Skip to content
Work
About
Blog
Talk to me

Engine guide: Meta AI

How to get cited by Meta AI

Meta AI lives in the apps people already open dozens of times a day. It answers in a chat window, often on a phone, which shapes what kind of source it can use and show.

The short answer

Meta AI answers inside WhatsApp, Instagram, Messenger and Facebook, and can search the web for current questions. Meta documents two AI crawlers: Meta-ExternalAgent for indexing and training, which honours robots.txt, and Meta-ExternalFetcher for fetches a user triggers, which may bypass it. Citations show as compact source links, so short, mobile-readable answers win.

What ranked on 26 September 2026: Meta's crawler documentation for webmasters and a handful of bot directories rank for related queries. Meta has published little about which search provider backs web answers, so pages claiming a specific partner should be read with caution.

What is specific to Meta AI

Distribution is the story. The assistant sits inside messaging, so questions are short, local and practical, and the answer is read on a small screen.

  • Meta-ExternalAgent crawls for AI indexing and training, and blocking it is Meta's documented opt-out
  • Meta-ExternalFetcher fetches links users share or ask about, and Meta notes it may not follow robots.txt because the fetch is user-initiated
  • facebookexternalhit, the link-preview crawler, is separate and controls how your link card looks when shared
  • Questions in chat skew towards recommendations, places, recipes and quick comparisons, which suits local and consumer businesses
  • Answers are brief, so only one or two sources usually appear

Things that sound right and are not

  • ClaimBlocking Meta-ExternalAgent removes us from Meta AI.

    It limits indexing and training. User-triggered fetches may still read your page. Treat the block as a training decision, not as a visibility switch.

  • ClaimOur Instagram presence covers this.

    Social posts and web answers are different sources. Meta has not documented that public posts ground Meta AI web answers, so keep a clear website as well.

  • ClaimLink previews do not matter here.

    Answers in chat are compact, and a clean title, description and image from your Open Graph tags make the cited link readable. Broken previews look like spam.

Where to start

  1. 01Decide on Meta-ExternalAgent access in robots.txt and record the reasoning, since it is a training trade-off.
  2. 02Check your Open Graph title, description and image on key pages with Meta's sharing debugger.
  3. 03Write the short answers people ask in chat, such as opening hours, prices, delivery areas and how to book, near the top of the relevant page.
  4. 04Ask Meta AI your category questions inside WhatsApp and Instagram and note which sources it shows.
  5. 05Make sure the details match across your site, Google Business Profile and Facebook page, because mismatches are what an assistant repeats.

Questions

Does Meta AI cite websites at all?
When it searches, it can show source links. Many answers come from the model alone and carry none, so test with questions about current or local information.
Which country versions have web search?
Availability of Meta AI features differs by market and changes often. Test from the countries where your buyers are.
Is facebookexternalhit an AI crawler?
No. It fetches pages to build link previews when someone shares a URL. Blocking it breaks previews and does nothing for AI training.

Related

Doing this yourself and want a second opinion on what the assistants say about you? Tell me what you are seeing, or read the AEO Playbook for the whole framework.