HomeInsights › More Than Half The Internet Is Now Bots
Industry Data · The Agentic Web

More than half the internet is now bots. Here's what that means for your AI visibility.

By Neil Harte, Founder · Genivista · July 2026 · 8 min read

The short answer

  • Cloudflare reports that agent traffic crossed 50% of all internet traffic for the first time this year — more than half of what hits a website is now a machine, not a person.
  • AI training now accounts for 52% of crawler requests, up from 22% just over a year ago, and mixed-use crawlers that blend search, training and live agent retrieval make up over 36% of activity.
  • For every hour spent researching online, only 15 minutes is now spent on the open web — the rest is a consolidated AI answer.
  • The effect is not limited to media and publishing. Cloudflare's own data shows retail, software, IT and finance seeing the same pattern, with some heavily crawled categories losing up to 40% of human traffic in under a year.
  • None of this tells you what AI is actually saying about your business. That requires looking directly at the answers, not at the traffic.

Cloudflare published a report this month with a number worth sitting with: more than half of all internet traffic is now non-human. Cloudflare sits behind more than 20% of the web, so this is not a small sample — it's one of the most complete views available of what is actually happening to web traffic as AI adoption accelerates. For a B2B business trying to work out whether AI visibility is a real commercial issue or an inflated worry, this report is useful because the numbers are independent, measured, and considerably starker than most people assume.

What the report actually found

Cloudflare's Content Independence Day report, published 1 July 2026, marks one year since the company changed its default settings to block AI training crawlers unless a site owner opted in. The anniversary report is a data release rather than an opinion piece, and three figures in it stand out.

50%+
of all internet traffic is now agent traffic — non-human — for the first time on record
52%
of crawler requests are now for AI training, up from 22% in Spring 2025
15 min
of every hour spent researching online is now spent on the open web — the rest is a generated answer

Cloudflare also reports that generative AI adoption has moved faster than any previous technology shift — more than 30% of humanity, some 2.5 billion people, now use generative AI regularly, reached in roughly 3.5 years, more than twice the speed of smartphone adoption. And mixed-use crawlers — bots that blend search indexing, AI training and live agent retrieval into a single, undeclared pass — now account for more than 36% of crawler activity, which makes it genuinely difficult for a site owner to know what any given crawl was actually for.

Why this is a different kind of statistic

Traffic reports usually measure something you can act on directly: more visitors, fewer visitors, where they came from. This one is different, because it measures a shift in what "visiting" even means. When Cloudflare says only 15 minutes of every hour spent researching online happens on the open web, it is describing a buyer who used to click through five or six sites to compare options, and now types one question into an AI tool and gets a single, consolidated answer — often without ever seeing a website at all.

That is precisely the mechanism Genivista exists to diagnose. If a buyer's entire research process can now happen inside an AI answer, then the content of that answer — whether it names you, describes you accurately, and recommends you ahead of competitors — matters as much as your website itself once did. Cloudflare's data is the infrastructure-level evidence for a shift Genivista sees at the account level, one company at a time.

Half the internet is now machines. The question that actually matters is not how much traffic is a bot — it's what the bot tells the human on the other end.

This isn't just a publishing problem

Cloudflare is explicit that the pattern first visible in news and media has become "industry-agnostic," with retail, software, IT and finance now seeing the same dynamics. Some of the most heavily crawled categories in Cloudflare's data have seen human traffic decline by as much as 40% in under a year — and Cloudflare notes that many publishers are now preparing for what they call "Google Zero," a scenario where little to no traffic arrives from search referrals at all.

For a B2B company, the relevant translation isn't "our traffic will drop 40%." It's this: if the category conversation is increasingly happening inside AI systems rather than on search results pages, then whatever those systems currently say about you — accurate or not, complete or not, ahead of or behind your competitors — is quietly shaping deals you will never see influenced. That's the argument made in our related piece on the deals you can't see you're losing.

The part that should concern site owners most

The detail in Cloudflare's report most relevant to AI visibility, specifically, is the rise of mixed-use crawlers — now over a third of all crawler activity. A mixed-use bot combines search indexing, AI training and live agent retrieval into one undeclared visit, so a business genuinely cannot tell whether a given crawl helped it appear in search, trained a model's long-term "memory" of the business, or answered a live buyer question in real time. Cloudflare singles out Google's crawler as the largest example of this, noting Google now holds roughly 2x more access than other leading AI companies precisely because its mixed-use bot makes it hard to participate in search without also feeding its AI systems.

The practical consequence: a business can no longer assume that "good SEO" and "good AI visibility" are the same investment, or even that they are being measured by the same crawl. Our piece on why AI search and Google search are different things covers this distinction in more depth.

What this means for what you do next

Cloudflare's report is about infrastructure and economics — who crawls, how much, and who gets paid for it. It does not, and isn't designed to, tell an individual business what AI systems currently say when a buyer asks about its category. That is a different question, and it can only be answered by asking AI the same questions a buyer would and looking, unedited, at what comes back.

That's the starting point of Genivista's AI Visibility Initial Review — a live, unedited look at what ChatGPT, Gemini and Perplexity currently say about your business, at no charge, before you decide whether a deeper independent investigation is justified.

Source: Cloudflare, "Content Independence Day, one year on: building the business model for the agentic Internet," published 1 July 2026, compiled from Cloudflare Radar and Cloudflare's 2026 Investor Day presentation.

Frequently asked questions

Is more than half of internet traffic really non-human?
According to Cloudflare's July 2026 report, yes — agent traffic crossed 50% of all internet traffic for the first time this year. Cloudflare sits behind more than 20% of the web, including 36% of the world's most-visited sites, giving it one of the largest vantage points available for this kind of measurement.
What is a mixed-use crawler, and why does it matter?
A mixed-use crawler blends search indexing, AI training and live agent retrieval into a single bot, so a site owner cannot tell which purpose a given visit served. Cloudflare reports mixed-use crawlers now account for more than 36% of crawler activity, which makes it harder for businesses to know whether they are appearing in search, being trained on, or being read live to answer a buyer's question.
Does this affect B2B companies, or just publishers and media?
Cloudflare's report is explicit that the effect is industry-agnostic: retail, software, IT and finance are all seeing the same pattern as media did first. Any B2B company whose buyers research online before making contact is affected, because the research increasingly happens inside an AI answer rather than on a website visit.
If AI is reading my site more than humans are, does that mean more visibility?
Not necessarily. Being crawled is not the same as being cited, and being cited is not the same as being recommended. A site can be crawled heavily by AI training bots while still being absent, misdescribed or displaced by a competitor in the answers those systems generate. Crawl volume is a data point, not a verdict on visibility.
What should a business do in response to this data?
Start by finding out what AI systems currently say when a buyer asks about your category — not by reading crawler logs, but by asking the same questions a buyer would. That is what an AI Visibility Initial Review does: a live, unedited look at whether you are being described accurately, found at all, and recommended ahead of or behind your competitors.

See what AI says about your business

A no-charge 30-minute AI Visibility Initial Review. Live on screen. Unedited. No preparation needed — just your company name.

Book an Initial Review →