Is Internet Becoming "Machine First"?
Hoomans Are Becoming The Minority...
Writing to you on a blistering hot day in Amsterdam, staring at my monitor and feeling a bit existential…
I was scrolling through my tech feeds earlier and stumbled upon a couple of pieces that honestly made me pause mid-espresso. First, there was an eye-opening article over at Techround discussing how we’ve officially crossed a bizarre digital rubicon: there are now more AI agents roaming the internet than actual hooman beings.
Then I dug into bunny.net’s 2026 Edge Security Report, which analyzed a massive sample of 42,5 billion web requests from the first half of the year. The data is incredibly solid and it points to a pretty wild reality. We are no longer living on a hooman-first internet. The web has quietly shifted into a machine-first playground.
If you look at the latest metrics from Cloudflare Radar, agentic AI bots are now driving roughly 57,4% of global web requests, leaving us flesh-and-blood hoomans with just 42,6%. We are officially the minority online. But why is it moving this way, what is the actual point and isn’t the whole ecosystem becoming a bit too convoluted?
Why is the web moving this way?
The shift isn’t just about traditional search engine crawlers or performance scrapers, those have eclipsed hooman traffic on smaller websites for over a decade. The real explosion we are seeing right now comes from agentic AI traffic. Every time a user asks a modern LLM to fetch real-time data, plan a trip or compare products, the system deploys autonomous agents to comb through the web on their behalf.
According to the bunny.net’s report, 1 in every 5 requests reaching websites is entirely automated. You’ve got aggressive crawlers like ByteDance’s Bytespider hogging almost half of all AI crawler traffic by itself, more than OpenAI, Google and Anthropic combined. We’ve essentially built a massive secondary layer of the internet where automated entities are constantly pinging other automated entities.
The convolution: What are we actually gaining?
Here’s where my inner tech nerd gets a bit skeptical about the architecture. The internet was originally designed as a protocol for hoomans to share knowledge with other hoomans. Now, we’re entering a hyper-verbose, circular loop.
Think about the current workflow: a hooman asks an AI a question. That AI sends out 10 agents to scrape 50 websites (many of which are already packed with AI-generated SEO filler content). The agents read the data, compress it and send a bloated summary back to the main LLM, which then reformats it into a digestible snippet for the hooman.
It is incredibly convoluted. We are introducing massive layers of algorithmic abstraction to handle the sheer volume of data, yet it feels like we aren’t creating anything genuinely new. It’s a digital echo chamber where machines are summarizing machine-written text for other machines.
The real-world cost of a machine web
This is the part that genuinely concerns me, because this machine-first internet doesn’t run on fairy dust, it runs on high-performance silicon. The specs required to keep up with this automated traffic are astronomical and the real-world side effects are starting to bite.
Skyrocketing infrastructure prices: Hosting a web application used to be relatively cheap. Now, because of the sheer volume of automated requests, infrastructure costs are creeping higher. Bunny.net highlighted that automated bot waves are scaling up aggressively; they even mitigated a massive Layer 7 DDoS attack peaking at 16,5 million requests per second from over 392.000 unique IP addresses. To survive this relentless automated background noise, businesses have to purchase heavier edge security and beefier cloud servers just to filter out the bots so that the 42,6% of hooman users can actually load a page.
Environmental toll: We are effectively burning through megawatts of electricity and driving up carbon emissions to keep data centers cool, all so bots can scan pages for other bots. The efficiency ratio here is terrible. We are sacrificing real-world planetary resources to sustain a loop of digital noise.
Unsolved problems: Are these agents actually fixing the fundamental bugs of society? It doesn’t feel like we are solving poverty, curing diseases or even fixing the delay on my local Amsterdam tram lines with this tech. Instead, we are introducing an entirely new set of high-maintenance problems, like advanced credential abuse, application-layer exploits and resource exhaustion, without solving the structural issues we already had.
I’m not trying to pull off a cynical Luddite rant here. I love optimization, sleek automation and raw computing power as much as any geek. But when the architecture of the world’s primary communication tool pivots to serve autonomous code rather than the people who built it, we have to wonder what the actual end game is…
What do you think? Are you ready to hand over the keys of the web to our new agentic overlords or should we start putting up “Hoomans Only” signs on our protocols? Drop a comment below and lets chat before a bot does it for us.