Track your site's AI traffic

Agent Analytics is the part of Searchly that shows the visits your site gets from AI assistants. It reads the sessions Google Analytics 4 attributes to ChatGPT, Perplexity, Gemini and other AI assistants, adds your Search Console data, and lines both up with the pages that tracked AI answers cite. Use Agent Analytics to see what was observed on your site, and Technical Optimization to see what an audit of your saved page finds.

GA4 already records a source for each session, and visits from an AI assistant often arrive with a source such as chatgpt.com or perplexity.ai. Sign in with Google and choose your GA4 property, and Searchly groups those sources into ChatGPT, Perplexity, Gemini and other AI assistants for you. Access is read-only, data syncs once a day, no new tracking code is needed, and the connection comes with every plan.

Searchly recognizes sessions whose GA4 source is ChatGPT, Perplexity or Gemini, and counts a short list of other assistants, including Claude, in an Other AI group. The match is made on the session source GA4 reports, which campaign parameters can also set. Visits from Google AI Overviews arrive in GA4 as Google traffic, so they cannot be counted separately, even though Searchly tracks Google AI Overviews answers.

GA4 can only attribute a session to an AI assistant when the visit carries a source it can record. Copied links, apps that open a link without a referrer, browsers that block analytics and assistants that withhold the source can all leave a visit unrecorded or without a source. Searchly reports what GA4 recorded and does not estimate the rest, so read the numbers as the part of your AI traffic that analytics can see. A day that has not synced is shown as unavailable, not as zero, and partial totals are marked as lower bounds.

Yes, from your own site's request logs or code. Generate a token in Setup and follow the guide for your platform: a Cloudflare Logpush job or Worker, a Vercel log drain, middleware for Next.js, a Netlify Edge Function or a WordPress plugin. Webflow and Framer sites can connect through Cloudflare on the plans that allow it, and any other sender can post request events to Searchly's HTTP API. Searchly keeps requests from the AI crawlers of 13 operators, including OpenAI, Anthropic, Perplexity and Google's user-triggered agents, and labels each by operator and by the purpose the operator declares: training, search index or user-triggered. A request is verified when the delivery includes its IP address and that address is inside the ranges the operator publishes; otherwise it is declared. Requests from people and other bots, including Googlebot and Bingbot, are discarded. Searchly stores up to 50,000 crawler requests per project each UTC day.

No. For each page, Searchly shows how often tracked AI answers cited it, the AI sessions it received and its Search Console clicks, matched on the exact host and path. Page evidence is a snapshot of one day within the period you select. The product labels it evidence, not attribution, and it does not report conversions or revenue.

Yes. Page evidence exports to CSV. The Searchly API returns page evidence, GA4 traffic and Search Console reports as JSON, and the Searchly MCP server offers the same reports as tools to MCP clients that support Streamable HTTP and custom request headers. The workspace owner turns on API access and creates scoped keys that expire.

GA4 records the sessions. It cannot show which of those pages AI answers cite. Searchly groups the AI sources for you and puts citations, AI sessions and Search Console clicks on one row per page. Citations come from the AI answers Searchly already collects for your tracked prompts, so they need no extra setup.