GeorgeBot

Every GeorgeBot visit starts with a person pressing a button in Baseline Labs. It reads the site that person named, measures what AI crawlers and search engines can see there, and reports back. When a tool compares a business with its competitors, it also reads the competitors' homepages, and the assistant can read a page it is asked about.

• No more than one new request a second to a site while it reads pages.
• One request every 5 seconds when the person running the scan turns on Reduce requests.
• At most 150 pages in a Site Pulse scan, and at most 250 in a full Schema Audit or AI Vision scan.
• After reading pages, a Site Pulse scan also checks a few files (robots.txt, the sitemap, llms.txt) and up to 5 images on each page it read.
• The same daily cap of 1,000 requests per site as BaselineLabsAIScraper, shared between the two.
• On a 429 it honours Retry-After: it waits as asked, up to a minute, once, and then carries on.

Requests come only from four static addresses: 94.130.130.190 and 204.168.236.25 (IPv4), 2a01:4f8:13b:3968::2 and 2a01:4f9:c014:c18f::1 (IPv6). A request using this name from anywhere else is not us.

On the site a scan was run for, GeorgeBot does not read robots.txt, because a person asked for that scan by name. On any other site, such as a competitor's homepage or a page from a search result, it reads robots.txt first and stays out where it is disallowed:

User-agent: GeorgeBot
Disallow: /

That keeps it off a site when somebody else's comparison names it. To refuse scans as well, block the user agent GeorgeBot at the server or firewall.

To let it in instead, see the next section.

A firewall that blocks GeorgeBot is doing its job: it has never seen us before. When that happens, Baseline Labs shows you which protection answered and does not score the scan or charge for it. One allow rule lets us in, and nothing else about your protection changes. Match both the name and the four addresses, because anyone can send a name.

Cloudflare

1. Open your domain and go to Security, then Security rules. Choose Create rule, then Custom rules.
2. Name it Allow Baseline Labs, choose Edit expression and paste:

   (http.user_agent contains "BaselineLabs" and ip.src in {94.130.130.190 204.168.236.25 2a01:4f8:13b:3968::2 2a01:4f9:c014:c18f::1})

3. Set the action to Skip. Tick All remaining custom rules and All Super Bot Fight Mode rules, then under More components to skip tick Browser Integrity Check and Security Level. Security Level covers I am Under Attack mode.
4. Choose Deploy and drag the rule to the top of the list.

On the Free plan, Bot Fight Mode ignores custom rules entirely. If it is on, either switch it off under Security, then Settings, or add the four addresses as IP access rules with the action Allow, which Cloudflare checks before Bot Fight Mode.

AWS WAF

1. Open the web ACL in front of the site. For a site on CloudFront it lives under the Global (CloudFront) region.
2. Choose Rules, then Add rules, then Add my own rules and rule groups, and use the rule builder.
3. Inspect Single header named user-agent, match type Contains string, value BaselineLabs, and add a second statement that the source address is in an IP set holding the four addresses. Set the action to Allow.
4. Give it a lower priority number than the Bot Control rule group and the common rule set, so it runs first.

Other hosts and plugins

• Vercel: Firewall, then Configure, add a rule on User Agent containing BaselineLabs with the action Bypass.
• Kinsta: Bot protection, then Exceptions, add the user agent and choose Allow.
• WP Engine: Web Rules, an Access rule that allows the User-Agent header containing BaselineLabs.
• Wordfence and Sucuri allow by address: add the four addresses to the allowlist.
• Anything else: ask whoever runs your hosting to allow requests whose User-Agent contains BaselineLabs and that come from the four addresses.

Then press Test again on the blocked scan in Baseline Labs. It makes one free request to check the rule, and a new rule can take a minute to apply.

Operator documentation: https://baselinelabs.ai/bots/georgebot

Requests come only from four static addresses: 94.130.130.190 and 204.168.236.25 (IPv4), 2a01:4f8:13b:3968::2 and 2a01:4f9:c014:c18f::1 (IPv6). A request using this name from anywhere else is not us.

A user agent is a claim, not a proof: any client can send any name. Where an operator publishes IP ranges, a reverse lookup is the only way to tell a real visit from an impersonation.