Enterprise AI in Numbers · AI search (GEO/AEO)
Share of AI crawler traffic collecting content for model training
44.0%
44.0% as of October 25, 2026, according to Cloudflare Radar, AI Insights. Share of AI bot requests over the last 28 days that come from crawlers declared for model training; parts give every purpose (training, mixed, search, user action, undeclared).
What it means for leaders
Most AI crawling still feeds model training, which sends no visitors back, while search and agent crawling, which can, is a smaller share. Set crawler rules by purpose, not by company.
What would change the reading: Search and user-action crawling growing as a share of the total.
The breakdown
Every reading
| Date | Value | Change | Source |
|---|---|---|---|
| 44.0% | — | Source |
Definition and method
- What it measures
- Share of AI bot requests over the last 28 days that come from crawlers declared for model training; parts give every purpose (training, mixed, search, user action, undeclared).
- Population
- HTTP requests from AI bots on Cloudflare's network
- Updated
- Each update (last 28 days)
- Source
- Cloudflare Radar, AI Insights · CC BY-NC 4.0
Caveats
- Purposes are those the crawler operators declare, as classified by Cloudflare.
- Covers sites on Cloudflare's network only.
Cite this
Enterprise AI in Numbers, "Share of AI crawler traffic collecting content for model training", https://lucianoayres.com/ai-brief/data/ai-crawling-for-training/, retrieved October 25, 2026. Data: Cloudflare Radar, AI Insights.