OpenAI’s Atlas Browser: Cool Technology That Lacks A Compelling User Experience for Consumers Today

Context: 

• OpenAI rolled out its own browser (Atlas) on October 21, 2025 - nearly three years after the launch of its ChatGPT powered by its GPT 3.5 series. 

• It didn’t define a target audience in its announcement - just used loose terms such as “us” and “you.”

• Core features include a type of omni-present chat interface, browser memory (not tracking), and an agent mode (i.e., a “machine” navigating one or more websites on your behalf). 

Why would they build their own browser?

Broadly speaking, most companies want to grow their share of consumers’ digital touchpoints and time spent online. Doing so gives them control (e.g., Alphabet, Amazon, Apple, Meta, etc.) and a phenomenal amount of data to allow advertisers to purchase targeted ads, deliver more relevant experiences, and even streamline experiences for consumers (i.e., create convenience). Conversational (text or voice) interfaces allow consumers to do many cools things, but they will not be the most convenient or best way to do all or many digital tasks. Ironically one of the guys demo'ing Atlas called out how tedious it is to type so much. 

However, investors will eventually expect a return on their investments through profits (less likely) or the sale of their equity stakes. Despite OpenAI’s 800M weekly active users (?), 10M paying Plus customers, and expected 2025 ARR of $13B (3rd party estimates based on monthly revenue), financial analysts estimated 2024 net losses at more than $5B and expect 2025 to be in the same ballpark or even higher. Yes, a  $500B valuation gives a company leeway. Any number of companies (especially Silicon Valley) prioritize growth over profits. This isn’t just software. Money is going to build infrastructure in data centers, hire talent, and train models. 

First, if consumers are willing to create “memories” with OpenAI (a.k.a. share data and allow themselves to be tracked), OpenAI like other LLM applications with reasoning capabilities will increasingly offer them convenience through simplified and relevant outcomes. No question. Consumers seek convenience when they want to get tasks done (e.g., write an Email, respond to a message, schedule a dentist appointment, etc.) Increasing consumer usage of OpenAI’s applications will create an obvious value flywheel - for users and the business. Recent court cases challenging Apple’s and Google’s business practices speak to the power of being a default browser or any type of portal to consumers.

Revenue opportunities are broad: paid services, revenue share, sale of data, training data to feed service improvement, etc. 

Second, more than 50% of OpenAI’s self-reported usage is for Practical Guidance (28.3%) or Seeking Information (21.3%) (Sept 2025 report). Brands want to target consumers in these moments. The global advertising market in 2024 was over $1T and grew nearly 10%. Digital advertising is nearly three quarters of that amount with Search representing about 40% of that. Search alone drove nearly $200B in revenue for Google in 2024. If OpenAI were to take even 10% of that number, they might eliminate their losses. Broadly, Google and Meta take home about half of all global digital ad revenue with TikTok not far behind. Digital advertising is more than 70% of all advertising. The overall market is growing. No reason OpenAI shouldn’t chase it. 

Ironically, if OpenAI were to challenge Apple’s and Google’s positions with consumers (e.g., smartphones, app stores, browsers, apps, etc.), it might help them in locations such as the UK where the UK’s Competition and Markets Authority (CMA) designated Apple and Google with strategic market status (SMS) in their respective mobile platforms. The decision will enable the regulator to take targeted actions to enhance competition in the space.

My experience: eh, not impressed - at least yet

Today’s Atlas experience did not offer me an improvement over my current browser experiences. Don’t expect consumers to flock to the platform yet. 

Agentic elements worked, but didn’t provide a better experience. The agentic elements of the browser are probably the least developed today. I asked it to do three tasks for me: 1) fetch information 2) research flights and 3) search. It didn’t complete any of these three tasks quickly with quality. I agree with OpenAI that today’s web experiences are less than ideal (i.e., have become Frankenstein-ish over time) - and have written a lot of research about the diminished importance of websites. Are there times when getting digital tasks done is tedious? Sure.  It’s also very first-worldy to claim that surfing the Internet or working on a computer is hard. Ironically, several of the demos were on Google services. 

I felt like I was signing away my life’s personal data in perpetuity when I downloaded and installed Atlas on my Macbook. On one hand, it asked for a lot of information (e.g., bookmarks, password access, browsing history, etc.) that it could use to make my transition from Chrome to Atlas more seamless. (FWIW, I gave into about half of their requests - and they asked for a LOT). Allowing them to track me (i.e., create memories) is a reasonable request if I want more contextual experiences that simplify tasks for me over time. Yes, I know this isn’t much different than what I have given to Google over the years - just a lot at one time. 

Security warnings and control features will exceed most consumers’ bandwidth and skill set. The set up came with a lot of language about the risks of using agentic services. Consumers just want to feel safe. They don’t want to actively manage security on their devices and accounts. For the most part, they don’t have the ability to think through the risks on a request by request basis. Seeking information or practical guidance won’t trigger too much concern. If consumers start transacting, they will simply want to feel safe. Apple gets this better than any other entity - simple or convenient + safe = high usage.

Final thoughts? I can't wait to see what they do next. I use their application nearly every day. My use cases continue to evolve. I understand that releasing products like Atlas early and often to learn is part of the process.