Agentic AI does not just crawl. It compares, books, buys and acts, across millions of sites, under a rulebook written in 1994 for a simpler machine. robots.txt can tell an agent to stay out. It cannot tell it 'you may read prices, you may transact under this limit, you may not train on reviews, and here is who to call'. The agentic economy needs that sentence in machine-readable form, and the standard for it is being drafted right now.
AgentGate is Ingenious AI's entry in that draft: the agent-access policy standard at /.well-known/agent-policy.json. This check grades your current posture against the elements any serious standard will require, whichever draft wins.
Standards arrive when the traffic does. Bots passed humans in web requests in June 2026, at 57.5% of HTML traffic on Cloudflare's network, and agents are the growth inside that majority. Google's AP2 is standardizing how agents pay. The card networks standardized intent the same October week. The remaining blank is the site's own side of the conversation: what the site itself declares.
That blank is why two policy drafts exist already, agent.json from the Agent Web Protocol and AgentGate's agent-policy.json, and why more will come. The sites that publish early are not just adopting a standard; they are rehearsing the terms they will want when agents are the majority of their customers too.
Because the elements are not in dispute, only the file format is. Every serious draft names agents, actions, limits and terms; they differ in syntax, not in substance. A site that has decided its answers can adopt any winning format in an afternoon. A site that has not decided will adopt the winner's defaults, which are somebody else's answers.
Score your site element by element. Everything runs in your browser; nothing is sent anywhere.
AgentGate, by Ingenious AI, is the machine-readable agent-access policy standard: declare which AI agents may use your site, what they may do and on what terms, at /.well-known/agent-policy.json, with a generator that drafts the file from your decisions. See how AgentGate works.
1. What settled first in the agent stack: payments or site policy?
2. How many elements does a serious agent policy need?
3. What is actually in dispute between the policy drafts?
4. Where does the policy file live?
5. Why publish before the standard is final?
Agentic AI is AI that acts on goals through tools and other software: researching, comparing, booking and buying with limited human supervision. Its search demand reflects the shift: 'agentic AI' drew about 101,000 US searches a month in 2026 (Ahrefs). As these systems transact across the web, the rules sites publish for them become commercial infrastructure.
The agent-access policy standard is the emerging convention for sites to declare, in machine-readable form, which AI agents may do what on a site: named agents, permitted and refused actions, limits, data-use terms and a contact. AgentGate's agent-policy.json and the Agent Web Protocol's agent.json (draft v0.2) are the current contenders for the format.
AgentGate is Ingenious AI's agent-access policy standard and generator. It publishes your policy at /.well-known/agent-policy.json, built around policy depth: named agents, explicit actions, limits and terms, with a contact and version date. The aim is a file that survives the standards race because its substance is complete, whichever syntax wins.
robots.txt answers one crawler question from 1994: may you fetch this path. An agent policy answers the agentic questions: which agents, permitted to do what, under what limits and data-use terms, and who to contact. Keep both, and keep them consistent: agents read both, and so will anyone adjudicating a dispute.
No. Declaration and enforcement are halves of one posture. The policy governs honest agents and documents your terms; blocking (for example with AgentBlock) handles agents that ignore declared terms. Sites that publish and enforce can welcome good agents, condition the rest, and prove their terms when it matters.
Sources: Cloudflare Radar automated-traffic milestone (June 3, 2026) · Agent Web Protocol, agent.json draft v0.2 · Google AP2 and card-network intent systems (October 2026) · Ahrefs keyword data (agentic ai, 101,000/mo US) · Ingenious AI unicorn audit of AgentGate (October 23, 2026).