Growth in traffic from AI agents and agentic browsers, 2025 vs 2024, from a very low base.
HUMAN SecurityRise in AI-driven bot attacks in 2025 compared with the previous year.
Thales (Imperva)Share of tested popular websites that stopped all ten test bots in 2026.
DataDomeAbout these numbers. Every figure here comes from HUMAN Security, Thales (Imperva), DataDome or Cloudflare. These companies sell bot protection or traffic management, and each measures its own network and customers. AgentBacker is a third party summarising their published findings. Each figure is quoted with a link to its source. We have added no estimates of our own.
1Agent traffic is rising fast
HUMAN Security reports that in 2025 traffic from AI agents and agentic browsers grew 7,851% year over year.
HUMAN adds that 2024 volumes started from a very low base.
(HUMAN, 2026 State of AI Traffic & Cyberthreat Benchmark Report)
Cloudflare: Over the past year, daily requests from AI agents on our network grew by more than 1,700%.
(Cloudflare, 30 Sep 2026)
- DataDome counted
17.7 billion AI agent requests in Q2 2026, up from 12.2 billion in Q1.
DataDome's count includes AI crawlers, not only agents acting for a person. (DataDome, AI Traffic Report Q2 2026) - Thales:
In 2025, bots made up more than 53% of all web traffic, up from 51% the previous year.
(Thales 2026 Bad Bot Report)
2Agents reach accounts, logins and checkout
Most agent activity is browsing. A meaningful share is not. HUMAN: In 2025, 77% of agentic AI activity was on product and search pages, while 8.8% occurred on account pages, 5% on authentication flows, and 2.3% on checkout pages.
- DataDome, 2025:
64% of AI bot traffic reached forms, 23% login pages, and 5% checkout flows.
(DataDome 2025 Global Bot Security Report) - DataDome, first half of 2026:
AI agents generated 605.6 million requests to login pages, forms, carts, payment flows, and account-creation pages, with login pages accounting for 51.7% of that traffic.
(DataDome, State of Bot & Agent Security Report 2026)
3Fraud and abuse are rising
Thales: In 2025, AI-driven bot attacks surged 12.5x compared to the previous year.
Its release adds that bots account for over half of all traffic, with 40% classified as malicious.
The other figures below cover bot activity of all kinds, not AI agents alone.
AI agents have already been used in carding-like attacks. HUMAN documents one case in which a threat actor ran 11 card-add attempts and 6 payment attempts across two sessions, followed by a pivot to loyalty-point redemption after the card paths failed, mirroring established carding workflows, but executed through an AI browser agent.
4Few sites are protected
DataDome tests popular websites each year with a set of simple bots. In 2026, 65.3% of tested websites stopped none of the 10 bot types evaluated. Only 2.4% stopped all of them, down from 8.4% in 2024 and 2.8% in 2025.
Among DataDome's own customers, sites that already pay for bot protection, 54% of DataDome customers have already implemented agent trust policies.
(DataDome, Q2 2026)
What this means for smaller vendors
The figures above come from networks that see billions of requests. A small online business sees none of that. It cannot tell an agent from a person, and it has no team watching its login and checkout pages.
Rejecting all agents is the usual answer. It has a cost. As Cloudflare puts it: Turn an agent away, and you may be turning away the person who sent it.
AgentBacker offers a third option: accept agents that have a known, accountable person behind them. To get an AgentBacker pass, the person behind the agent passes an ID check with a selfie and signs a liability agreement. You receive the verified name and the signed agreement. You do not need to detect or classify the agent. You know who answers for it.
AgentBacker does not replace your payment checks or bot protection. It gives you a reason to say yes to the agents that come with a name attached.
agentbacker.com
Sources
All figures were checked against the pages below on 14 October 2026. Where a company's own site blocked automated access, we checked the text of its press release as distributed by Business Wire.
- HUMAN Security, The 2026 State of AI Traffic & Cyberthreat Benchmark Report (covers 2025), published 26 March 2026. Report PDF · report page
- Thales, 2026 Bad Bot Report: Bad Bots in the Agentic Age, press release, 29 April 2026. imperva.com
- DataDome, 2025 Global Bot Security Report, press release, 30 September 2025. datadome.co · Business Wire
- DataDome, The AI Traffic Report Q2 2026, 16 July 2026. datadome.co
- DataDome, The State of Bot & Agent Security Report, 2026 Edition, press release, 22 September 2026. Business Wire
- Cloudflare, The Internet has a second audience, 30 September 2026. blog.cloudflare.com
AgentBacker · agentbacker.comThird-party summary of published findings. October 2026.