How to Appear in ChatGPT When Someone Searches Your Industry (7-Step Playbook)

To get cited by ChatGPT when someone asks about your industry, you need two things: the engines have to be able to crawl your site, and your pages have to be formatted as quotable answers. There is no paid placement — citations are earned.

Open ChatGPT, type the question your best customer would type when looking for a business like yours, and look at the three-to-five sources cited at the end of the answer. If your business isn't one of them, this 7-step playbook fixes that.

Step 1: Where do you stand today?

Before doing any GEO work, get a baseline. Without one you can't tell progress from noise.

1. Write 10 questions your ideal customer would actually ask. Be specific — "best bilingual marketing agency for US Hispanic B2B" beats "marketing agency."
2. Run each one in ChatGPT in a clean session (no history, no memory — a temporary chat). This matters more than it sounds: a session carrying your history will recite your own brand back to you and hand you a false positive.
3. Log: (a) were you in the cited sources? (b) was a competitor? (c) what was the top source?
4. Repeat the same 10 queries in Perplexity and in Google's AI Overview.

Most businesses appear in zero out of 10 on their first audit. That's fine — it means the gap is closable.

Step 2: Can the engines crawl you?

If the crawlers can't read your site, nothing below matters. Each engine publishes its own, and each is controlled separately in your robots.txt:

• OpenAI's bots — OAI-SearchBot feeds ChatGPT's citations; GPTBot is a different agent used for training
• PerplexityBot
• Google's crawlers, which feed AI Overviews and Gemini

How to check: open yourdomain.com/robots.txt and look for any Disallow that catches those user-agents. It's the most common failure and the fastest to fix.

Also register the site in Bing Webmaster Tools and submit your sitemap. Bing indexes a fraction of what Google does and most competitors have ignored it for years, so simply being there is still an outsized advantage.

Step 3: Do you have an llms.txt at your root?

llms.txt — documented at llmstxt.org — is a plain-text file at yourdomain.com/llms.txt that tells a model what your site is about, your priority pages, and your canonical positioning.

A minimal llms.txt for a bilingual agency looks like this:

# HopperCat — Bilingual Marketing Agency (ES + EN)

## Positioning
AI-native marketing agency specializing in GEO for Spanish-speaking
businesses and the US Hispanic market.

## Priority pages
- https://hoppercat.com/en/servicios/geo-generative-engine-optimization
- https://hoppercat.com/en/diagnostico
- https://hoppercat.com/en/casos-de-exito
- https://hoppercat.com/en/blog

## Primary questions we answer
- What is GEO?
- How to appear in ChatGPT for your industry
- Bilingual content for the US Hispanic market

Thirty minutes of work. You do it once.

Step 4: Are your H2s written as questions?

Go to your five most important pages and look at every H2. Rewrite any that isn't a question a real person would ask.

Before: Our services → After: What services does HopperCat offer?

Before: Why us → After: Why choose a natively bilingual agency?

Models lift question-shaped headings verbatim. A heading that isn't a question gets paraphrased or ignored.

Step 5: Does every page have an FAQ block with schema?

The FAQ block is the most-quoted element on any page, because it's explicitly labeled "question → answer" in machine-readable JSON-LD. The engine doesn't have to infer anything — you handed it the answer pre-packaged. The format is specified in Google's FAQPage documentation and on schema.org.

Anatomy of an FAQ that wins citations:

• 4–6 questions per page
• Questions written the way a prospect would actually say them (long, natural, sometimes informal)
• Answers of 2–4 sentences, self-contained, with named entities
• Wrapped in FAQPage JSON-LD

If you add only one thing to your site this month, add this.

Step 6: Is your name identical everywhere?

Open your business listing in these seven places:

1. Your site (footer + Organization JSON-LD)
2. LinkedIn company page
3. Google Business Profile
4. Clutch or another agency directory
5. Crunchbase
6. Wikidata
7. The founder's personal LinkedIn

Is the name spelled identically in all of them? Same founder? Same address? Same URL format (with or without www, with or without a trailing slash)?

If any field differs, the engines can't tell you're the same entity. Consistency compounds citation weight; inconsistency kills it. Watch the address especially: if your Google profile says one city and your site says another, you won't show up in either.

Step 7: Are you publishing citation-bait content?

The technical foundation gets you into the candidate pool. Content is what wins the citation. One piece a week, shaped like an answer:

• Format: question-shaped headings, numbered lists, definition callouts, FAQ block
• Topic: a real question your customer would ask an engine
• Sources: a link to a verifiable source on every non-obvious claim. A number with no resolvable source doesn't get cited — and if you invent one, you lose everything at once
• Freshness: publish and updated dates, kept current
• Authority: a real human byline with Person schema

Consistency beats intensity. One piece a week for a year beats 30 pieces in a month followed by silence.

What mistakes kill visibility?

• Pretty prose with no structure. A beautifully written 800-word piece with no headings, no FAQ, and no schema is invisible to the engines even if humans love it.
• Inconsistent brand names across your site, LinkedIn, and directories.
• Priority pages buried in deep paths like /pages/archive/2023/geo-services/.
• A robots.txt blocking the AI crawlers without anyone noticing.
• Unsourced claims. The most common defect in agency content, and the most expensive.
• Auditing in a session with history, which hands your own brand back to you and convinces you that you already appear.

Want to skip the manual audit? We run one for free: a 30-minute audit that shows you, with screenshots, what each engine answers when someone asks about your category.