You can now see every AI crawler that reads your website, which company operates it, and whether it came to train a model, index you for search, or answer a reader's question. Analytics tools that run as a page script can't show you this, because crawlers never run page scripts. We record it at the edge instead, so the visit is counted as the page is served.
Two kinds of traffic, kept apart: People who found you and bots that read you are different things, so they're counted separately rather than blended into one "referral" number.
Who's reading: 21 AI crawlers from 12 companies, including OpenAI, Anthropic, Google, Perplexity, Meta, Apple, and Amazon. Each hit is named, not bucketed as "bot traffic".
Why they came: Every crawler is labeled by intent. Training collects data for a model, search indexes you for later, and assistant means a reader asked something and the bot fetched your page to answer it.
Which posts they take: A most-crawled list shows which pieces the bots keep coming back to, which is rarely the ones you'd guess.
Where readers come from: Visits from 11 AI assistants, including ChatGPT, Perplexity, Claude, and Gemini, are separated from 11 search engines, including Google, Bing, DuckDuckGo, and Kagi.
Down to one post: Narrow any of it from your whole publication to a single piece to see how that one is being found.
Crawler tracking needs a live website on Paragraph, since the crawl has to reach a page we serve. The search and AI visitor split works on any publication, with no website required. Neither needs a paid plan.
Some AI assistants strip the referrer, so the real number of readers they send you is higher than what's shown. And crawler visits usually take a few days to appear after you publish, because the bots have to come back.
We're building an owned media engine for your company. To take advantage of AI crawlers today, simply open Analytics, then visit Search & AI, to see who's been reading. See what else Paragraph can run for you at paragraph.com.
Colin Armstrong
