How to appear in ChatGPT: 2026 technical guide

Appearing in ChatGPT when a potential customer asks about your category isn't magic. It's engineering, content, and discipline applied to a system that can be understood. The problem is that almost no one is applying it well in Spanish, so the space remains open for those who do.

This post is the concrete technical guide. If you want to understand GEO in depth, first read What is GEO (Generative Engine Optimization): 2026 guide. If you want to know how to be inside ChatGPT's response this week, keep reading.

How ChatGPT decides which sources to cite

There are two citation modes in ChatGPT worth distinguishing:

1. Real-time browsing: when ChatGPT activates web search (with Bing as base engine) and cites the pages it finds to answer. Behavior here is similar to classic SEO, but with the additional filter of "what source can I extract and rewrite confidently."
2. Trained model knowledge: when ChatGPT responds without browsing, using what it learned in its training corpus. There's no explicit citation link here, but your brand can appear mentioned if you were part of the corpus or if you appear in authority publications that did enter training.

The two paths require different things:

• For browsing: your site must be accessible to the crawler (ChatGPT-User or GPTBot), have readable static HTML, and extractable content.
• For trained knowledge: you need presence in authority sites (Wikipedia, large publications, citations in other recognized blogs) that did enter the corpus.

The 3 categories of presence

A brand can appear in ChatGPT in three ways, and it's worth knowing which one you're fighting:

1. Explicit mention in response: the user asks "recommend SEO agencies in Mexico", ChatGPT responds with a list, your name comes up.
2. Cited as source: the user asks "what is GEO", ChatGPT responds with a definition and cites your article as source with visible link.
3. Implicit attribution: ChatGPT gives an answer that clearly uses your conceptual framework, vocabulary, or examples but doesn't mention you explicitly. This is the hardest to measure.

What you're fighting for determines the strategy: for explicit mention you need brand authority + presence in listicles + external reviews. For source citation you need dense and citable content. For implicit attribution you need consistency and volume.

What DOES work (proven)

An analysis of 700,000 conversations with citations in ChatGPT during Q4 2025 shows reproducible patterns that apply today:

1. Pages with 5+ statistics with source and link are cited approximately 3 times more than pages without data.
2. Self-contained sentences (that make sense read alone) are what the LLM extracts to cite. Example: "94% of B2B buyers used an LLM in their purchase process in 2025" works. "As we saw before, the percentage keeps rising" doesn't.
3. Clean H1-H2-H3 structure helps the LLM understand what's answer and what's subtopic. H2 with questions (## How is GEO measured?) is especially extractable.
4. Wikipedia, government sites, and authority publications dominate citations. Being mentioned in one of those (with link to your site) accelerates your brand authority.
5. Enriched schema.org: Article with Person author, Organization or ProfessionalService for the entity, FAQPage when there's Q&A. Reduces ambiguity about who you are and what you talk about.
6. External linking to primary sources: if you cite a datum, link to the study, not to a blog citing it. The chain of trust matters.
7. Explicit robots.txt: User-agent: ChatGPT-User and User-agent: GPTBot with Allow: /. If you block them, you don't appear.

What does NOT work

Three common practices that actively hurt you in GEO:

1. Keyword stuffing: the LLM detects abnormal density and distrusts. What for classic Google was "aggressive but worked", for LLMs is low-quality signal.
2. Synthetic content without opinion: AI-made articles that don't add own data, don't cite sources, don't have voice. LLMs don't cite articles like that because they don't contribute new information.
3. Pure CSR without SSR: the LLM reads the HTML the server returns. If your site is a React/Vue/Svelte SPA without pre-render, the crawler sees <div id="root"></div>. It doesn't exist for it.
4. Blocking AI bots out of fear: some brands blocked GPTBot thinking "Google is the only thing that matters." Result: zero presence in ChatGPT, zero citations, zero mentions.
5. Pages without schema or with bad schema: Article without author, Organization without contactPoint, BreadcrumbList badly structured. The LLM can't infer entity.

How to audit your own ChatGPT presence in 5 minutes

Without paid tools, in order:

1. Open ChatGPT in browser or app, without prior context (new chat).
2. List 3 real questions your ideal client would ask. Agency example: "recommend SEO agencies in Mexico", "what's the difference between SEO agency and GEO agency", "agencies in Mexico that do technical SEO for Next.js".
3. Ask each one, log the literal response.
4. Mark:
   • Did your brand appear mentioned? (yes/no)
   • Did competitors appear? (list of names)
   • If no one in your category appeared, what came up instead?
   • Were there citation links (Sources button)? If yes, which domains?
5. Repeat with Claude, Perplexity, and Gemini. That's 4 engines × 3 questions = 12 datapoints. Takes 5-10 minutes.

Result: you have baseline. If in no engine you appeared mentioned, the problem is bigger than technical: it's absence of presence. If you appeared in one and not in others, your authority is uneven.

90-day strategy to lift mentions

It isn't magic. It's directed work:

Day 1-30:

• Resolve technical blockers: SSR if you're in CSR, enriched schema, robots.txt allowing AI bots, dynamic sitemap.
• Setup of weekly measurement dashboard (the 12 questions on the 4 engines).

Day 31-60:

• Publish 4-6 dense articles with verifiable data on your category. Each with 5+ statistics with link to source, self-contained sentences, clear H2s.
• Outreach to 3-5 Mexican authority publications (Marketing4eCommerce, Forbes Mexico, Expansión) for mention or interview.

Day 61-90:

• Weekly dashboard review: did mention count rise on any question?
• Refine content: the articles being cited are the clue of which type works most in your category. Replicate format.
• If the dashboard doesn't show improvement, there's a deeper problem (insufficient domain authority, undifferentiated content, technical still broken).

At 90 days you know if the GEO bet is paying off or needs adjustment. It's not the timeline for "first client from ChatGPT", it's the timeline for having measurable visibility signal.

What to do if you don't appear in any question

Three options:

1. Accelerate content + authority: double the blog bet, get mention in 2-3 authority publications, improve schema.
2. Attack more specific niches first: instead of "SEO agencies Mexico" (saturated), fight "Next.js SEO agency" or "GEO agency LATAM" where competition is lower.
3. External audit: if you've spent 90 days investing and it doesn't move, write us. We audit your technical setup + content + presence and tell you what's blocking.

GEO isn't for every business, but for those that depend on being found, the next 24 months are the ones that move most.