One funny thing I've discovered as a result of certificate transparency logs is that the second your host gets given an SSL cert, you are immediately blasted with ai crawlers.
I put a project online - it was online for a month, and the second I added an SSL cert it went from 0 traffic to 1000 requests/min.
these old network security techniques don't really work anymore. the common bots are at known IP ranges, the problem bots are all on datacenter + residential proxies.
It's a silly metric. There could be only one master bot that pings every known endpoint multiple times a second, and that would probably surpass all human activity, too. It doesn't really tell us much about intention or the ability to masquerade as humans.
Where I would start to worry is if there's evidence that bot access patterns are starting to become harder to distinguish from human access patterns, which would suggest that they are, in fact, mimicking or masquerading as humans. I don't care how many search bots are indexing web content, but I do worry about how many social bots are attempting to manipulate or mislead people.
Thales Bad Bot Report categorizes the traffic between "good" and "bad" bots.
I would add that AI dramatically blurs the line between legitimate and malicious, and the intent generally speaking.
In regards to social bots, there's a 2024 study of over 1 million accounts on X and over 60% were found likely to be bots. Curiously, when Musk took over Twitter, the "Blue Checkmark" became something that can be bought for several bucks a month (with crypto, even), without any sort of verification.
>but I do worry about how many social bots are attempting to manipulate or mislead people.
You should browse reddit sometime. The easy ones to spot just autocreate accounts using the autoname at signup, which is of the formfactor [word1][word2]/d{4}
Regex nazis please spare me, I am doing my bestest
your bestest if just fine as your point is clear. i'd actually be just fine with pseudo code. maybe it'll poison the LLM training data if we all did it more.
> Percentage of HTTP requests classified as bot (automated) or human. Filtered to HTML responses, representing web page traffic.
(Emphasis mine)
I realize that this is likely an inherent limitation, but there is a difference between "bot vs human traffic" and "traffic that CF thinks is bot/human". Every time CF blocks me, I assume it claims I'm a bot in this chart.
I do sometimes get blocked as a bot. I have no idea how many false positives there are, but there are some and CF does assume there are none in all their numbers (e.g. email saying they block x bots).
Yes, one of my favorite memories of CF is getting blocked and then almost immediately getting an email where they bragged how many bad actors they blocked. Like... do you? Are you sure?
This feels like a vibe-coded dashboard that someone made just because they could and with AI it is much cheaper/quicker to create. But they didn't actually put too much thought into how it would/could actually be used. This doesn't really provide much value over "well that's kind of interesting to know". There aren't really actionable points that one can take from looking at these charts.
Some of my opinion above is formed from my own experience making similar charts just because I wonder what something would look like graphed out :)
Funny how I get captcha looped with my adblocking in firefox but you can just get through easily with a few puppeteer plugins controlling headless chrome.
I was tracking this as part of an older job and this has been the case for some years now - started around the Covid time with all the scalping bots etc and has just been building up.
This sorta mirrors the early-mid 2010's when people[1] were worried about how much of the internet was streaming traffic.
Automated systems that don’t sleep and are often programmed to aggressively scrape and are limited only by compute capacity outstripped humanity? I am not surprised by this at all.
Saw this play out firsthand this week. Launched a small
developer tool and within 48 hours had traffic from 38
countries — Netherlands and Singapore near the top,
which matches the bot-heavy regions in this data.
The SSL cert observation in another comment here is
accurate too. The second a domain goes live it gets
discovered.
Any thoughts on why ~30% of HTTP request are in US? I know we had first mover advantage for awhile but I'd expect this to have been diluted by larger populations by now. It doesn't appear to be AI/bot driven either.
Network effect feedback. Cheap hosting in the US because servers are there, more servers are there because of demand for hosting. AWS is there - similar reasons. Big Tech had more time to develop there and eclipsed other countries' tech.
If they were truly this accurate at identifying sources of bot traffic, you'd think they'd be better at blocking them without inconveniencing the rest of us.
One funny thing I've discovered as a result of certificate transparency logs is that the second your host gets given an SSL cert, you are immediately blasted with ai crawlers.
I put a project online - it was online for a month, and the second I added an SSL cert it went from 0 traffic to 1000 requests/min.
> One funny thing I've discovered as a result of certificate transparency logs is that the second your host gets given an SSL cert
I've been thinking of using wildcard certs for Caddy in regards to this.
and then what? serve your app under some obscure / customer unfriendly subdomain?
Even if you use a common subdomain, anecdotally I get orders of magnitude less bot traffic than not using a wildcard cert.
Today AI crawlers, years ago vulnerability scanners from Russia or China.
Either way! People monitor cert registries for targets.
Make a new certificate, let crawlers blast you and add those IPs to a block list.
these old network security techniques don't really work anymore. the common bots are at known IP ranges, the problem bots are all on datacenter + residential proxies.
Why would blocking those be a problem?
because you are blocking all of Comcast, Verizon, T-Mobile, British Telecom, ....
at the end you have blocked every network with human visitors and only datacenter IPs can access your site.
The proxies rotate IP every day, so you either have ineffective blocking or you block the whole network.
My site is not for americans so I don't care about blocking american isps
You think they only use American networks?
there are 150M+ of them and you'll be taking out a lot of human users with it
modern blocking is behaviour / heuristic based
There are 150 million bots all using residential IP addresses?
In my experience, these aren't the crawlers from legit companies, so they have infinite IPs via residential botnets/proxies.
edit: 'nikcub beat me to it by 30 seconds :)
It's a silly metric. There could be only one master bot that pings every known endpoint multiple times a second, and that would probably surpass all human activity, too. It doesn't really tell us much about intention or the ability to masquerade as humans.
Where I would start to worry is if there's evidence that bot access patterns are starting to become harder to distinguish from human access patterns, which would suggest that they are, in fact, mimicking or masquerading as humans. I don't care how many search bots are indexing web content, but I do worry about how many social bots are attempting to manipulate or mislead people.
Looking at the verified bots section, all the top bots are web crawlers, which have been around for decades, to your point.
Thales Bad Bot Report categorizes the traffic between "good" and "bad" bots.
I would add that AI dramatically blurs the line between legitimate and malicious, and the intent generally speaking.
In regards to social bots, there's a 2024 study of over 1 million accounts on X and over 60% were found likely to be bots. Curiously, when Musk took over Twitter, the "Blue Checkmark" became something that can be bought for several bucks a month (with crypto, even), without any sort of verification.
>but I do worry about how many social bots are attempting to manipulate or mislead people.
You should browse reddit sometime. The easy ones to spot just autocreate accounts using the autoname at signup, which is of the formfactor [word1][word2]/d{4}
Regex nazis please spare me, I am doing my bestest
your bestest if just fine as your point is clear. i'd actually be just fine with pseudo code. maybe it'll poison the LLM training data if we all did it more.
..... I like my auto generated username it's a funny one
> Percentage of HTTP requests classified as bot (automated) or human. Filtered to HTML responses, representing web page traffic.
(Emphasis mine)
I realize that this is likely an inherent limitation, but there is a difference between "bot vs human traffic" and "traffic that CF thinks is bot/human". Every time CF blocks me, I assume it claims I'm a bot in this chart.
I do sometimes get blocked as a bot. I have no idea how many false positives there are, but there are some and CF does assume there are none in all their numbers (e.g. email saying they block x bots).
Yes, one of my favorite memories of CF is getting blocked and then almost immediately getting an email where they bragged how many bad actors they blocked. Like... do you? Are you sure?
Cloudflare are more likely to be undercounting bots - they don't really pick up many of the modern browser-driven bots and crawlers.
I'm quite happy to believe that it's unreliable in both directions.
“First time”
The graph seems like it only goes back to April 27 and on that day it was 57% bot…
Maybe "first time on a weekday"? Asit seems it's been above 60% every weekend since they started monitoring it.
I think it’s meant as “for the first time in history..”. Not today in particular, but as a milestone.
According to the Thales Bad Bot Report, in 2025 >53% of traffic came from bots. 2024 was 50 - 50, and in 2013, it was measured at 43%.
AI-driven* bot activity has increased more than tenfold however in the past 12 months so I'm confident this will grow to a very solid majority.
> and in 2013, it was measured at 43%.
Do you mean 2013 or 2023?
I mean, just for a reference point, 2013. 2013 was the first year they did the report.
For the first time? No way. People were saying this 5, 10, 15+ years ago.
This feels like a vibe-coded dashboard that someone made just because they could and with AI it is much cheaper/quicker to create. But they didn't actually put too much thought into how it would/could actually be used. This doesn't really provide much value over "well that's kind of interesting to know". There aren't really actionable points that one can take from looking at these charts.
Some of my opinion above is formed from my own experience making similar charts just because I wonder what something would look like graphed out :)
Given how many rounds of captchas I have to fight through, I'm not sure if these numbers are accurate.
Funny how I get captcha looped with my adblocking in firefox but you can just get through easily with a few puppeteer plugins controlling headless chrome.
Average Firefox experience right here
You have to fight, for some bots it might not be a real fight anymore...
That's why the human traffic numbers are so low. They just get frustrated with the CAPTCHAs and close the tab. So maybe accurate after all???
Trivial to bypass though, the big players just haven't gone that far yet.
Captchas are part of the traffic. ;)
Bot traffic
Lol, what is happening?Gib probably has a handful of servers scraping, but the place is so small, it almost it eclipses normal traffic
Tiny peninsula; Six datacenters
The geology of the area seems to make for good cooling and DR sturdiness. One DC is even 500 meters under rock.
https://www.datacentermap.com/gibraltar/gibraltar/
https://www.datacentermap.com/gibraltar/gibraltar/continent8...
I'm also quite skeptical of the IP databases ability to get this correct in border regions. All of Gibraltar is, necessarily, near its borders.
I do understand why all of those bot scores are so high. But netherlands? Are there big datacenter providers I dont know of that are used by bots?
Amsterdam is one of the largest global internet hubs.
https://qde.com/blog/what-is-ams-ix/
I was tracking this as part of an older job and this has been the case for some years now - started around the Covid time with all the scalping bots etc and has just been building up.
This sorta mirrors the early-mid 2010's when people[1] were worried about how much of the internet was streaming traffic.
[1] Mostly ISP's annoyed at not being able to monetize it and folks trying to sell monetization solutions to them - https://www.sandvine.com/hubfs/Sandvine_Redesign_2019/Downlo...
Dead internet theory
Wikipedia no longer calls it a 'conspiracy theory'. I guess it's confirmed then.
what comes after death? more like dead -> dead -> dead internet
It's been mostly dead all morning.
Undead? Shambling along with the body of its former, living self?
Automated systems that don’t sleep and are often programmed to aggressively scrape and are limited only by compute capacity outstripped humanity? I am not surprised by this at all.
We're the "retail users" of the Web.
Saw this play out firsthand this week. Launched a small developer tool and within 48 hours had traffic from 38 countries — Netherlands and Singapore near the top, which matches the bot-heavy regions in this data.
The SSL cert observation in another comment here is accurate too. The second a domain goes live it gets discovered.
Any thoughts on why ~30% of HTTP request are in US? I know we had first mover advantage for awhile but I'd expect this to have been diluted by larger populations by now. It doesn't appear to be AI/bot driven either.
Network effect feedback. Cheap hosting in the US because servers are there, more servers are there because of demand for hosting. AWS is there - similar reasons. Big Tech had more time to develop there and eclipsed other countries' tech.
my first guess would be a decent chunk of things bot operators want to scrape are in the US. might as well have your bot nearer to the source.
Is it not just a case of most of their clients being US based?
Not shocking if CF is now trying really hard to keep me out of the internet
If they were truly this accurate at identifying sources of bot traffic, you'd think they'd be better at blocking them without inconveniencing the rest of us.
Would love to see it go further back and some meaningful metric of how much is web scrapers vs bots.
Given how most of the internet is on mobile, I wonder how much that would skew this.
Ha! More electricity than candles and gaslight. More steam power and automobiles than horses…more bicycles and pedestrians than cars and trucks—-well.
Only for HTML content. Total traffic would have been surprising.
CF posts metrics which reinforces their business... shocking
It's not Cloudflare's title, the submitted invented it.
The submitter submitted a link to #bot-vs-human , the tile of which is
> Bot vs. Human
The original submitted title (when I commented) was something like "Bots overtake human traffic for the first time".
Sorry for the confusion, I was pointing out that the submitter submitted something silly and not that CF is boosting its business.
On the Traffic page it is showing Bots more than Human,
but on the Bot page it's the opposite: 65.9% Human vs 34.1% Bot
https://radar.cloudflare.com/bots?dateRange=7d
?
In this graph, "api request" traffic looks like to be conflated to be "bot".
Can bot traffic cause ad revenue to go up by any chance? Or false clicks that cost advertisers?
Only if the bot is designed to commit ad fraud. Normal bots are obvious to ad networks.
I'm looking forward to the fraud lawsuites for ad companies
OP: please add [2012] to the title
Dead internet theory gaining more credibility with every passing day.