Agent Attention: Measurable demand from AI systems for your content. When an AI agent crawls your page and cites it in an answer, that's not a bot visit - it's someone asking a question and your content being selected as the answer. It is page-level, global, and near real-time.
Bot Activity Types: Agent Analytics classifies every verified bot visit into one of three intent types:
Agent Analytics vs. GA4: GA4 runs JavaScript in the browser. AI agents don't use browsers and never trigger GA4, so GA4 is blind to agent traffic. Agent Analytics reads your CDN's network logs (Cloudflare, Vercel, Fastly), capturing every HTTP request from every AI system. Agent Analytics shows you the input (what AI is crawling and citing). GA4 shows you the output (what humans do after AI sends them to you). Use both together.
Four headline metrics at the top: AI Citations, AI Training, Human Referrals, and % of total traffic from AI systems.
Read the relationships, not the individual numbers:
Every row is a page on your site, with columns for citations, training, indexing, human referrals, and referral traffic share.
Look for mismatches:
Watch Pages lets you pin specific URLs to track over time. Use for pages where you've made deliberate AEO changes so you don't have to hunt for them.
Sankey Diagram shows how different AI platforms path to your pages. Different AI systems have different crawling patterns - some may focus on your blog, others on product pages.
Popular Pages by Bot breaks down page distribution per individual bot. Notice whether different platforms are seeing different slices of your site.
Top-line metrics break out total bot visits by intent type. The platform table underneath shows each AI platform with its user agents, total visits, and citation visits.
Key read: the gap between total bot visits and citation visits per platform. A platform with lots of training visits but zero citations has a fundamentally different relationship with your content than one citing you constantly.
Click into any platform to see individual user agents (e.g., OpenAI runs ChatGPT-User, GPTBot, and OpenAI-SearchBot, each serving different purposes).
Shows how many real humans arrived because an AI agent cited you, and from which platforms.
Compare against bot visits: a platform might cite you thousands of times, but how many resulted in a human click? That ratio reflects how each platform presents citations - some surface links prominently, others embed information so completely that users never need to click.
Human Visits by Page is a different ranking than the Pages view. Pages with the most bot citations aren't necessarily the ones with the most human clicks.
Every verified bot visit: platform, bot type, user agent string, path, HTTP status code. You won't check this daily, but go here to debug unexpected patterns.
The status column is especially useful:
Profound doesn't just read user agent strings. Bot classification uses multi-step verification: user agent matching, IP range checks, reverse and forward DNS lookups, ASN lookups, and behavioral pattern matching.
