DeepSeekBot

DeepSeek โ€” Training

All crawlers ยท Agent Accessibility Index

Purpose: Training. Bulk crawling to build model training data.

DeepSeek's crawler, collecting content for training its models.

4.22%

of the panel names this token in robots.txt and disallows it, for the week of 2026-10-05.

How sites treat this token

StanceDomainsWhat it means
FullBlock38Every path is disallowed for this token.
PartialBlock4Some paths are disallowed for this token; others are not.
Allowed4The token is named and explicitly allowed.
Unmentioned949The file never names this token. This is not consent โ€” the site may still close it off with a blanket rule for every crawler.

DeepSeekBot's three-year history

DeepSeekBot over time, 16 quarterly periods from 2023-01 to 2026-08.DeepSeekBot rose from 0.00% at 2023-01 to 5.04% at 2026-08.0%2%4%6%2023-012023-092024-052025-012025-102026-085.04%

A site in the historical panel first named and disallowed this token in 2025-10; its most recent reading is 5.04%.

Note: this chart reads Common Crawl's own 611-domain historical panel, not the panel behind the stance table above โ€” a different population from the live index panel, so levels are not comparable between the two indices.