Something strange is happening to the internet. Websites can still publish thousands of pages, search engines can still return millions of results, and information can still travel around the world in seconds. Yet the people who created much of that information are increasingly finding that fewer humans are actually showing up to read it.

Gisele Navarro has watched that change happen from inside her own business. Her website, HouseFresh, spent years testing air purifiers in real homes and publishing independent reviews. Then its Google traffic collapsed, while automated systems continued arriving at the site in enormous numbers.

HouseFresh Lost Almost All Of Its Google Audience

HouseFresh was built around a straightforward idea: test products properly before telling people whether they are worth buying. The team purchased air purifiers, ran them in real homes, recorded their performance, and published detailed reviews based on those tests. For a small publisher, the work required time and money, but search traffic made that investment possible.

That relationship changed dramatically. Navarro wrote that HouseFresh had gone “from welcoming 4,000 people from Google Search each day to just receiving 200,” a decline of roughly 91%. She said the consequences reached the company’s income, its ability to sustain its team, and its plans for the future.

Navarro eventually stopped believing that search traffic could provide the stability it once offered. “Whatever stability we had from search traffic, from Google, let’s not count on that at all,” she wrote.

Her experience offers a glimpse into a much larger shift taking place across the web.

More Than Half Of Web Traffic Is Now Automated

The exact percentage depends on what is being measured, but the direction is hard to ignore. Imperva’s 2026 Bad Bot Report found that automated traffic represented more than 53% of all web traffic during 2025, up from 51% the previous year.

Cloudflare measured a different category of traffic in its 2025 review. Human traffic accounted for 47% of HTML requests, while non-AI bots represented 44%. AI bots averaged another 4.2% during the year.

Those figures should not be treated as one definitive measurement of the entire internet. The companies use different methodologies and examine different kinds of traffic. That is why estimates ranging from roughly one-third to more than half can appear without necessarily contradicting each other.

Some automated traffic is useful. Search crawlers, security systems, accessibility tools, price checkers, and monitoring services all perform legitimate jobs. The issue is the changing balance between automated consumption and human attention.

The Dead Internet Theory Sounds Different Now

For years, the Dead Internet Theory lived mostly in online conspiracy culture. Its basic idea was that the internet had become increasingly dominated by automated accounts, machine-generated content, and artificial activity while genuine human participation declined.

The available data does not prove the theory in its broadest form. It does make the basic image easier to understand.

Luis Prada, who wrote about the earlier estimate that around one-third of internet traffic came from bots, described the internet as “the hollowed-out shell of a Pottery Barn.” His comparison captures the strange possibility of a digital environment that remains packed with information while containing fewer genuine interactions between the people who create it and the people who consume it.

That distinction matters because the web was shaped by human curiosity. Someone wrote a review because another person might need it. Someone documented a hobby because another enthusiast might discover it. Someone published specialist knowledge because another person might eventually find it through a link.

The value was created through connection. Now, more of that same material can be collected by machines, processed elsewhere, and returned to users as an answer without the original page receiving a visit.

The Old Internet Relied On A Simple Exchange

Navarro described the relationship between Google and publishers as a kind of social contract. Publishers allowed Google to crawl their work, while Google sent people back to the sites that had created it.

“Google created a social contract with the online publishing industry: publishers provide content for Google to crawl and, in exchange, Google sends valuable clicks back to publishers,” Navarro wrote.

That exchange supported a huge independent publishing ecosystem. It gave small websites a reason to invest in original reporting, product testing, photography, tutorials, reviews, and specialist knowledge.

The system was never perfect. Search rankings could change, algorithms could punish publishers, and large companies had advantages that independent websites did not. Still, the basic mechanism was understandable: create something useful, attract readers, and use that audience to keep producing useful work.

AI systems complicate that model because they can consume information without necessarily returning comparable attention. The publisher supplies the material, the machine processes it, and the human reader may receive the information somewhere else.

That creates a very different economic relationship.

One AI Crawler Made The Imbalance Hard To Ignore

Cloudflare’s measurements provide one of the clearest examples. During a week in June 2025, the company examined how frequently AI crawlers requested HTML pages compared with the number of referrals they sent back to publishers.

For Anthropic’s Claude, Cloudflare measured nearly 71,000 HTML page requests for every HTML page referral.

The number is difficult to imagine in ordinary publishing terms. A traditional reader visits an article, reads it, perhaps clicks another page, and may return later. A crawler can request information at a completely different scale because it is not behaving like a person browsing a website.

That difference becomes crucial for publishers whose businesses depend on human attention.

A HouseFresh review exists because people spent time buying and testing an air purifier. If a machine can absorb that review and use its information to answer thousands of questions, the work may continue influencing people without those people ever visiting HouseFresh.

The information travels. The connection to its creator may not.

What Happens When Original Work Stops Paying?

This is where the story becomes larger than website traffic. Independent publishing depends on people being able to justify the time and expense required to create something original.

Consider the work involved in a detailed product review. Someone has to purchase the product, set up the test, operate it, collect results, compare those results with competing products, write the findings, edit the article, and maintain it when information changes.

If original publishers repeatedly lose their audiences while automated systems continue consuming their work, the financial incentive to perform that work can weaken.

The consequences may not appear immediately. A website does not vanish the moment traffic falls, and a writer can continue publishing even when revenue declines. A small team can keep going through determination and sacrifice.

But there is a limit to how long any business can operate without enough people supporting it.

Cloudflare’s researchers have pointed toward that concern, stating that the changing traffic ratios “directly impact the viability of content publication on the Internet.”

That is ultimately a question about people, not bots.

The Internet Still Needs Someone To Do The Work

AI can make finding information faster. It can summarize complicated subjects, answer questions, compare ideas, and help people access knowledge they might otherwise struggle to understand.

But the usefulness of those systems still depends on reliable information to consume. Someone has to conduct the experiment before there can be a summary of it. Someone has to test the product before an AI can describe its performance.

That is the tension underneath the growing bot numbers. Machines can become extraordinarily efficient at consuming human knowledge, but the knowledge itself still begins with human effort.

HouseFresh offers a particularly clear example because its work depends on firsthand testing. Navarro has continued publishing despite the collapse in search traffic, but her experience shows how difficult that decision becomes when the traditional route to readers stops working.

The internet may become increasingly automated without becoming useless. It may even become more convenient. But convenience cannot replace the people who investigate, test, document, question, and create the information that makes the whole system worth using.

The future of the web will depend on whether those people can continue to find an audience. If they cannot, we could end up with an internet that answers more questions while giving fewer creators a reason to keep asking them.

Loading...