217: The Browser Wars Are Back. This Time With AI Agents!

In the 1990s, Microsoft and Netscape fought for control of the browser, the gateway between humans and the internet. Netscape went from 90% market share to zero in five years. Now, with over 30 agentic browsers launching in under 18 months, the same war is playing out again, only this time the stakes are higher. This episode breaks down the 90s browser wars, compares the tactics to what's happening today, and explains what website owners should do about it.

Key takeaways

1. The playbook hasn't changed - Bundling, free products, proprietary lock-in, and distribution deals decided the 90s browser wars. The same tactics are playing out with agentic browsers today.
2. Google is running Microsoft's 1995 playbook - Microsoft embedded IE into Windows to protect its OS monopoly. Google is embedding Gemini into Chrome to protect its search monopoly. The browser is the defensive weapon, not the product.
3. The Chromium trap is deeper than IE bundling ever was - Most agentic browsers (Comet, Atlas, Neon) run on Google's Chromium engine. Even competitors are built on Google's foundation.
4. The prize shifted from attention to transactions - The 90s fight was about what people see. The agentic browser fight is about what AI agents buy, book, and do on your behalf.
5. Your website is the new Netscape - If AI agents mediate every user interaction, your site risks becoming invisible infrastructure rather than a destination.
6. Regulation will be too late - The DOJ took 6 years to settle with Microsoft. Netscape was already dead. The same timeline is playing out with Google's antitrust case.

What to do today

• Don't optimize for one agentic browser. Build for web standards: semantic HTML, ARIA labels, structured data, server-side rendering.
• Build direct audience relationships (email, communities, subscriptions) so you're not dependent on browser intermediaries.
• Make your site worth visiting, not just worth scraping. Offer value an AI agent can't replicate.
• Treat accessibility as an agent strategy. Screen reader compatibility = AI agent compatibility.
• Test your site with an agentic browser to see what works and what breaks.

Read the full agentic browser landscape breakdown: nohackspod.com/blog/agentic-browser-landscape-2026

Chapters

• 00:00 - Introduction
• 01:34 - The First Browser War
• 09:15 - The Agentic Browser Explosion
• 12:48 - Why Is This Happening Now?
• 16:15 - Where the 2026 Version Gets Worse
• 21:27 - What This Means for Your Website
• 23:14 - What to Do About It
• 26:49 - Closing

Connect

Website: https://nohackspod.com

LinkedIn: https://www.linkedin.com/in/slobodanmanic/

Newsletter: https://nohackspod.com/subscribe

No Hacks is a podcast about web performance, technical SEO, and the agentic web. Hosted by Slobodan "Sani" Manic.

Transcript

Agentic Browsers
===

[00:00:00] In 1995, A then 24-year-old named Mark Anderson said that his company, Netscape would reduce windows to a poorly debugged set of device drivers. Microsoft's response, internal emails. That said, and I'm quoting from DOJ trial records, you're cut off. Their air supply. Within five years, Netscape went from 90% market share to zero.

[00:00:28] Gone. A $3 billion company turned into a history footnote. Now fast forward to early 2026. I recently published a guide cataloging the Gentech browser landscape, and I counted over 30 of these products. 30 AI power browsers that don't just show you webpages. They navigate, they click, they try to fill out forms, compare products, even make purchases on your behalf.

[00:00:55] 18 months ago, this category did not exist 12 months ago it barely existed and now there are over 30 of them backed by the biggest companies in tech fighting to control how AI interacts with the web on our behalf. We've seen this movie before and honestly, this tracks, if you look at Hollywood, they cannot stop remaking the nineties movies right now.

[00:01:18] Twister, twisters, Beetlejuice, Beetlejuice Scream seven, whatever it is, Jumanji, the Crow. Apparently the tech industry got inspired by Hollywood

[00:01:28] Because they're remaking the browser wars. ​

[00:01:34] let me take you back to 1993, a college student. Mark Anderson is at the University of Illinois, and he built something called Mosaic. It is the first web browser that puts text and images together in the same window. The first browser that does that 1993. Before Mosaic, the web was just text after Mosaic, regular people could actually use the internet. Anderson leads Illinois and teams up with Silicon Graphics founder Jim Clark, and they start Netscape in April, 1994. Netscape Navigator was launched in December that year. It was free for non-commercial use by 1995. Netscape has about 90% of all browser usage.

[00:02:18] 90% they own the web and then. August 9th, 1995. They go public. The stock was supposed to be offered. This is from the Wikipedia page, the stock was supposed to be offered at $14 a share. The last minute they double it to 28. On opening day, it went up to 75. A 16 month old company valued at nearly $3 billion.

[00:02:41] I'm seeing a lot of these. In 2026 that IPO is widely credited with kick-starting the entire.com bubble. We know how that ended. One week later on August 16th, seven days later, Microsoft reacts by releasing Internet Explorer 1.0. This was not a coincidence. Of course, bill Gates at that time had already written his internet tidal wave memo.

[00:03:09] In May. So a few months earlier, 3000 words sent to every executive at Microsoft calling the internet the most important single development since the IBM PC and even singling out Netscape by name, the Constitution Center has the full memo archive. It is worth reading, but let's just focus on this. This is the moment where Microsoft redirected the entire company.

[00:03:34] And their entire strategy toward the internet by December 7th, 1995. That is four months later, bill Gates, and he's in the news today, isn't he? Poor Bill? He's holding a press conference declaring Microsoft is hard core about the internet. That date was deliberate because December 7th is Pearl Harbor Day.

[00:03:56] Netscape veterans actually called it dare Pearl Harbor Day. Now, what happened next is the important part because the tactics are what matters for our story we're telling today. Microsoft first move was bundling Internet Explorer , shipped free with every copy of windows, which ran on 90% of the world's species.

[00:04:15] this is 1996. Dial up internet. Downloading a browser, takes time, takes effort. The connections were slow. Most people just used whatever they had on their computer and they did for, for a long time after. Now, second, the strategy that got exposed during the DOJ trial as embrace, extent, extinguish.

[00:04:38] Step one, adopt open standards like HTML and JavaScript. Step two, add proprietary features that only work in your browser. ActiveX VB script. Step three. Once those features become widespread, competitors can't keep up because they can't support your proprietary extensions. Relevant today. And the third thing, distribution deals.

[00:04:57] Microsoft paid a OL. A great company as well. Yes, we we'll hear more about them in this episode. They paid a OL 25 cents per user to make IE Internet Explorer the default browser for AOL's five plus million subscribers. In return, and you may remember this, if you had a computer, the nineties, a OL got an icon on the Windows desktop.

[00:05:21] They just swapped distribution at scale. The result, Netscape market share went from 90% in 1995 to 80, in 96 to 42 by mid 98. To under 5% by 2001 Internet Explorer. God bless its soul peaked above 90%. The detail I really like about this story, and I said I'll talk about a OL more. They bought Netscape for $4.2 billion in November, 1998.

[00:05:50] Day, continued bundling Internet Explorer as a default browser inside their software because they signed a contract, but they paid $4.2 billion for a browser. They couldn't even recommend. How are they doing today? These days, the web was completely fragmented. Netscape had its proprietary features.

[00:06:10] Microsoft had its proprietary features. Developer had to choose sides or build everything twice. If you ever built a website in the IE six and IE seven era, you know exactly what I'm talking about. Do you remember those best viewed in Netscape or best viewed Enter Explorer badges? That that was the web In late nineties and early two thousands, the DOJ filed.

[00:06:32] The antitrust against Microsoft. In May, 1998, the judge ordered a breakup In 2000, it was reversed on appeal. The settlement in 2001 was as these things always end up being just a wrist lap, but by then Netscape was dead. And the part that connects directly to where we are today after winning, Microsoft essentially stopped developing, ie.

[00:06:57] There were five years between IE six in 2001 and IE seven in 2006, and those are I think, the two of the worst browsers ever made. Five years of nothing. And that stagnation created the open opening for Firefox in 2004 and then Chrome in 2008, and then Chrome went from zero to 71% market share using the same playbook Microsoft had used bundling with Android, making it a default, controlling the infrastructure underneath.

[00:07:25] Now. Why did I tell you all of that? In January, 2026, Google Ship Gemini Auto browse started shipping not shipped to all of them, to all the users. They started shipping Gemini Auto browse into Chrome. It's an AI agent built into the browser that 3 billion people already use. If that sounds familiar, it should think about what Google was facing.

[00:07:48] Chat, GPT is answering questions people used to. Google perplexity pulling search traffic. OpenAI launched its own browser Atlas. The DOJ just found Google guilty of maintaining any legal search monopoly. Google's gateway search is completely under threat for the first time ever in 20 years. So what do they do?

[00:08:10] The same thing Microsoft did when Netscape threatened windows. In 1995, gates wrote the internet tidal wave memo and redirected the entire company. Google had their own code read. Remember that in 2026, Google embedded an AI agent using, I believe, one of the smaller models, faster models, cheaper models, into the browser, and on their way to roll it out to 3 billion people who already had that browser installed.

[00:08:36] Microsoft didn't care about browser. What they cared about was protecting windows. Google doesn't care about building the best AI agent. They care about protecting search. They care about protecting their revenue. The browser. In this case is the defensive weapon and not the product. Microsoft shipped IE with every copy of Windows. Google is starting to ship an AI agent with every copy of Chrome, different decade, same reflex. Hollywood movies being remade. Just like in 1995 when Microsoft's move triggered an arms race, Google's move.

[00:09:08] We'll probably trigger one too, so lemme show you what that arms race will look like.

[00:09:12] no. Hatch,

[00:09:15] I spent weeks cataloging agentic browser landscape for a blog post on no hacks bot.com. And what struck me wasn't just the number of products, it was the speed. 'cause as of early 2026 and it's February 10th, as I'm recording this, there are over 30 agentic browsers from major companies. 18 months ago, zero, 12 months ago, maybe a handful, not 30 plus.

[00:09:38] Backed by the largest companies in tech, as I mentioned earlier. So you have. New standalone browsers built from scratch. You have ity Comet, it's free, it handles form fulfilling product comparison, all that stuff. Chat, GPT Atlas from OpenAI, subscription based with autonomous agent mode for multi-step tasks.

[00:09:57] Gens Spark, which raised a hundred million dollars at a 530 million valuation, runs on device models fellow, which lets you inspect what AI is doing before it executes there. There are more. There are a lot more. Then you have the big tech company integrations beyond Chrome. Google has something experimentally called disco.

[00:10:18] You can check it out. I think it's google disco.google.com it's a pretty fun project where they replace the URL bar with a prompt composer. Edge has copilot mode with multi-tap context awareness and tropic as CLO for Chrome as an extension on the developer side. And this is the part that that really gets me excited.

[00:10:36] There are open source frameworks like. Browser use and stagehand, cloud infrastructure from browser base, hyper browser enterprise APIs from every major AI company. Now, here's the part that matters. If you run or optimize a website, all these agents already interact and will interact with your website differently than a human does.

[00:11:01] They navigate by reading your HTML structure. They fill out forms by interpreting form labels. They click buttons based on what those buttons say, not where they're positioned on the screen or what color they are. What helps them, and I cannot say this enough times, is semantic HTML with proper button and form elements, clear descriptive labels, accessible design patterns, server rendered content.

[00:11:26] Basically the same things that help screen readers. And humans help AI agents and what breaks them? Well, if you have an aggressive capture, anything that requires hovering to load, infinite scroll , unlabeled buttons, and heavy client side rendering, if your site depends on JavaScript to load its core content, basically a large chunk of these agents will not see your content, and this traffic is.

[00:11:57] Not future traffic. It's real. Today. According to Adobe Analytics, data traffic from Gen AI browsers or AI browsers or gentech browsers to US, retail sites increased 4700% year over year in July, 2020 5, 4, 7, 0 0, and these AI directed visitors spent 32% more time on site and viewed 10% more pages than traditional visitors.

[00:12:23] I go way, way deeper into all 30 plus tools. In the blog post. I will link to it in the episode description, but for this episode, I want to zoom out because the specific products matter less than the pattern they're forming, and it raises a question worth answering. Why are 30 plus companies all making the same batch at the same time?

[00:12:46] no. Hatch,

[00:12:48] Three reasons. The first one is that the technology simply wasn't ready until late 2024. For years, AI models could generate text, summarize articles, answer questions, but reliably interacting with a webpage, clicking the right button, filling out a form, navigating a multi-step checkout that was beyond what the models could do.

[00:13:09] Then October, 2024, an anthropic releases their computer use. Agent in public beta, this is the first major proof that an AI model could actually see and control screen. November, 2024, anthropic releases MCP, the model context protocol, a standardized way for AI to connect to tools and services. January, 2025, open AI launches.

[00:13:30] Operator March, Amazon releases Novak July, hyper perplexity launches, comma. January 26th, Chrome starts shipping Gemini Auto Brow. Look at a timeline. Once one company proved it was possible, everybody else piled in the capability threshold was crossed 15, 16 months ago, in late 2024. Everything since then is nothing but land grab.

[00:13:55] The second reason search is cracking, at least the search that we used to have for 20 years. Google is aware of this and. The browser is the next position to defend yourself. But it's not just Google Search as a category is simply fragmenting chat.

[00:14:12] G PT answers questions directly. Perplexity is building an entire business on bypassing Google. The traditional model of type of query you get a list of links is. It's still there, but it's losing ground to AI systems that just give you the answer or do the task for you. Even Google, they have AI overview, they have AI mode, which by the way is great.

[00:14:32] I've been using AI mode a lot lately, and I, I kind of like it.

[00:14:36] So if search is no longer the guaranteed chokehold it used to be, where do you go from there? Well, you go one layer deeper. You build the agent that acts on the user's behalf. That's the browser. That's why OpenAI didn't stop at chat. GPT. They built Atlas. That's why Perplexity didn't stop at search. They built Comet.

[00:14:51] None of these companies are building browsers because they love browsers. They're racing to control the next gateway before somebody else controls it. The third reason. The most important one, the price has changed. The nineties browsers and the wars around them. It was about the eyeballs. It was about controlling what the people see.

[00:15:14] Ad impressions, the agent browser war. It's about wallets. An AI agent, books your flights, fills out your mortgage application, compares insurance quotes, and clicks. Buy on your groceries, controls a commerce layer that could be worth trillions of dollars. Just look at what Amazon has done in late 2025. They get this, they sued perplexity.

[00:15:36] They sent a cease and desist letter over Comet's automated shopping feature. They, they demanded that Comet declares it's an AI shopper. So why did they do that? Well, they, they're trying to protect their own intellectual property. An existential threat to Amazon's marketplace and they're trying to protect themselves.

[00:15:56] So we went from fighting over who shows you the search results to fighting over who spends money for you? Same playbook, higher stakes. But that is not the whole story because, the 2026 remake a Browser Wars is not just a remake.

[00:16:12] no. Hatch,

[00:16:15] By now the parallels between 1990s browser wars and 2020s agentic. Browser wars are obvious. We have bundling, we have pre products as weapons, we have fragmentation distribution deals. Yes, you've heard the playbook repeating for the last 10, 15 minutes, so I'm not going to re-list all the similarities.

[00:16:34] You get it. The playbook is virtually the same. What I want to talk about is where. The current version gets worse and yeah, remake, of a nineties classic can be worse. Shocking in three specific ways. The new browser war is a bigger problem and a more difficult problem than the original. The first one is the chromium trap, because Microsoft never achieved true and complete infrastructure lock-in, they bundled.

[00:17:04] Explorer, internet Explorer with windows. Yes, but Netscape and Firefox, they could still run on windows just fine. If you wanted an alternative, you could just install one and use it. A lot of people did. Google did something smarter. Chromium is open source and sure, that's great until you realize that.

[00:17:24] Perplexity Comet is built on chromium Chad, GPT. Atlas is built on chromium opera, and neon is built on chromium. Google's competitors are running on Google's rendering engine. Yes, it's open source, but it's controlled by Google. They control the web standards to the Chromium projects. So even if you switch to a rival, a gen browser, you're still on Google's foundation.

[00:17:46] They own the building, the Agentic AI features. Just the floor that's added to the building. Microsoft bundled explorer with windows and won the first browser war. Google got its competitors to build on its engine. That's not bundling. This is way, way, way deeper and more serious. Second, second thing I wanna talk about is the invisible fragmentation.

[00:18:06] So I mentioned, and I remember these vividly, best viewed in Netscape, best viewed in Internet Explorer badges. Ugh. They were so bad. You had to build for both and, and the CSS didn't work. And the jobs it, it was terrible. That was the fragmentation that you could see and that most people can't see. A user hits a page and they get a badge telling them, switch a browser, and you knew something was broken.

[00:18:28] It was annoying, but it was visible. You knew what was wrong with Thetic browsers, the fragmentation is completely, completely silent. Each agent is going to interact with websites differently based on. The training data, the system, product, whatever else. There's no standard for how an AI agent should navigate a checkout flow or parse a product page.

[00:18:49] A website, it works perfectly with Claude for Chrome, might break completely in charge a PT atlas and the user will never see the error 'cause it's completely, it's inside a black box. The agent will just quietly fail to extract the right price. It will miss a product variation or skip your website entirely and.

[00:19:07] That's scary. The user will get a bad result and they don't know why your website loses the transaction, and you don't know why. It This is an invisible tax on your business. And the third point, the ticket defaults that we have today are at a scale that the nineties could not even dream of.

[00:19:25] So remember the A OL deal, Microsoft paying 25 cents, 25 cents per user. Damn for default browser status, Google pays Apple $20 billion a year to beat a default search engine on Safari. Same play, but the price kind of went up. And it's not just about the money. The defaults have always mattered in tech, but they have never felt this.

[00:19:50] Sticky like they do now in the nineties. Switching a browser, just download an alternative over dial up. Annoying, doable. You can export your bookmarks and your set with egen browsers with, with how deeply integrated browsers are into everything we do. Switching means losing your trained ai, your saved workflows, your automations, your purchase history, your preferences.

[00:20:11] We're not at that level yet, but we are getting there. AI personalization will create locking that. Just wasn't there in the nineties. Most people will not switch. I mean, remember when chat GT four oh was replaced with chat GT five and, and people rioted? Imagine the entire browser changing personality.

[00:20:32] It's not going to be good. Most people will never switch because of this. And with agen browsers, the cost of switching is is higher than it's ever been. So Google shipping, Gemini auto browsers to every Chrome user on the planet. It makes Internet Explorer bundling with windows look tiny. By comparison, Chrome holds to this day, 65, over 65% of the global browser market. It's not just a browser for Google,. That might be the most powerful default advantage in the history of technology. The nineties browser war, it had the same playbook, but the 2026 version has a way deeper lock in. It has silent failure modes and distribution at a scale that would've been unthinkable.

[00:21:16] Not only 1995, but 10 years ago. So what does that mean for you specifically if you run, operate or optimize a website?

[00:21:25] no. Hatch,

[00:21:27] And this is the part I I keep coming back to in this episode. Netscape went from a $3 billion company to a footnote because. Its entire business dependent on being the gateway between the users and the internet. When someone else took that position, they had nothing else. Left websites. Don't hate me, but I think they face a version of that same risk.

[00:21:53] Don't hate me for saying that if AI agents can mediate every user interaction, and I really think that they will, not tomorrow, not in a month, but maybe six months, a year, three years, your website stops being a destination. It becomes invisible infrastructure, a data source that an agent queries behind the scenes.

[00:22:14] So you go from being the storefront to being a warehouse.

[00:22:18] The question every website owner needs to ask themselves is, if an AI agent is the one interacting with your website, what is the value you're providing that the agent cannot eventually cut out?

[00:22:32] And if you're thinking, well, surely the regulators will step in, look at a track record. Well, the DOJ took three years to file against Microsoft after explorer, bundling began. The settlement came three years after that, Netscape was dead. The Google Land Trust found a monopoly in August, 2024. The court rejected it in September, 2025.

[00:22:51] Some behavioral remedies were imposed, but by the time the regulators fully understand this agentic browsers. And the monopoly that I believe Google is building that's going to be more dangerous than any other monopoly they've ever had. The market will be locked up already, so if you're waiting for regulation to save you, your Netscape into 1997, you're just hoping the courts will intervene while your market share.

[00:23:14] Evaporates talking to the browsers, not to you, not not to website operators. And what do you do as a website owner with all of this? Number one, don't bet on one platform. In the nineties, developers went in on ActiveX and they were locked with Internet Explorer. Then I stagnated Firefox Rose, their sites broke.

[00:23:36] Remember those with notices that ActiveX doesn't work properly. So today, don't optimize your website for one specific agent. Built for the common layer, semantic, HTML proper, a aerial labels, structured data server side rendering. All those things work across all agents because. They're the web standards.

[00:23:55] They're not proprietary features. These agents are trained on web standards. They're not trained on errors and mistakes. The website that survived this transition will be the ones who built on fundamentals.

[00:24:08] Number two, if you haven't already. You need to start building direct relationships with your audience.

[00:24:14] Now, Netscapes fatal mistake. Wasn't technical. It was strategic because they depended entirely on being the gateway. When someone else took control of that gateway, they had nothing left. If your business depends a hundred percent on people visiting your website through a search engine or a browser and AI agent sitting between you and your user is

[00:24:36] and existential risk build those direct channels. Email lists, communities, subscriptions, accounts, things that connect you to your audience without an energetic browser in the middle. The agent can browse your website, but it should not be your only relationship with your user.

[00:24:55] Number three, well make your website worth visiting and not just worth scraping.

[00:25:01] If an AI agent can extract every bit of value from your site without a human ever seeing it, no human will ever see it. Think about what a human gets from actually visiting your page. That a data extraction call cannot deliver interactive tools, personalized experiences, community trust. The websites that thrive will have to offer something an AI agent cannot replace.

[00:25:25] Number four. Accessibility should be your agent strategy because this is the practical insight that connects everything together. The same things that make a website accessible to screen readers, make it easy to navigate. For AI agents, use proper labels, use semantic structure, keyboard navigation, clear hierarchy.

[00:25:43] Accessibility has been a compliance checkbox by a lot of organizations for decades. In the Agentic browser era, it is a competitive advantage. I guarantee you that the best optimized websites for AI agents are almost, by definition, the most accessible ones.

[00:26:00] Number five, don't wait. Act now. In the nineties, the browser war went from zero to decided in about five years.

[00:26:09] I think the agentic browser market is moving faster than that and waiting to see who wins before you adapt is the wrong play. The winner might be decided before you've even started. So here's the thing. The cost of making your website agent friendly right now is super low. Semantic, HTML, structured data, server side, rendering proper labels.

[00:26:28] These are good practices regardless of what happens with agent browsers. There's no scenario, not a single scenario where these improvements will hurt you. They will help you with traditional SEO. They'll help you with accessibility. They help you with performance, and they make you ready for whatever happens with this new browser war

[00:26:46] no. Hatch,

[00:26:49] we've seen. Browser wars before. The tactics don't change. Bundling free products to undercut competitors, proprietary features, distribution deals, all that. The difference this time is that the fight isn't about what people see. It's about what the agents, AI agents do on their behalf. The stakes have moved from attention to transaction, from ad impressions to purchases, and your website is sitting in the middle of this war.

[00:27:16] You didn't start it, but you can't set it out however. You do get to decide whether you're ready for it. So go and read the full agentic browser breakdown on the blog. It's no hpo.com/blog/agentic-browser-landscape-twenty 26. The link is in the description. It will cover all 30 plus tools in detail. And please do yourself a favor, pick one of these agent browsers and try to use your website through it.

[00:27:44] See what works. See what breaks. You'll learn more in 10 minutes of testing than any article or podcast episode can ever teach you. That's it for today. I'm sunny and this is no hacks.

[00:27:56] ​