Platform guide

Grok, Meta AI & DeepSeek SEO: How They Find and Cite Sources

Three big AI assistants, three very different stories. Here's what each company has actually documented, and what it hasn't.

Facts checked on October 19, 2026. Every claim links to its source at the bottom.

Worth a check-up, not your first priority

By their own numbers, these assistants are big. Grok had about 117 million monthly users of its AI features in March 2026, according to an SEC filing. Meta says Meta AI has about 1 billion monthly users across its apps (May 2025).

Clicks are another story. StatCounter's September 2026 chatbot referral data, measured on the websites that run its tracking code, doesn't list Grok or Meta AI at all, so nobody publishes how many clicks they send. DeepSeek sent 0.02%.

The three assistants at a glance
Assistant	Users (company figure)	Share of chatbot visits to websites (StatCounter, Sep 2026)	Documented crawler
Grok	About 117 million a month (March 2026)	Not listed	None
Meta AI	About 1 billion a month (May 2025)	Not listed	Yes: meta-webindexer
DeepSeek	No official figure found	0.02%	None

So check the basics below once. Then spend your time on ChatGPT and Gemini, where most of the chatbot referrals StatCounter tracks come from.

Meta AI: the only one with a search crawler you can control

Checked on October 19, 2026

Where its answers come from

Meta runs a crawler called Meta-WebIndexer. Meta says it "navigates the web to improve Meta AI search result quality". In 2024, news reports said Meta AI also pulled live results from Google and Bing, while Meta's own post only said "from across the web". Whether that's still true today is unknown.

Can you control it?

Meta documents several crawlers, each with its own job. You can treat them separately in robots.txt.

meta-webindexer: builds Meta AI search. Allowing it "helps us cite and link to your content", says Meta.
meta-externalagent: trains AI models and indexes content for other Meta products.
meta-externalfetcher: fetches a page when a user asks for it. Meta says it may ignore robots.txt.
facebookexternalhit: makes link previews. It might skip robots.txt for security checks.

So you can say no to training and still stay in Meta AI search:

User-agent: meta-externalagent
Disallow: /

User-agent: meta-webindexer
Allow: /

Blocking meta-externalagent also stops its indexing for other Meta products. Changes can take up to 24 hours, because Meta caches robots.txt that long.

Does it link to you?

Yes, according to Meta: allowing Meta-WebIndexer helps Meta AI cite and link to your content.

What nobody has documented

Which search partners Meta AI uses today. And how often its answers, mostly given inside WhatsApp, Instagram, Facebook and Messenger, lead to a click.

Grok: web search plus X, no documented crawler

Checked on October 19, 2026

Where its answers come from

Grok combines web search with posts from X, according to an SEC filing. xAI's developer docs offer a web search tool and an X search tool. They don't say which search index or crawler sits behind them.

One thing is unique to Grok: it reads X. So being discussed on X plausibly helps. That's our reasoning, not something xAI has said.

Can you control it?

xAI publishes no crawler name and no robots.txt instructions. Names like "GrokBot" do show up in third-party bot lists, but those lists disagree and xAI hasn't confirmed any of them. One tester (StackFox, February 2026) reported that Grok fetches pages with ordinary browser user agents, so a robots.txt rule may not match anything.

Does it link to you?

In xAI's API, yes: search results come back with a list of citations. How the Grok app shows sources isn't documented.

What nobody has documented

Which search engine Grok uses, whether it has its own crawler, and how it picks sources.

DeepSeek: two search systems, no documented crawler

Checked on October 19, 2026

Where its answers come from

DeepSeek's privacy policy (updated 10 February 2026) says it uses third-party APIs for search and shares your search keywords with them. Its newer Search API runs on "DeepSeek's in-house search engine". Which of the two powers the chat app today isn't documented.

Can you control it?

DeepSeek documents no crawler and no robots.txt rules. "DeepSeekBot" only shows up in third-party lists. Its Fetch API says fetching is "subject to the target site's availability and access policies", without saying how.

Does it link to you?

Its Search API returns a URL, title and snippet for every result, plus an "authority" label. So developers get the sources; how the chat app shows them isn't documented.

Rules in Europe

Data: DeepSeek's privacy policy says it stores personal data in China.
Italy: on 30 January 2025, the data protection authority ordered a definitive restriction on processing Italian users' data. We found no sign it has been lifted.
Netherlands: national civil servants may not use DeepSeek for their work since February 2025 (news report). For everyone else it's legal; the Dutch Data Protection Authority only urged caution.

What nobody has documented

Which search engine the app uses, whether DeepSeek crawls the web itself, and whether the API's "authority" label affects which sources get cited.

What works for all three

None of these companies publishes a ranking guide. But like every AI assistant, they need to find your page and understand it fast.

Be in the big search indexes, with nothing blocking their crawlers.
Keep your main text in the HTML itself, not only after JavaScript runs.
Answer the question near the top, in plain sentences.
Name your sources and add dates, so a claim can be checked.

For the full approach, read these next: LLM SEO guide · Multi-platform AI strategy

Frequently asked questions

Can I block Grok?

Not reliably. xAI documents no crawler name, so there's no official robots.txt rule. Rules for names from third-party lists may not match anything.

Can I block Meta AI training but stay in Meta AI search?

Yes. Disallow meta-externalagent and allow meta-webindexer in robots.txt. Meta says the change can take up to 24 hours.

Which search engine does DeepSeek use?

Its Search API uses DeepSeek's own search engine. Its privacy policy says the app uses third-party search APIs. Which one powers the app today isn't documented.

Is DeepSeek banned in the Netherlands?

No. According to news reports, only national civil servants may not use it for their work. For consumers it's legal.

Should I optimize for these before ChatGPT or Gemini?

No. In StatCounter's September 2026 data, ChatGPT and Gemini together sent over 90% of the chatbot referrals StatCounter tracks. DeepSeek sent 0.02%, and Grok and Meta AI aren't measured at all.

Does GEO-Score track citations in Grok, Meta AI or DeepSeek?

No. GEO-Score checks whether a page is easy for AI to read and quote. It doesn't track citations in any assistant.

Keep reading

Sources

Every fact on this page comes from one of these pages. All checked on October 19, 2026.

Is your page easy for AI to quote?

GEO-Score checks whether a page is clear, well built and easy for AI to quote. It doesn't track citations in Grok, Meta AI or DeepSeek.

Check your page

Worth a check-up, not your first priority

The three assistants at a glance
Assistant	Users (company figure)	Share of chatbot visits to websites (StatCounter, Sep 2026)	Documented crawler
Grok	About 117 million a month (March 2026)	Not listed	None
Meta AI	About 1 billion a month (May 2025)	Not listed	Yes: meta-webindexer
DeepSeek	No official figure found	0.02%	None

So check the basics below once. Then spend your time on ChatGPT and Gemini, where most of the chatbot referrals StatCounter tracks come from.

Meta AI: the only one with a search crawler you can control

Checked on October 19, 2026

Where its answers come from

Can you control it?

Meta documents several crawlers, each with its own job. You can treat them separately in robots.txt.

meta-webindexer: builds Meta AI search. Allowing it "helps us cite and link to your content", says Meta.
meta-externalagent: trains AI models and indexes content for other Meta products.
meta-externalfetcher: fetches a page when a user asks for it. Meta says it may ignore robots.txt.
facebookexternalhit: makes link previews. It might skip robots.txt for security checks.

So you can say no to training and still stay in Meta AI search:

User-agent: meta-externalagent
Disallow: /

User-agent: meta-webindexer
Allow: /

Blocking meta-externalagent also stops its indexing for other Meta products. Changes can take up to 24 hours, because Meta caches robots.txt that long.

Does it link to you?

Yes, according to Meta: allowing Meta-WebIndexer helps Meta AI cite and link to your content.

What nobody has documented

Which search partners Meta AI uses today. And how often its answers, mostly given inside WhatsApp, Instagram, Facebook and Messenger, lead to a click.

Grok: web search plus X, no documented crawler

Checked on October 19, 2026

Where its answers come from

One thing is unique to Grok: it reads X. So being discussed on X plausibly helps. That's our reasoning, not something xAI has said.

Can you control it?

Does it link to you?

In xAI's API, yes: search results come back with a list of citations. How the Grok app shows sources isn't documented.

What nobody has documented

Which search engine Grok uses, whether it has its own crawler, and how it picks sources.

DeepSeek: two search systems, no documented crawler

Checked on October 19, 2026

Where its answers come from

Can you control it?

Does it link to you?

Its Search API returns a URL, title and snippet for every result, plus an "authority" label. So developers get the sources; how the chat app shows them isn't documented.

Rules in Europe

Data: DeepSeek's privacy policy says it stores personal data in China.
Italy: on 30 January 2025, the data protection authority ordered a definitive restriction on processing Italian users' data. We found no sign it has been lifted.
Netherlands: national civil servants may not use DeepSeek for their work since February 2025 (news report). For everyone else it's legal; the Dutch Data Protection Authority only urged caution.

What nobody has documented

Which search engine the app uses, whether DeepSeek crawls the web itself, and whether the API's "authority" label affects which sources get cited.

What works for all three

None of these companies publishes a ranking guide. But like every AI assistant, they need to find your page and understand it fast.

Be in the big search indexes, with nothing blocking their crawlers.
Keep your main text in the HTML itself, not only after JavaScript runs.
Answer the question near the top, in plain sentences.
Name your sources and add dates, so a claim can be checked.

For the full approach, read these next: LLM SEO guide · Multi-platform AI strategy

Frequently asked questions

Can I block Grok?

Not reliably. xAI documents no crawler name, so there's no official robots.txt rule. Rules for names from third-party lists may not match anything.

Can I block Meta AI training but stay in Meta AI search?

Yes. Disallow meta-externalagent and allow meta-webindexer in robots.txt. Meta says the change can take up to 24 hours.

Which search engine does DeepSeek use?

Its Search API uses DeepSeek's own search engine. Its privacy policy says the app uses third-party search APIs. Which one powers the app today isn't documented.

Is DeepSeek banned in the Netherlands?

No. According to news reports, only national civil servants may not use it for their work. For consumers it's legal.

Should I optimize for these before ChatGPT or Gemini?

No. In StatCounter's September 2026 data, ChatGPT and Gemini together sent over 90% of the chatbot referrals StatCounter tracks. DeepSeek sent 0.02%, and Grok and Meta AI aren't measured at all.

Does GEO-Score track citations in Grok, Meta AI or DeepSeek?

No. GEO-Score checks whether a page is easy for AI to read and quote. It doesn't track citations in any assistant.