.
The idea, once again, is to extract context from the active tabs. And thanks to multi-modal awareness, the AI can understand text and images, too. For example, by looking at the picture of a meal, the AI can tell you the ingredients and nutritional info, or list all the tourist spots in the city by looking at an image of a Jordanian market.
You just have to ask.
AI skills offer an unprecedented level of control and convenience, while combining with the sidebar facility described above. This is an immensely powerful automation system that can save you a lot of time, tiresome browsing, back-and-forth tab switching, and clutter.
One of the most promising sides of AI browsers is their ability to connect with third-party services and hand over the reins to AI chatbots. For example, Perplexity’s Comet and ChatGPT can also connect with your Gmail inbox. WhatsApp, Slack, cloud storage services, and more.
With agent mode, you can go shopping or make reservations with just a text prompt. It works autonomously by handling the clicks and types on your behalf. It’s pretty surreal to witness in action, but that’s where the problems begin.
Can you trust an AI agent (and an integrated browser) with sensitive data, such as login credentials? In ChatGPT Atlas, for example, the browser remembers not only your web surfing activity, but also your chat history. And the way it logs all that information is extremely scary. It opens a whole new world of hyper-personalized advertisement and privacy risks.
Experts are also wary of risks, not just for the traditional web activity pattern, but for the whole new surface of attacks that has been opened by AI chatbots and AI browsers.
“The more power users give to AI, the higher the risks. If one’s browser can order products online, it would always be possible for a malicious attacker to force it to ship another product — potentially something unpleasant or even dangerous — to that address,” Leo Feinberg, co-founder and CEO of Verax AI, told Digital Trends.
Aside from the obvious risks of hallucination (aka giving out wrong information or made-up facts) and bias, there are serious security risks that tag along with AI-driven web browsing. The folks over at Brave Software Inc. discovered a prompt injection risk in Perplexity’s Comet browser.
In a follow-up report, the maker of the eponymous privacy-centric browser detailed a similar vulnerability in another AI-powered browser. “If you’re signed into sensitive accounts like your bank or your email provider in your browser, simply summarizing a Reddit post could result in an attacker being able to steal money or your private data,” it reported. The target could be your banking operation, medical data, stuff stored in the cloud drives, email inbox, or more.
Johnny Hughes, CMO and Chair of the AI Council at AI-driven communications company Avenue Z, tells DigitalTrends that web browsing is evolving beyond manual search and keywords to conversation format, just the way we interact with AI chatbots to find answers. Interestingly, he noted that there isn’t a definite good or bad side of stuffing AI into the core web browsing experience.
“These tools are contextual engines trained on our behaviors, inputs, and queries. What happens when your browser knows more about you than your partner?” he adds. The risks are grave. In August, user conversations with ChatGPT and xAI’s Grok chatbot were leaked, exposing thousands of conversations on Google Search. Back then, experts pointed out that chatbots are a deeper privacy disaster than browsers.
Mixing the two sure sounds like a calamitous recipe. “If anyone is considering being an early adopter, I’d recommend using these tools cautiously, giving them minimal permissions and blocking their abilities for potentially damaging actions,” warns Feinberg, whose company works at the intersection of AI and security.
An engineer at an AI company, talking on condition of anonymity, told me that despite working on consumer AI products, they are wary of connecting all the services you use with a single chatbot ecosystem. AI assistants are a lot more personal than web browsers, and linking them to other products that you use on a daily basis — from Amazon to Spotify — is like letting an AI company profile your entire life.
One has to draw the line, one where they can balance the convenience of AI browsers with their privacy perils. The shift, however, is inevitable.
The idea of a paid web browser is pretty controversial, and so far, there haven’t been any that have gone truly mainstream. After all, when you have feature-packed options such as Chrome and Safari that are freely available, why pick something less functionally charming?
So far, only Brave has managed to find a foothold with a premium product by offering a VPN and enhanced privacy measures. But the browser landscape is changing rapidly, all thanks to AI. And it’s not just the new entrants, but also the heavy-hitters like Chrome that are adapting.
The ethics of talking to an AI chatbot, and what kind of information they can return in return, is a topic of hot debate. The risks of misleading medical information, incitement to violent actions, and detachment from real-world experiences stir intense conversations. But it seems the language you use while talking with AI tools such as ChatGPT and Gemini also affects the quality of answers you get. As per fresh research, being rude could be more useful than acting polite.
The big picture
The problem of biases has plagued AI chatbots ever since ChatGPT landed a few years ago, and changed the whole landscape of conversational assistants. Research has repeatedly uncovered how chatbot responses show gender, political, racial, and cultural bias. Now, OpenAI says that its latest GPT-5 model for ChatGPT is the least biased, at least when it comes to politics.
What’s the big story?
Upgrade your lifestyleDigital Trends helps readers keep tabs on the fast-paced world of tech with all the latest news, fun product reviews, insightful editorials, and one-of-a-kind sneak peeks.
