xSeek/Docs/Llama
Back to documentation

Llama user agents, explained.

Meta operates two user agents for Llama and Meta AI. Understand what each does, control access, and optimize for Meta's AI surfaces.

Llama · MetaUpdated Apr 2026~5 min read2 user agents
TL;DR

Meta ships Meta-ExternalAgent (formerly FacebookBot/MetaAIBot) for general AI operations and Meta-ExternalFetcher for user-initiated requests. Both respect robots.txt.

Overview

Meta's Llama model is used across Meta AI (WhatsApp, Instagram, Facebook) and the meta.ai assistant. The crawlers behind these surfaces visit the web either to maintain Meta AI's knowledge or to fulfill a specific user request.

Both agents honor robots.txt. Block them individually depending on whether you want to be reachable in Meta AI responses.

How to identify the agents

2 crawlers

Llama identifies itself with specific user-agent strings. Verify each through a published IP range.

MA

Meta-ExternalAgent

Meta AI crawler · general
TrainingKnowledge
meta-externalagent/1.1 (+https://developers.facebook.com/docs/sharing/webmasters/crawler)

Meta's general AI crawler. Gathers content that may be used to build and improve Meta AI products (including Llama). Previously observed as FacebookBot or MetaAIBot.

VerifiableMeta docs
MF

Meta-ExternalFetcher

User-initiated fetches · on-demand
User-triggered
meta-externalfetcher/1.1 (+https://developers.facebook.com/docs/sharing/webmasters/crawler)

Fetches URLs when Meta AI users ask for content from a specific page. On-demand, not used for background training crawls.

VerifiableMeta docs

How Llama accesses your content

3 access patterns, mapped to the agents above.

1

AI knowledge

Meta-ExternalAgent crawls content that feeds Meta AI products.
2

Live fetches

Meta-ExternalFetcher pulls URLs when a Meta AI user asks for a page.
3

Ecosystem

Meta AI lives inside WhatsApp, Instagram, Facebook, and meta.ai — all share this infrastructure.

Control Llama's access

robots.txt

You control each agent independently. Common configurations below.

Allow on-demand, block trainingRecommended

Stay reachable in Meta AI answers, stay out of training.
# Allow user fetches, block general crawler User-agent: meta-externalagent Disallow: / User-agent: meta-externalfetcher Allow: /

Block all Meta AI accessRestrictive

Removes your content from every Meta AI surface.
# Block every Meta AI crawler User-agent: meta-externalagent Disallow: / User-agent: meta-externalfetcher Disallow: /

Section-level control

Protect sensitive paths while staying citable.
# Meta AI: public only User-agent: meta-externalagent Disallow: /private/ Disallow: /members-only/ User-agent: meta-externalfetcher Disallow: /private/

Optimize content for Llama.

Meta AI runs on Llama. Five rules that help your content surface in answers.

Semantic HTML. Proper heading hierarchy, lists, and tables for clean extraction.
Render without JavaScript. Meta's crawlers read HTML — content gated by JS often won't be parsed.
Mobile-first. Meta AI is primarily accessed through mobile apps. Layouts that break on mobile lose citations.
Schema markup. Article, FAQPage, and HowTo structured data help Meta verify authority.
Factual accuracy. Meta AI emphasizes sourcing — make claims verifiable, cite your references.

Track Llama visits with xSeek.

Track every Meta AI visit.

Monitor Meta-ExternalAgent and Meta-ExternalFetcher hits. See which pages feed Meta AI answers across WhatsApp, Instagram, and meta.ai.

Start free →

Frequently asked questions

Meta's Llama and Meta AI surfaces use meta-externalagent/1.1 (general crawler) and meta-externalfetcher/1.1 (user-initiated fetches). Older logs may also show FacebookBot or MetaAIBot.
Source: Meta's official crawlers documentation.
PreviousBing AI user agentsNextBaidu ERNIE user agents