AI Browsers Are Convenient—but Could They Kill the Open Web?
Matthew Leo · Published July 22, 2026 · AI
AI Browsers Are Convenient. They May Also Kill the Web.
The web browser used to have one simple job.
You told it where you wanted to go, and it took you there.
AI browsers want to remove the journey.
Instead of opening five websites, reading them and deciding what matters, you can ask one question and receive a neat answer built from all five. You can ask the browser to compare products, summarize an article, search your open tabs, read your email or complete a task while you do something else.
That is genuinely useful.
It is also a serious threat to the websites doing the work underneath it.
If an AI browser reads a review, extracts the important points and gives you the conclusion, why would you visit the review?
If it studies ten recipes and creates one for you, what happens to the food blogs?
If it summarizes reporting from several newspapers, who pays the journalists next year?
AI browsers may not literally destroy the internet. Amazon, Netflix, banks and government services will survive without blog traffic.
What they could destroy is the part of the web created by people who depend on visits to keep producing useful information.
That includes independent publications, local newsrooms, reviewers, forums, tutorials and small websites like this one.
What Is an AI Browser?
An AI browser combines ordinary web browsing with an assistant that can understand pages, search across information and sometimes take action for the user.
Perplexity’s Comet is built on Chromium, the same open-source foundation used by Chrome and several other browsers. It includes AI search, page summaries, questions about open websites, commands for managing tabs and connections to services such as Gmail and Calendar. It can also use browsing history to retrieve information relevant to the user. (perplexity.ai)
Dia, from The Browser Company, can search across tabs and connected workplace tools such as Slack, Notion, Google Workspace and calendars. It can turn scattered information into reports, prepare meeting material and build presentations from the user’s context. (diabrowser.com)
Google is moving in the same direction without asking people to adopt a completely new browser. AI Mode is now integrated more deeply into Chrome and can use selected tabs as context for questions. Google Search can also answer a question through AI Overviews before a user opens any result. (blog.google)
The difference between these products and a traditional search engine is important.
A search engine gives you a list of places that may contain the answer.
An AI browser tries to become the place where the answer is consumed.
The Convenience Is Not Fake
There is no point pretending the old way of browsing was perfect.
Opening twelve tabs to compare laptops is tiring. Half the websites are covered in pop-ups, autoplaying videos, newsletter requests and advertisements that move the article while you are trying to read it.
Recipe websites sometimes make you scroll through a family history before revealing how much flour you need.
News websites repeat the same basic information in slightly different words.
AI browsers can remove much of that frustration.
A good summary can save time. Asking questions about several documents at once can be far better than manually searching each one. An agent that completes a repetitive online form may spare someone twenty minutes of administrative misery.
For people with disabilities, language barriers or limited time, these features may make online information more accessible.
The technology has value because the problem it addresses is real.
Many websites became unpleasant because publishers were trying to squeeze enough money from each visit to survive.
Now AI browsers are offering to solve the unpleasant experience by removing the visit altogether.
That is a brilliant solution for the browser.
It is less brilliant for the website.
The Web Had an Unwritten Deal
Traditional search engines also copied and indexed website content, but there was an understood exchange.
Google’s crawler visited a website.
Google placed the page in search results.
A person clicked the result and visited the website.
The publisher received attention, advertising revenue, subscriptions, sales or at least the chance to build a relationship with the reader.
The arrangement was never perfectly equal. Google became extraordinarily powerful by organizing information created by everyone else.
Still, traffic moved in both directions.
AI changes the exchange.
A company can crawl thousands of pages, use them to generate an answer and return very few readers to the original sites.
Cloudflare measured this difference in June 2025. It estimated that Google crawled websites about 14 times for every visitor it referred. OpenAI’s ratio was approximately 1,700 crawls for every referral, while Anthropic’s was about 73,000 to one. Cloudflare warned that this broke the old relationship in which websites tolerated crawling because search platforms sent people back. (cloudflare.com)
Those figures are estimates, and traffic from some apps may not always be identified correctly.
The scale of the gap is still difficult to ignore.
AI companies want to read the web at industrial speed.
They are much less enthusiastic about returning the reader.
People Click Less When the Answer Is Already There
The concern is no longer theoretical.
Pew Research Center studied browsing activity from 900 American adults in March 2025. When a Google results page contained an AI summary, users clicked a traditional search result in 8% of visits. Without an AI summary, they clicked a result 15% of the time.
Only 1% of visits to a page with an AI summary led to a click on one of the sources cited inside that summary. Users were also more likely to end their browsing session after seeing an AI answer. (pewresearch.org)
A separate 2026 study examined more than 161,000 matched Wikipedia articles across languages. It estimated that exposure to Google’s AI Overviews reduced daily traffic to the English versions by about 15%. The decline was greater for topics where a short summary could easily satisfy the user’s question. (arxiv.org)
That final detail matters.
The more completely an AI answers the question, the less reason the user has to visit the source.
AI companies often present citations as the solution.
But a citation nobody clicks is not compensation.
It is a small sign saying where the value came from after the value has already been taken.
Google Says AI Can Still Help Publishers
Google rejects the idea that AI search is simply destroying web traffic.
The company says AI Overviews and AI Mode encourage people to ask more complicated questions and visit a wider variety of websites. It argues that the users who do click through are often more engaged and spend more time on the destination page. Google has added more inline links, website previews, preferred-source tools and subscription labels to encourage visits. (blog.google)
Google reported in June 2026 that AI Overviews had more than 2.5 billion monthly active users and that AI Mode had passed one billion. It has also started testing controls allowing some website owners to opt out of having their content used to ground generative Search responses. Sites that opt out would also lose any traffic or visibility from those AI features. (blog.google)
Google’s argument deserves to be heard.
A smaller number of highly interested visitors may sometimes be more valuable than a larger number of people who leave immediately.
AI may also create new searches that would never have happened before.
But “better clicks” cannot solve the problem if the total number of clicks becomes too small.
A shop would not survive because its remaining customers stayed longer after 90% of the neighbourhood stopped entering.
Publishers need enough traffic to pay writers, editors, photographers, developers and hosting costs.
Quality matters.
Quantity still pays bills.
AI Browsers Make the Problem Bigger
An AI Overview appears on a search page.
An AI browser can follow the user everywhere.
It may summarize the article currently open, compare it with other pages, remember the result and use the information later without returning to the source.
Comet includes a built-in button specifically for condensing lengthy content. Dia promotes the ability to gather information scattered across websites and connected tools, then turn it into a finished report. (perplexity.ai diabrowser.com)
From the user’s perspective, this is excellent.
From a publisher’s perspective, the browser is becoming a well-dressed middleman who enters the shop, reads every label and then sells directions to the customer outside.
The browser no longer needs to display the website as the final product.
The website becomes raw material used to produce something else.
That shift changes who receives the reader’s attention, who controls the experience and who earns money from it.
The Biggest Websites Will Survive
The web itself is not about to disappear.
People will still visit social networks, shopping platforms, streaming services and online banking.
Large publications may negotiate licensing agreements with AI companies. Some can build subscriptions, apps, events and direct audiences strong enough to survive declining search traffic.
The sites most at risk are the ones with the least bargaining power.
A local newspaper cannot negotiate like Reddit.
An independent reviewer cannot demand a multimillion-dollar licensing agreement.
A small travel blog cannot stop an AI from summarizing its guide without also risking invisibility.
These sites often depend on a simple process:
Someone searches for information.
The website appears.
The person visits.
The publisher earns a fraction of a cent from an advertisement, recommends a product or gains a subscriber.
Remove the visit and the economics collapse.
The information may continue appearing in AI answers for a while because it already exists.
The problem arrives later, when fewer people can afford to create the next article.
The Web Could Become Smaller and More Boring
AI systems need information.
They do not automatically create the reporting, testing and lived experience that make information valuable.
An AI can summarize a phone review.
It did not buy the phone, use it for a month, test the battery, take photographs or discover that the hinge becomes annoying after three weeks.
It can summarize a restaurant guide.
It did not visit the restaurants.
It can explain a new law.
It did not attend the hearing, interview experts or read hundreds of pages of evidence unless someone first placed that work online.
If the people doing original work lose their traffic and income, fewer of them will continue.
The web then fills with content made by large organizations, hobbyists who can afford to work unpaid and AI systems rewriting material produced earlier.
The result may be an internet with more answers and less knowledge.
Everything will be clearly summarized.
Less of it will be worth summarizing.
Independent Voices May Be Hit First
Search traffic has plenty of problems.
It rewards clickbait, repetitive articles and pages designed around keywords instead of readers.
But it also gives small publishers a chance to compete.
A useful article from an unknown writer can rank beside one from a giant publication. A niche expert can reach people without owning a television network or paying for a large advertising campaign.
AI answers may weaken that opportunity.
Research into AI search has found that citations can become concentrated among a smaller group of sources. A 2026 study covering thousands of queries across hundreds of countries also found that AI search exposed users to fewer long-tail sources and less variety than traditional search. (arxiv.org)
That means the browser may give you an answer from “the web” while quietly choosing which tiny section of the web counts.
Readers see a smooth paragraph.
They do not see the dozens of sources that were ignored, the judgment used to rank them or the commercial relationships that may influence the final result.
A traditional search page at least exposes some of that choice.
An AI answer hides it inside the machine.
The Browser Also Learns More About You
AI browsers become more useful when they know what you are doing.
Comet can use browsing history for personal search and connect to email and calendar services. Dia can work across tabs, Slack, Notion, Google Workspace and other tools. (perplexity.ai diabrowser.com)
That context allows the browser to answer questions that a normal search engine cannot.
It also places more sensitive information inside one system.
The browser already sees where you go online. Add AI memory, email access, work documents and permission to take action, and it may understand a large portion of your personal and professional life.
Dia says users control memory and connected tools, synced browser data is end-to-end encrypted and personal data is not sold or used to build advertising profiles. Its privacy policy also says some question-and-answer content may be used to improve the product by default unless sharing is disabled. (diabrowser.com)
Those controls matter.
Users still need to read them.
The phrase “AI browser” sounds like a smarter version of Chrome. In reality, it may be closer to giving an assistant a desk inside your digital life.
Letting the Browser Act Creates New Security Risks
Summarizing a page is one thing.
Allowing an AI to send an email, complete a purchase or interact with an account is much more serious.
Researchers have shown that AI browsing agents can be manipulated through prompt injection. A malicious instruction may be hidden inside a webpage, document or other content the agent reads. The agent can mistake that content for a command and take an action the user never requested.
The BrowseSafe research project tested realistic attacks embedded in web pages and found that browser agents need several layers of protection rather than relying only on the model to notice suspicious instructions. (arxiv.org)
This does not mean AI browsers are unusable.
It means convenience should come with limits.
An AI may be trusted to summarize an article.
It should not quietly move money, reveal private information or send messages because a webpage told it to.
The more capable the browser becomes, the more damaging a mistake can be.
Blocking AI Is Not a Complete Answer
Website owners can try to block AI crawlers.
Cloudflare has introduced tools allowing sites to allow, block or charge certain AI crawlers for access. Its Pay Per Crawl system uses the internet’s “402 Payment Required” response to tell a crawler that access has a price. (cloudflare.com)
That is an interesting attempt to build a fairer exchange.
It also reveals the publisher’s impossible choice.
Allow the AI to crawl your work and risk receiving little traffic.
Block it and disappear from a growing part of how people find information.
Charge for access and hope AI companies agree to pay.
Large publishers may have enough power to make those choices meaningful.
Small publishers may simply become less visible.
An opt-out is not real control when refusing participation means becoming invisible to the audience.
What a Fair AI Browser Should Do
AI browsers do not need to disappear.
They need a business model that does not depend on slowly starving their sources.
A fairer system would:
Show sources prominently, not as tiny numbered footnotes
Encourage readers to open the original page
Avoid reproducing so much that visiting becomes pointless
Let publishers set clear rules for summaries, training and agent access
Pay creators when their work provides substantial value
Distinguish original reporting from sites merely repeating it
Tell users why particular sources were selected
Allow websites to opt out of summaries without disappearing from normal search
Separate paid placements from editorial source selection
Give users control over browser memory and connected accounts
Require confirmation before agents take sensitive actions
None of these ideas is technically impossible.
They are commercially inconvenient.
The AI company benefits when the user stays inside its product.
The publisher benefits when the user leaves.
That conflict cannot be solved with a prettier citation box.
Publishers Also Need to Change
The old web model was already weak before AI arrived.
Publishers became too dependent on Google, social media and programmatic advertising. Many built audiences they did not truly own.
AI browsers are making that dependency more dangerous.
Websites need stronger direct relationships through newsletters, memberships, communities, podcasts, events and distinctive reporting readers actively seek out.
They also need to stop producing disposable articles that an AI can replace with three sentences.
A page called “What Time Does Walmart Close?” is vulnerable because the answer is the entire product.
Original reporting, useful tools, personal experience and strong analysis are harder to replace.
That does not make AI extraction fair.
It does mean publishers must give readers a reason to care where the answer came from.
For Mapletechie, that means an article cannot merely repeat specifications or summarize a press release. The value has to be in the judgment, the explanation and the voice.
AI can repeat the facts.
The opinion has to earn the visit.
Convenience Has a Cost
AI browsers will continue growing because they solve a real problem.
People are overwhelmed by information. Websites are cluttered. Search results are crowded with ads, copied articles and pages written mainly to rank.
Asking one clear question and receiving one clear answer feels better.
The danger is that we judge the system only by how convenient it feels today.
The web is not an endless natural resource.
People create it.
They investigate, test, explain, photograph, review, argue and publish. Some do it for fun. Many need an income to continue.
AI browsers can hide that labour by making every answer appear instantly, as though the information simply existed.
It did not.
Someone made it.
A browser that uses the web without returning attention or money is not organizing information. It is consuming the system that feeds it.
AI browsers may make the internet easier to use.
But if they succeed by making websites unnecessary, we may eventually discover that the web they summarized so efficiently is no longer being built.