How to get your site cited by ChatGPT, Perplexity, Google AI Overviews and Grok

To get cited by ChatGPT, Perplexity, Google AI Overviews, and Grok, make your identity machine-readable, publish answers crawlers can fetch in plain text, answer real questions with visible FAQ content (and matching FAQPage schema when the page already answers them), keep language and sitemap signals consistent, keep search crawlers allowed, ping the indexes that actually listen when pages change, and measure each assistant yourself. That is the short version.

I am Angel Baez, Applied AI Engineer and co-founder of Prepa IN and Recupera. I work on AI agents for growth: SEO, AI visibility and RevOps. The steps below are the practical checklist. angel-baez.com uses the same pattern as one small example, not as the story of this post. Nothing here invents traffic, clients, quotes, or product results.

1. Write one identity sentence and put it everywhere that matters

Pick one name, one role, and one niche. Put that sentence in:

• The homepage title and meta description
• About and resume (or equivalent) page titles
• Person or Organization JSON-LD (name, jobTitle, url, sameAs, and worksFor or founder when true)

If two pages disagree about who you are, answer engines have less reason to trust either one. Keep social sameAs URLs real and stable. Do not invent alternate names just to capture more queries.

2. Publish crawlable plain text for agents

Ship an /llms.txt that points to the pages you want cited. Prefer canonical URLs with the trailing slash your site actually uses. Treat it as a proposed agent map (llmstxt.org), not as a Google ranking lever. Google’s own AI features guidance says Search does not use special AI text files.

When the pages matter, add a full-text companion (for example /llms-full.txt) with the about page, resume, pillar content, and the niche posts you want models to quote. Link that file from llms.txt. Agents that can ingest plain text should not have to guess from HTML chrome alone.

3. Answer real questions with FAQ content and FAQPage schema

List the questions people already ask about your topic. Answer them in visible prose first, then emit FAQPage JSON-LD that matches that copy.

Good FAQ answers are specific, short enough to quote, and honest about what you do not know. Skip FAQ schema on pages that do not actually answer questions. Visible answers help quotability across assistants; the schema is a mirror, not a Google AI Overviews shortcut.

4. Keep bilingual and sitemap signals honest

If you publish more than one language:

• Use one hreflang set in the page head and the same set in the sitemap (en, es-MX, and x-default on this site, for example)
• Do not emit conflicting English codes for the same URL

For the sitemap, set lastmod from real content dates (post updatedDate / pubDate, resume last modified, and similar). A stamped “now” on every URL teaches crawlers nothing useful.

5. Keep search crawlers allowed, then ping the indexes that listen

For ChatGPT search specifically, OpenAI documents that sites should allow OAI-SearchBot in robots.txt if they want to appear in ChatGPT search answers. Blocking training crawlers is a separate choice from blocking search retrieval.

After deploy, submit new or updated sitemap URLs through IndexNow for participating engines such as Bing, and use Google’s own Search Console / URL tools for Google. Fail soft: a ping failure should not break the deploy.

IndexNow does not create a citation by itself, and it does not reach Google. It shortens the delay between a real content change and a fresher crawl on the engines that participate. Google AI Overviews still depend on Google’s index.

6. Make the page fast enough to earn the visit

Citations still send people to your site. Keep the largest contentful paint competitive on the pages you want shared: right-sized images, fonts that do not block first paint, and titles that make the source obvious once someone lands.

Speed is not a citation hack. It removes friction while identity and answers do the naming work.

7. Measure each assistant on the queries you care about

Check the same prompts over time in:

• ChatGPT (signed out and signed in, if both matter to you)
• Perplexity
• Google Search, including whether an AI Overview appears
• Grok

Log what you see: named or not, which URL was cited, whether an overview showed up. Do not invent the rest. Patterns take repeated checks, not one screenshot. For console-side instrumentation, see Tracking AI visibility with GSC and Bing Webmaster Tools.

xAI documents that Grok can search the web and return citations. It does not publish ranking factors or an official public crawler policy, so treat Grok the same way you treat any assistant you do not control: observe, log, and avoid invented rules.

A brief example on angel-baez.com

On angel-baez.com the same checklist shows up as Person JSON-LD, bilingual titles and meta, /llms.txt plus English /llms-full.txt, matching hreflang, IndexNow on production deploys, FAQPage on the pillar and on posts that need it, accurate sitemap lastmod, and a faster homepage. That is an implementation of the steps above, not a claim that every query now cites the site.

What to do first

1. Lock the identity sentence and Person JSON-LD.
2. Ship llms.txt and a full-text file for the pages you want quoted.
3. Add FAQ sections only where the prose already answers the question.
4. Align hreflang and sitemap lastmod with reality.
5. Allow search retrieval bots you care about, then ping IndexNow (and Google’s tools) on deploy.
6. Re-check ChatGPT, Perplexity, Google AI Overviews, and Grok on your target queries and write down what you observe.

For the longer system around agents and growth work, read AI agents for growth: SEO, AI visibility and RevOps.

También disponible en español.