How Much of the Internet Is Bots or AI in 2026? The Best Current Estimate
Recent data suggests machines now generate roughly three-fifths of web page requests. AI is part of that shift, but it is still much smaller than total bot traffic.
The best current answer is that bots appear to generate around 60% of web page requests, while humans account for roughly 40%. Cloudflare Radar measured worldwide HTML requests at 61% bot and 39% human on August 23–24, 2026. That is newer than Imperva’s full-year 2025 figure of more than 53% automated web traffic.
AI is only part of that automation. There is no authoritative global figure for “how much of the internet is AI,” but the available data suggests AI-specific and dual-purpose AI/search crawlers are still a minority of web requests.
Worldwide HTML page requests classified as bots by Cloudflare on August 23–24, 2026.
Automated web traffic across Imperva’s full-year 2025 dataset.
A cautious 2026 working range for identifiable AI-oriented and dual-purpose HTML crawling.
ESTIMATE, NOT A GLOBAL MEASUREMENTWhy around 60% is a better current bot estimate
Imperva’s 2026 Bad Bot Report is still the strongest full-year global benchmark. It says automated traffic rose from 51% in 2024 to more than 53% in 2025, leaving humans at about 47%.
But we now have a more recent public snapshot. Cloudflare Radar measured worldwide requests for HTML pages at 61% bot and 39% human over August 23–24, 2026. Cloudflare’s figure covers HTML page requests on its network rather than every byte transferred across the internet, so it should not simply replace Imperva’s annual result. It does, however, show that the majority-bot trend has continued.
Cloudflare Radar worldwide HTML request classification for August 23–24, 2026.
Imperva measured automated traffic above 53%. This is useful as a broad annual benchmark because it smooths out daily fluctuations.
Cloudflare measured bots at 61% of worldwide HTML page requests over a recent one-day window.
For September 2026, “around three-fifths of web page requests are automated” is more defensible than continuing to headline the older 53% figure as if it were current.
Different providers classify bots differently and measure different parts of the web. A range is more honest than claiming there is one exact global percentage.
This range is an interpretation of recent large-network measurements, not a single published global statistic.This still does not mean 60% of every internet byte is bots
The phrase “internet traffic” can hide several different measurements. Cloudflare’s 61% figure counts requests for HTML pages. Imperva measures web and application traffic seen through its security infrastructure. Neither number is a measurement of every video stream, game download, private cloud link, VPN tunnel or backbone packet in the world.
A bot can make thousands of tiny page requests while one human downloads a multi-gigabyte game. The bot may dominate the request count while the human dominates the number of bytes transferred.
The accurate claim is that bots now appear to generate the majority of web page requests. Saying “60% of all internet bandwidth is bots” would go beyond what the current evidence shows.
How much of that traffic is actually AI?
This is much harder to measure. “AI traffic” can mean AI crawlers visiting websites, autonomous agents making requests, model inference, API calls or huge transfers between data centers. No company currently publishes a single global percentage covering all of those categories.
The cleanest public baseline comes from Cloudflare’s 2025 analysis of HTML requests. It found identifiable AI bots averaged 4.2% of HTML requests during the year. Cloudflare tracked Googlebot separately at 4.5% because Googlebot is dual-purpose: it supports conventional search indexing as well as AI-related uses.
Average share of HTML requests Cloudflare attributed to identifiable AI bots.
Googlebot’s average HTML request share. It cannot be treated as purely AI traffic because it also serves traditional search.
Akamai’s September 2026 bot controls now split AI automation into three categories: training crawlers, AI search crawlers, and AI fetchers or agents. That is another sign that AI-generated web activity has become large enough to need its own classifications, but Akamai does not publish a single global percentage for those categories in that announcement.
The lower end roughly reflects identifiable AI crawlers growing from Cloudflare’s 2025 baseline. The upper end allows for some dual-purpose search-and-AI crawling. It should be treated as a reasonable current range, not a measured global total.
Inference, private API activity and data-center-to-data-center AI traffic are not fully captured by public web-crawler percentages, so “AI share of all internet traffic” remains unknowable from public data.Our separate guide to what AI traffic actually is explains the difference between AI crawlers, agents, model requests and large data transfers without treating them as one type of bot.
Why the AI number will probably keep rising
The reason to expect more AI-related traffic is not simply that more people are using chatbots. AI systems increasingly act as clients themselves. Search crawlers index pages, training crawlers collect data, and agents can fetch live information or interact with APIs while completing tasks.
One human request can trigger many machine-generated calls to models, websites and APIs.
AI search products need fresh web content, which creates recurring crawler and fetcher traffic.
Businesses are moving larger datasets and using external models, creating traffic that never appears as a public website crawl.
Carriers are adapting to less predictable machine-generated demand. Lumen’s new Intelligent Internet service is one example of an operator building around elastic AI and cloud traffic.
That makes a sudden jump in AI traffic plausible, but it also makes measurement harder. The more AI moves through private APIs and data-center links, the less useful a crawler-only statistic becomes as a measure of total AI network activity.
The bottom line
For September 2026, around 60% is the most useful current estimate for the share of web page requests generated by bots. Cloudflare recently measured 61% bot traffic across worldwide HTML requests, while Imperva’s broader full-year 2025 dataset measured automated activity above 53%.
AI is a much smaller part of that total. The last clean Cloudflare baseline put identifiable AI bots at 4.2% of HTML requests in 2025, with Googlebot’s 4.5% share kept separate because it serves both search and AI. Based on those figures and the continued growth of AI crawlers and agents, a cautious 2026 working range of roughly 5% to 10% for AI-related HTML requests makes sense, but it remains an estimate rather than an official global measurement.
Bot and AI traffic FAQs
What percentage of web traffic is bots in 2026?
Recent Cloudflare Radar data measured bots at 61% of worldwide HTML page requests on August 23–24, 2026. Imperva’s full-year 2025 dataset measured automated traffic above 53%. Because the datasets are not identical, around 60% is a useful current estimate rather than an exact global total.
What percentage of internet traffic is AI?
There is no authoritative global percentage. Cloudflare measured identifiable AI bots at an average 4.2% of HTML requests in 2025 and Googlebot separately at 4.5%. A cautious 2026 working estimate for AI-related HTML crawling is roughly 5% to 10%, depending on which dual-purpose crawlers are counted.
Sources
- Cloudflare Radar: Worldwide bot vs. human HTML requests, August 23–24, 2026
- Imperva: 2026 Bad Bot Report, based on full-year 2025 traffic
- Cloudflare Radar 2025 Year in Review: AI bots, Googlebot and human HTML request shares
- Akamai: AI training crawlers, AI search crawlers, fetchers and agents, September 3, 2026