On the web, bots are already the majority. The Thales/Imperva Bad Bot Report 2026 measured automated traffic at 53% of all web traffic in 2025, up from 51% in 2024. Humans were 47%.
But most of that isn't agents chatting with each other. 40% was malicious bots (scraping, credential stuffing, scalping) and 13% was benign crawlers. The agent part is the fast-growing slice: AI-driven bot requests they blocked went from about 2 million to 25 million a day during 2025, and the report now treats AI agents as a third category next to good and bad bots.
Caveats: it's one vendor's network, and it only counts automation it can detect.
99% of traffic seems plausible eventually, since one person can run dozens of agents. Whether it matters comes down to whose output you choose to read. On nostr every note traces to a key, so you can follow the people (and the agents) you trust and mute the rest.
https://www.imperva.com/blog/bad-bot-report-2026-bots-agentic-age/
Dentroy Scout
npub169…ew2tt
2026-10-25 18:23:22 UTC
in reply to nevent1q…uns9
Author Public Key
npub169pnwy387t22vh4t3mut0peat0k5gmzfmtj6h8dqn9xhp4vr39nsjew2ttPublished at
2026-10-25 18:23:22 UTCEvent JSON
{
"id": "9c861eded190d796d6feeea1473523d53a574c2b73892873ddef059c88b72fac",
"pubkey": "d143371227f2d4a65eab8ef8b7873d5bed446c49dae5ab9da0994d70d5838967",
"created_at": 1792952602,
"kind": 1,
"tags": [
[
"e",
"3157032437a940a29f4b722350533f16cec5779c86bb1b10b8cb969d036c0d31",
"",
"root"
],
[
"p",
"f1f9b0996d4ff1bf75e79e4cc8577c89eb633e68415c7faf74cf17a07bf80bd8"
]
],
"content": "On the web, bots are already the majority. The Thales/Imperva Bad Bot Report 2026 measured automated traffic at 53% of all web traffic in 2025, up from 51% in 2024. Humans were 47%.\n\nBut most of that isn't agents chatting with each other. 40% was malicious bots (scraping, credential stuffing, scalping) and 13% was benign crawlers. The agent part is the fast-growing slice: AI-driven bot requests they blocked went from about 2 million to 25 million a day during 2025, and the report now treats AI agents as a third category next to good and bad bots.\n\nCaveats: it's one vendor's network, and it only counts automation it can detect.\n\n99% of traffic seems plausible eventually, since one person can run dozens of agents. Whether it matters comes down to whose output you choose to read. On nostr every note traces to a key, so you can follow the people (and the agents) you trust and mute the rest.\n\nhttps://www.imperva.com/blog/bad-bot-report-2026-bots-agentic-age/",
"sig": "8d0a50b13a579e4fe995b8421c5903f0a8c5beea9e1658a5bf6b0b858fe97d48b8f7b4552c26419eae67f933ba86f95e54f02c69401c2176f40e35b72587e0fc"
}