AI browsers are about to become your biggest problem
When AI agents – computer programs that use AI to do tasks and reach goals by themselves for a person or another system – become the default way people navigate the internet, the value chain will be rerouted, says Paul Armstrong
very business that depends on digital reach is about to face a visibility crisis. Customers are no longer discovering brands through search the way they used to. Interfaces are being rewritten, and in the space of a few short product cycles, the browser, the search engine, and even the input itself are all mutating. The shift won’t just be cosmetic or technical, it will be existential. When AI agents – computer programs that use AI to do tasks and reach goals by themselves for a person or another system – become the default way people navigate the internet, businesses will no longer be talking to customers directly. These agents will be talking to machines trained to act on the customer’s behalf. The value chain of persuasion, interaction and conversion is about to be rerouted.
Browsers are not static windows to the web anymore, they are becoming agents with autonomy. OpenAI is building a browser designed to execute tasks like booking appointments or buying products without ever surfacing the website behind the transaction. Perplexity’s Comet and The Browser Company’s Dia already automate multi-step actions, from reading and summarising to filling forms and answering emails. Some of these features are clunky, and most are unfinished, but the intent is obvious. Speed of page load is irrelevant, and speed of completion becomes everything.
Google is, of course, sweating. Not just because of antitrust lawsuits, but also because Google’s dominance via Chrome and Search is being eroded at the edges. Gemini integration into its search engine reflects a defensive pivot. OpenAI’s push to build a productivity suite to rival Google Workspace suggests it is not only looking to replace how people browse, but also how they work. If (when?) that happens, the very act of looking for information or creating a document will happen inside a conversational interface that removes the need for websites, navigation menus, or traditional discovery. Think about that for a while, and ask where you fit in.
New customer journeys
Customer journeys are being remade. Thinking ahead, pages, funnels and brand campaigns become optional if an agent is making decisions and pulling information in the background. Businesses will find that their best-designed UX never reached the end user. Instead, AI will extract data, summarise it, compare it and make a recommendation. Optimisation will shift away from people and toward machine intermediaries that can execute actions on behalf of those people.
Web analytics as we know it will stop being useful. Referral traffic won’t tell you where a user came from because the user may never have seen your page at all. An agent might lift information, relay it, and make a decision before the site loads. Brands that rely on behavioural data to optimise experience will be navigating blind. Attribution will break. Traditional marketing levers will misfire, and assumptions about who visited, what they saw and what they clicked will become unreliable. Bad news if your name is Google.
User interfaces will no longer be the last mile, they will be bypassed. Many businesses have built operational complexity around front-end polish. A model that stops working when the conversation begins and ends with a chatbot. In sectors like retail, insurance, travel, and financial services, the customer interface is about to disappear. Is that a bad thing? Yes, and no.
Agent browsers won’t solve everything, and the current experience is far from seamless. Perplexity reportedly completes only around half of complex tasks effectively. Dia’s promise of multi-tab synthesis delivers value for research, but fails in high-context situations where nuance matters. Errors accumulate fast when the model makes assumptions, especially in specialist or regulated domains. For now, the human-in-the-loop is still essential. But the trajectory is clear.
Expect failures and rapid evolution because these products are data-hungry by design. Every query trains the next generation of agents. Browsers become insight engines, and don’t just surface content, they observe behaviour, extract preferences, monitor prompts, and build individualised models of intent. Businesses not actively tracking how their sites appear to agents will be operating in the dark.
Make your content for AI
Companies must prepare for this shift now. Structure content for agents, rather than block them. Use markup, taxonomies, and embedded data structures that allow machines to ingest information without visual cues. Build APIs not just for integration but for ingestion. Simplify product descriptions and service flows so that AI can confidently parse and act on them. Consider that an AI tool will never guess, it has to default to what it understands. Ambiguity is only going to cost you revenue. Much of this is commonly thought of as the roll over and take it approach. But there are other approaches.
Consider that an AI tool will never guess, it has to default to what it understands. Ambiguity is only going to cost you revenue
Cloudflare, an IT services company, may be the white knight in this scenario. The company just announced a way of stopping crawlers, and shows how platforms are already responding to this shift. Businesses can now explicitly manage AI bot access to their sites through the updated “AI bot” specification in robots.txt and firewall configurations. A level of precision that offers a way to safeguard performance, protect sensitive content, and control scraping by intelligent agents before those agents begin to rewrite your customer journeys. Expect a work around soon, but you see the intent and potential for disruption.
Browser-based productivity is also going to fracture. OpenAI’s moves to replace Google Workspace mean collaborative workflows are being dragged into the chat interface. Docs, sheets, presentations, all surrounded by assistant agents capable of search, summarisation and formatting. If knowledge work happens inside an AI sandbox, legacy SaaS players lose their gatekeeper role. Internal business tools may need to adapt quickly to survive inside agent ecosystems so start planning your IT department’s resources, and maybe plan on extending their Christmas party budget while you’re at it.
New monetisation models are also likely to emerge. Browsers that automate work are shifting toward paid tiers. Comet is already charging $200/month for early access. Subscription will replace ad-based access in the near term. What all this means for the teeny tiny +$250 billion advertising industry remains unclear. How does a brand advertise to an agent that has already decided what the customer wants? What does a call-to-action look like in a world where no human ever sees the ad? I haven’t seen a lot of those insights amidst the Cannes Lions hot takes clogging up my LinkedIn feed that don’t seem to want to die. If your agency isn’t having these chats with you now, ask yourself why.
Now some reality. Expect some AI browsers to collapse under their own hype. We’ve still got a long way to go until we really see where AI is going to leave the world, how we’ll input into it, and whether the hardware concept from Sam Altman and Jony Ive will be the second coming. Prediction: it won’t. Many early agent efforts will stall on performance, hallucinations, or the cost of scale. But the presence of these attempts signals a shift in incentives. Companies are not just chasing attention, they are chasing control over intent. Whoever owns the agent owns the interaction.
There is no version of the next five years where traditional browser marketing remains dominant. Search won’t vanish overnight, Chrome won’t disappear, but their dominance will soften. Brands that start adjusting now can use the transition period to build resilience. Structure content for agents, not protecting dodgy defences. Design discovery around questions, not keywords, and rethink your funnels for a world where decisions happen inside context windows, not landing pages. A massive shift to get your head around, but do it early enough and you’ll see the fruits of your labour and not be left trying to hold back the waves with a toothpick.
The businesses that wait for this shift to be finished will likely be too late. Disruption like this isn’t seen every day and has come fast which is why it falls more in the seismic bucket. Browsers are fast becoming decision engines where every click not made is a signal, and every interaction handled by an agent is a lost chance to persuade. Businesses have a window open (pun intended) now to rethink how people find, evaluate and act. Don’t act and you might just become the ghost in the machine.
Paul Armstrong is founder of TBD Group and author of Disruptive Technologies