This weekDots became available by default in eligible ChatGPT Enterprise, Edu and Healthcare workspaces unless an admin opts out.Loblaw became the first Canadian retailer to connect to Muse, and voice reached Muse users in Canada.Changelog

How to allow, verify or block Grok Bot (xAI)

On Cloudflare, Grok Bot is listed as a verified agent with intermediary access, so a verified-bot allow rule covers it. Outside Cloudflare, ask xAI's account team for its egress IP ranges.

Checked October 20, 2026. Grok Bot acts as the signed-in member and hands logins, 2FA and payments back to the person.

How it identifies itself

SignalValueDocumented by the maker
user-agentGrokAgentNo, third-party only radar.cloudflare.com
web-bot-authSignature-agent key directory hosted on cursorusercontent.com (per Cloudflare Radar)No, third-party only radar.cloudflare.com
ip-listShared static egress IP ranges (not published; available from account team)Yes docs.x.ai
Allow it
On Cloudflare, Grok Bot is listed as a verified agent with intermediary access, so a verified-bot allow rule covers it. Outside Cloudflare, ask xAI's account team for its egress IP ranges.
Verify it
Cloudflare Radar lists a signature-agent key directory for Grok Bot, so a Web Bot Auth check through Cloudflare is the strongest test. xAI's egress IPs are shared by every Grok Bot customer, so they confirm the service, not the user.
Block it
Add a robots.txt group for GrokAgent with Disallow: /, the directive Cloudflare Radar lists, or block Grok Bot through Cloudflare's bot rules. xAI's own docs do not print the GrokAgent string, so pair robots.txt with a firewall rule.

Check your robots.txt for every agent at once

robots.txt lines for Grok Bot (xAI)

Copy the group you want for GrokAgent. The most specific matching group wins over your * group.

Let it in

User-agent: GrokAgent
Allow: /

Keep it out

User-agent: GrokAgent
Disallow: /

What we can and can't confirm

2 of the 3 signals above come from third parties, not from the maker, so treat them as hints. Matching a user agent string alone is easy to fake. A published IP list or a signed request is the stronger check, where the maker offers one.