Call · 30 min
MarketingMattia EspositoUpdated 10 November 202614 min read

How to show up in Meta AI as a business. The crawler that decides whether WhatsApp, Instagram and Facebook cite your website.

Meta AI sits in the search bar of WhatsApp, Instagram and Facebook, and the guides out there explain how to use it. How a business gets cited is still uncovered. Here is the crawler to let in and the checklist, in Word and PDF.

In brief

To show up in Meta AI answers, your site has to let in meta-webindexer, the crawler Meta uses to improve search results in its assistant. Meta writes that allowing it in robots.txt helps it cite and link to your content.

Training is a separate choice. In Meta’s documentation, updated on 21 May 2026, meta-externalagent is also used to train foundation models: you can block it and still be cited.

Meta AI has been available in Italy since March 2025, as a chat with web access inside WhatsApp, Instagram, Facebook and Messenger.

The checklist can be downloaded below, in Word and PDF, with no email required.

Written by: Mattia Esposito, AI and marketing consultant, founder of Itria AI. From Puglia, based in Turin since 2023, he works remotely in Italy and abroad. LinkedIn profile.

The “Can Meta AI see my site?” checklist, to download

A sheet with Meta’s five crawlers, ten lines to fill in and the robots.txt lines, in Word to complete and in PDF to print. You need access to robots.txt, the hosting logs, the Facebook Page, the Instagram profile and a phone with WhatsApp. It takes half an hour.

The Can Meta AI see my site checklist, with Meta’s five crawlers and the ten checks for showing up in Meta AI answers
FileContentsLink
Wordto fill in, in Italian

Meta’s five crawlers, the ten lines with where to look and a result column, the robots.txt lines.

come-comparire-su-meta-ai-scheda.docx

PDFto print, in Italian

The same checklist, to tick off with a pen.

come-comparire-su-meta-ai-scheda.pdf

N.What to checkWhere to look
1

meta-webindexer is not blocked, either by name or by a blanket rule.

yoursite.com/robots.txt

2

You have made a separate decision on meta-externalagent, the training crawler.

yoursite.com/robots.txt

3

The CDN or firewall does not stop Meta’s crawlers before they reach robots.txt.

CDN or hosting dashboard

4

meta-webindexer has visited in the last 30 days.

Hosting logs

5

The Open Graph title, description and image are present within the first MB of the page.

Meta’s Sharing Debugger

6

Text, prices, opening hours and contact details sit in the HTML, readable without JavaScript.

Page source (Ctrl+U)

7

Name, category, address, opening hours and website match across the site, the Facebook Page and the Instagram profile.

The two profiles, side by side with the site

8

After every robots.txt change, 24 hours have passed before you test.

Calendar

9

Three buying questions to Meta AI in WhatsApp, in your customers’ language: does your site appear among the sources?

Chat with Meta AI

10

Repeat the test every month with the same questions, noting who gets cited instead of you.

This checklist

Meta’s apps are where I learnt how attention is won. At 14 I started @artartism, an Instagram page of paintings, and in seven months it collected 632,288 organic likes. In August 2026, while preparing Itria’s first Meta campaign, I read around a hundred ads in a week to see what was running in the feed.

With Meta AI, what decides things for a website lives somewhere else. It lives on a technical page for developers, updated on 21 May 2026, which I could not find quoted in any Italian guide to Meta AI. I read it in full, in the Italian version, and the rest of this guide sets out what it says, alongside other Meta sources and the Italian Competition Authority.

On 7 August ChatGPT replied that it could not find a Mattia Esposito linked to itria.io. Between 11 October and 7 November 2026 the site’s pages appeared 3,818 times in Google and Microsoft AI answers. For Meta AI I have no comparable count: the Meta pages I read offer no report of how often a site is cited, so the checklist ends with a manual test.

What is Meta AI, and where does it answer?

Meta AI is Meta’s assistant inside WhatsApp, Instagram, Facebook and Messenger: you open it from the blue circle, from the search bar, or by typing @MetaAI in a WhatsApp group. Meta has described it this way since 2024: “You can use Meta AI in feed, chats, search and more across our apps”.

For a business, where it appears matters. A customer looking for a supplier, a restaurant or a product inside WhatsApp or Instagram asks the question there, without opening a search engine. In Italy, according to the Italian Competition Authority (AGCM), Meta AI has been pre-installed in WhatsApp since March 2025, placed prominently and built into the search bar.

In July 2025 the Authority opened an investigation into an alleged abuse of a dominant position, case A576, in close cooperation with the European Commission. Its press release says Meta “ha deciso di pre-installare il proprio servizio di intelligenza artificiale sull’app Whatsapp”, having decided to pre-install its AI service in WhatsApp. For anyone selling, the practical upshot is simple: the assistant is already on your customers’ phones.

What is Meta AI based on, and where do its answers come from?

When it searches the web, Meta AI reads pages collected by meta-webindexer, the crawler Meta declares for the quality of search results in its assistant. At launch, in the United States in September 2023, Meta wrote that real-time information came through a search partnership with Bing.

The April 2024 announcement, the Llama 3 one, mentions real-time information without naming an engine. The Meta pages read on 10 November 2026 name no search partner. What they document is Meta’s own crawler, so the accurate statement today is that Meta AI has a door of its own, and your site needs to keep it open.

In its Help Center, Meta explains that when you chat with Meta AI it may use sources and links from the internet to answer, and that readers can report them. This is where a business wins or loses the customer: the source under the answer links either to your site or to a competitor’s.

How a page reaches the sources of a Meta AI answer: meta-webindexer reads it, Meta AI search stores it, the answer in WhatsApp, Instagram or Facebook cites it. Training goes through meta-externalagent, a separate branch The pagetext in the HTML meta-webindexerrobots.txt and CDN The searchin Meta AI The answerWhatsApp, Instagram The sourcesunder the answer meta-externalagenttraining, set apart lines 5-7lines 1, 3, 4line 8 line 9lines 9-10line 2
The path of a page to the sources of a Meta AI answer, with the training branch decided separately, and which checklist lines act on each step.

The path is the same as for other assistants, with a different name on the door. How ChatGPT, Copilot, Gemini, Perplexity and Claude read the web is covered in the guide on how to show up in ChatGPT; here we go deep on Meta.

Meta’s crawlers: which one to allow to get cited

Meta lists five crawlers, and for citations what counts is meta-webindexer. The documentation puts it in one sentence: “Allowing Meta-WebIndexer in your robots.txt file helps us cite and link to your content in Meta AI”. The other four have different jobs.

Name in robots.txtWhat it does, according to MetaWhat to decide
meta-webindexer

Improves search results in Meta AI and lets Meta cite and link to content in its answers.

Let it in

meta-externalagent

Training foundation AI models and improving products by indexing content directly.

Your call, made separately

meta-externalads

Advertising and other business-related products and services.

Your call

facebookexternalhit

Reads pages shared on Facebook, Instagram or Messenger to build the link preview. For security checks it may ignore robots.txt.

Let it in

meta-externalfetcher

Fetches individual links at a user’s request, including for AI that browses sites on the user’s behalf. It may ignore robots.txt.

Firewall, if needed

A robots.txt that does not name these crawlers already lets them in. The lines matter when there is a blanket block higher up, or when you want to stay in citations and opt out of training.

In robots.txt: in for citations, out of training

User-agent: meta-webindexer
Allow: /

User-agent: meta-externalagent
Disallow: /

Meta says robots.txt changes can take up to 24 hours to apply. It also says it prefers robots.txt to non-standard formats such as NoAI tags, and gives an address for crawler questions, [email protected]. Every AI crawler name, Meta’s included, is in the guide to robots.txt for AI crawlers and in the glossary entry on AI crawlers.

Before robots.txt there is often another door. A CDN or firewall with bot rules can stop meta-webindexer before it reads the file, and the visit leaves no trace in your logs. The check sits in your provider’s dashboard: find the AI crawler rules and make sure Meta’s search crawler gets through.

Does Meta AI use my website for training?

It can, if your site lets in meta-externalagent, which Meta describes as crawling for uses such as training foundation AI models. One Disallow line with that name takes you out, and citations stay with meta-webindexer.

In the European Union that line also carries legal weight. Directive (EU) 2019/790, Article 4(3), allows text and data mining of online content unless the rightsholder has reserved it in a machine-readable way, and a robots.txt file qualifies. The full picture, AI Act included, is in the guide to robots.txt for AI crawlers.

meta-externalfetcher stays outside robots.txt. Meta writes that it fetches individual links at a user’s request and can therefore bypass the file’s rules. If you want to keep it out of part of your site, a firewall is the way, bearing in mind you would also block a customer who asked Meta AI to open your page.

Link previews: Open Graph and facebookexternalhit

When someone shares a link to your site on Facebook, Instagram or Messenger, facebookexternalhit reads the page and takes its title, description and image. If it fails, the link arrives bare, and this becomes the first impression a customer gets of you inside Meta’s apps.

Meta puts the requirements in writing. Your server must use gzip or deflate compression. Open Graph properties must sit within the first MB of the page, or they get cut off. The content must be readable within a few seconds. If a page has changed, the Sharing Debugger forces a fresh read.

These are the same basics every assistant needs: text and data in the HTML, straight away. AI crawlers generally read a page as the server sends it, and what happens to content loaded by a script is explained in the guide on JavaScript and AI crawlers. For the data that describes your business to machines, see the page on structured data for AI.

Meta AI for small business: getting cited, or letting it answer for you

Searches for “Meta AI business” mix two different things: showing up in Meta AI answers when a customer asks, and using a Meta agent that answers your customers for you. The first runs through your website and the crawler. The second is a business product with its own rules.

The second has a name, Meta Business Agent. According to TechCrunch, on 3 June 2026 Meta made it available worldwide inside WhatsApp, after nearly two years of customer-support trials in countries such as India and Mexico; it will be included in some tiers of the WhatsApp Business Premium subscription. It answers your customers, and it leaves untouched what Meta AI tells people who have never heard of you.

Then there is Muse, the personal agent Meta launched in September 2026. Meta says it is available in the United States and Canada, and on 29 September it introduced Muse for Small Business, which connects Instagram professional account analytics, Facebook Pages and ad accounts. Muse launched with access to the entire Shopify catalogue, and on 24 September Meta announced it is coming to its AI glasses.

An agent that buys and books on a user’s behalf comes into websites through meta-externalfetcher. For an Italian business today this signals where things are heading: pages an agent can read and use are the subject of the guide to websites ready for AI agents. How Meta uses Meta AI chats for advertising is a separate matter, covered in Meta AI, chats and ads.

Meta AI in Italy, the EU, the UK and the US: what is available

Meta AI has been in Italy since March 2025, as a text chat with web access. On 19 March 2025 Meta announced its arrival in 41 European countries, the EU ones included, starting with chat in six European languages; personalisation and memory remained US features. The rules for your website are the same everywhere.

WhereMeta AI and the rules for your siteDate and source
Italy

Chat with web access in WhatsApp, Instagram, Facebook and Messenger, pre-installed in WhatsApp; a competition investigation into an alleged abuse of a dominant position.

March 2025; AGCM A576, July 2025

European Union

Chat in six languages across 41 European countries. A Disallow for meta-externalagent counts as a reservation of rights over text and data mining.

Meta, 19 March 2025; Dir. 2019/790, Art. 4(3)

United Kingdom

Same crawlers, same robots.txt. The text and data mining exception covers non-commercial research only.

Meta, 21 May 2026; CDPA 1988, s. 29A

United States

Meta AI since 2023, with personalisation and memory; Muse and Muse for Small Business available, along with Canada.

Meta, 27 September 2023 and 29 September 2026

For an Italian business the practical difference comes down to one thing: here Meta AI answers and cites, while the agents that shop on a user’s behalf have been announced elsewhere. The doors to open on your site are the same as in the countries where Muse is already live.

Three cases measured on itria.io

The email address that wasn’t there. Cloudflare’s obfuscation took the address out of the HTML and put it back with a script: for an assistant, the contact page had no contact details. The case, and the same flaw on 14 of 121 producers’ sites, is in the piece on the email address AI assistants can’t see.

From one page to five. On the morning of 25 September 2026, Claude’s web search knew a single page of itria.io; by the evening, after I asked Brave to recrawl seven, it knew five. Before and after in Brave and Claude.

The pages that didn’t exist. Without a real 404 page, five URLs that had never existed were getting impressions in first position: the case of the phantom URLs.

The result. Between 11 October and 7 November 2026 Bing’s AI Performance report counted 268 citations of itria.io pages in Copilot, and Search Console 3,550 impressions in AI Overviews and AI Mode.

How to check if Meta AI cites your website

With two checks you can run yourself: the server logs, to see whether meta-webindexer visits, and a test in WhatsApp, to see whether your site appears among the sources. The first tells you the door is open, the second whether there is something inside Meta AI chooses to cite.

In the logs, filter for the name meta-webindexer and note the date of the last visit. For the test, send Meta AI three questions a customer would ask, in their language, along the lines of “who makes…” or “where can I buy… in Turin”, and open the list of sources under the answer. Next to it, write down who gets cited instead of you.

On my own site I learnt that the name cited is rarely the one with the best explanation. I wrote it on 21 September: “AI cites the body that issues the rule or the firm that does the job. Never the explainer.” It holds here too: a page with the product, the price, the area and the documents has more reasons to end up among the sources than a page describing the sector. The full post (in Italian).

Repeat the test every month with the same questions, because answers change. To track Meta AI, ChatGPT, Copilot and AI Overviews in a single sheet, see the guide on how to measure GEO, with the monthly Excel file.

What Meta does not say

The Meta pages read for this guide contain no list of factors that get a site cited by Meta AI. Reviews, followers or activity on your Facebook Page may count, but Meta does not say so, and anyone selling them as a rule does so without a source.

Italian usage figures are missing too: Meta’s public figures on Meta AI users are global. What you can control is written down: the right crawler let in, a readable page, business details that match on the site and in the apps, and a test you repeat.

How to build this for a real business, with the customer questions worth holding, is set out on the page about presence in search engines and AI assistants. The other guides in the series are in the marketing for SMEs index.

Questions and answers

How do I get my business to show up in Meta AI?

Let meta-webindexer into your site: it is the crawler Meta uses to improve search results in Meta AI. Meta writes that allowing it in robots.txt helps it cite and link to your content in its answers.

After that the page does the work: text readable in the HTML, facts with a source, and the same business name and details on the website, the Facebook Page and the Instagram profile. There is no form to sign up to Meta AI.

If I block meta-externalagent, do I disappear from Meta AI?

No. According to Meta, meta-externalagent is used among other things to train foundation models, while citations come through meta-webindexer. You set the two names separately in robots.txt, and Meta says a change can take up to 24 hours to apply.

Does Meta AI use Google or Bing to search the web?

At launch, in September 2023, Meta announced a search partnership with Bing. The Meta pages read on 10 November 2026 name no partner search engine, and Meta documents its own crawler for Meta AI search, meta-webindexer.

Is Meta AI available in Italy?

In Italy, yes. On 19 March 2025 Meta brought it to 41 European countries, the EU included, as a text chat with access to information from the web inside WhatsApp, Instagram, Facebook and Messenger. Muse, the personal agent launched in September 2026, is available in the United States and Canada, according to Meta.

How can I tell if Meta AI cites my website?

With two checks. In your server logs, look for visits from meta-webindexer. In WhatsApp, ask Meta AI the questions a customer would ask and see whether your site appears among the sources of the answer. Note the date, the question and who gets cited, and repeat the test every month.

Notes on sources

  1. Meta for Developers, Meta Web Crawlers, Italian version updated on 21 May 2026: the five crawlers, the Meta-WebIndexer sentence quoted above, robots.txt timing, NoAI tags, preview requirements. Read as raw text on 10 November 2026; the quotation follows the English wording of the same page.
  2. Meta Newsroom, Introducing New AI Experiences, 27 September 2023 (the Bing partnership) and Meta AI, Built With Llama 3, 18 April 2024, source of the quotation about the apps.
  3. Meta Newsroom, Europe, Meet Your Newest Assistant: Meta AI, 19 March 2025: 41 European countries, six languages, web access, features kept in the United States.
  4. Meta Help Center, Report a source or link used in a Meta AI response, read on 10 November 2026.
  5. Italian Competition Authority (AGCM), press release on case A576, July 2025, source of the quotation.
  6. Meta Newsroom, Muse for Small Business, 29 September 2026, and The Biggest News From Connect 2026, 24 September 2026.
  7. TechCrunch, Meta’s AI agent for WhatsApp Business is now available globally, 3 June 2026: a news outlet, not a Meta source.
  8. Directive (EU) 2019/790, Art. 4(3); Copyright, Designs and Patents Act 1988, s. 29A.
  9. The itria.io figures come from Search Console (generative AI report) and Bing Webmaster Tools, window 11 October to 7 November 2026.
·Who does it for you

Letting in the right crawler is the first step. Becoming the name Meta AI gives instead of your competitors is the work that follows. Positioning is built starting from the market.

Mattia Esposito, who wrote this guide, also works with Itria on positioning: in-depth market analysis to find where competitors are exposed and take that ground. He did it on his own site first. In August, asked who had founded itria.io, ChatGPT couldn’t find him. From 11 October to 7 November 2026 itria.io pages appeared 13,803 times in Google and Bing results and 3,818 times in AI answers: Google’s AI Overviews and AI Mode, and Microsoft Copilot.