AI traffic
The AI Traffic section sits on your dashboard between Pages and Locations. It reports AI crawlers (GPTBot, ClaudeBot, PerplexityBot, CCBot and around 50 others) fetching your pages.
This section needs a second install, on your server. AI crawlers never run JavaScript, so the tracking snippet cannot see them. See AI Crawlers for why, and Install for how. Until it’s installed, the section shows setup instructions instead of data.
Activity over time
One line per crawler operator, day by day, with the full operator list beside it: who fetched, how many times, and what share of the total. The five busiest get their own line; everything else collapses into a muted Other crawlers band so the chart stays readable, and the list still accounts for every one.
Click an operator in that list to filter the whole section to it.
The category tabs above the chart (Answers, Indexing, Training, Other AI, Markdown) narrow the chart and the pages table below it, so the two never describe different subsets. Switching the card to its Pages view is the one exception: those lines are per-path and carry no intent, so the chart notes which scope it could not follow rather than implying it did.
Pages
Which pages get taken, and what happens next.
| Column | Meaning |
|---|---|
| Page | The path fetched. |
| Crawlers | Which operators took it. |
| Read for | The intents behind those fetches. |
| Fetches | Total requests. |
| Answers | How many were answer fetches. |
| Verified | Share Snow could confirm. |
| Cited | Visitors who arrived from an AI assistant on that page. |
Cited is the join worth studying: it puts “GPTBot crawled /pricing 40
times” next to “and sent 3 visitors” on one row.
Three independent filters (crawler, intent, verification) narrow the table, and they compose: every number is recomputed inside the filter rather than showing the page’s untouched total.
Reading the crawl-to-visit ratio
A page crawled heavily but never cited is being read and not recommended. A page crawled rarely but citing well is punching above its weight, and probably deserves more internal links pointing at it.
Don’t read a high ratio as failure on its own. Training crawls are bulk collection and produce no immediate visits by design. Filter to Answers to see the intent that actually correlates with referrals.
Verified only
The Verified only toggle sits in the activity card’s header, beside the category tabs. It restricts every panel in the section, chart and table alike, to hits Snow could confirm.
Use it when you need numbers you can stand behind; leave it off for the honest total. Expect the figures to drop noticeably: that is not a sign anything was fake.
Nothing here?
- Not installed yet. The section shows setup instructions with a copy-paste snippet for your stack.
- Installed but empty. A real zero. Googlebot and Bingbot are excluded by design, so a well-indexed site can still show nothing until an AI crawler visits. New sites often wait days.
- Installed, and you expected more. A CDN or full-page cache may be answering requests before your server sees them. See Common problems.