Ambit Signal

api | install-astro | install-caddy | install-cloudflare | install-express | install-js-tag | install-netlify | install-nginx | install-shopify | install-squarespace | install-sveltekit | install-vercel-next | install-webflow | install-wix | install-wordpress | mcp | outcomes | privacy | quickstart | troubleshooting

Ambit Signal documentation

Ambit Signal measures how AI assistants find, read and send people to a website, and what those people do next. These docs are written for developers and AI coding agents: every page gives exact file paths, environment variable names, commands and a verify step.

Method and definitions: https://ambit.agency/signal/ . Open AI-agent registry: https://ambit.agency/signal/registry.json

What Signal records

Class Meaning
ai_fetch An AI assistant fetched the page live to answer a real person's question (ChatGPT-User, Perplexity-User, Claude-User, and others).
ai_crawler AI training or index crawler (GPTBot, OAI-SearchBot, ClaudeBot, and others).
search_bot Classic search crawler (Googlebot, bingbot).
ai_agent A request carrying a Web Bot Auth Signature-Agent (an agent browsing for a person).
ai_referral A human who arrived from an AI surface (utm_source=chatgpt.com, or a chatgpt.com, perplexity.ai, gemini, copilot or claude.ai referrer).
ai_sticky A returning human whose hashed visitor key had an AI referral in the previous 30 days.
ai_estimated A direct visit (no referrer, no utm) to a page an AI fetched for a person within the previous 10 minutes.
human / other_bot Everything else.

Bots are verified when their IP is in the vendor's published ranges, their reverse DNS forward-confirms, or their Web Bot Auth signature verifies. A bot claiming an AI identity from the wrong place is spoofed.

The three tiers (never added together)

Tier Class How it is known
confirmed ai_referral The request itself says it came from an AI (referrer or utm).
probable ai_sticky The same hashed visitor came from an AI within 30 days.
estimated ai_estimated Timing inference: an AI fetched the page for someone, then a direct visit landed on it.

Reports always show the tiers side by side. Do not sum them in anything you build on top of the API.

Two ways to collect

Path Sees Needs Use for
Server collector bots and humans site secret (server only) Cloudflare, Vercel/Next.js, Express, Astro, SvelteKit, Netlify, WordPress, nginx, Caddy
JS tag humans only public site_id Webflow, Squarespace, Wix, Shopify, any static site

AI crawlers and live AI fetches do not run JavaScript, so only a server collector sees them. On a hosted builder, put the domain behind Cloudflare and add the Worker to see bots too. Server collector plus JS tag together is the most complete setup: the tag adds first-touch attribution on forms.

Pages