How to get cited by ChatGPT & Perplexity in 2026

What GEO actually is

Generative Engine Optimization (GEO) is the practice of structuring your content and presence so AI answer engines — ChatGPT, Perplexity, Google AI Overviews, Gemini and Claude — cite, quote or recommend you inside the answers they generate. Where classic SEO chases a ranked blue link, GEO aims to be part of the answer itself. It's real enough that Adobe agreed to acquire Semrush for around $1.9 billion in late 2025, explicitly framing the deal around brand visibility in the AI era — the clearest market signal yet that this is a channel, not a fad.

The reason it matters is a shift in how people find things. AI answer engines now serve well over a billion users a month and are growing fast, and they often resolve a question without sending a single click to a website. In that world, the old goal of ranking to earn a visit is incomplete: if the AI answers directly, the only way to be present is to be cited in the answer. You either exist in that response or, for that user, you effectively don't exist at all.

How AI engines choose who to cite

To optimize for citations, you have to understand the machinery, which is called retrieval-augmented generation (RAG). Rather than reading your whole page, an engine extracts small chunks — typically 200 to 500 characters — from your paragraphs, scores each chunk on semantic match, source authority, recency and factual density, then weaves the top-scoring chunks into an answer that cites usually three to seven sources. The crucial implication: your paragraph, not your page, is what competes. A brilliant 2,000-word article whose individual passages don't stand alone as answers may never get pulled.

Each engine also has its own plumbing. ChatGPT's browsing pulls from Bing's index, Claude uses Brave Search, Perplexity runs near-real-time live search, and Google AI Overviews sit on Google's index. That's why there's no single "GEO score": you track citation share across each engine separately, because the same content can appear in Perplexity within days and stay invisible to base ChatGPT for months.

Step 1: let the AI in

None of this works if the engines can't reach or index you, and this is where most sites quietly fail — by one estimate, 82% of non-optimized sites get zero AI citations. So the first, unglamorous step is access. Your robots.txt should explicitly allow the AI crawlers — GPTBot, ClaudeBot, PerplexityBot, Google-Extended and Applebot-Extended — rather than blocking them by default or accident. Many sites unknowingly shut these bots out and then wonder why they never appear.

Then make yourself indexable where the engines actually look. Because ChatGPT relies on Bing, submitting your sitemap to Bing Webmaster Tools is one of the highest-return ten-minute tasks in GEO — it directly unlocks ChatGPT citation eligibility and most businesses have never done it. Verify your site in Brave's webmaster tools too, to support Claude. Both are free. Finally, add an llms.txt file at your domain root: an emerging standard that gives AI crawlers a clean, markdown map of your most important pages.

Step 2: write content AI can actually cite

Once the engines can read you, the content has to be built to be pulled. Because retrieval works on chunks, the single most important habit is to front-load a direct, self-contained answer right under each heading — roughly 40 to 60 words that resolve the question on their own, before you elaborate. That passage is what gets extracted and cited; the elaboration supports it but rarely gets pulled by itself. Writing every section so its opening answers its own headline is the core craft of GEO.

// Make every passage extractable, verifiable, authoritative
{
  "front_load": "40-60 word self-contained answer per heading",
  "verifiable": "stats with source + year, original data when possible",
  "structure": "clear claims, Q&A sections, data tables",
  "schema": "FAQPage + Article JSON-LD",
  "eeat": "named author + bio, visible dates, inline references",
  "chunk_test": "does each paragraph stand alone as an answer?"
}

Two things make a chunk disproportionately citable. First, verifiable data: statistics with a clear source and year, and especially your own original data — even a tiny survey makes a page the only source for that fact, which engines love. Second, structured trust signals: FAQ and Article schema, a named author with a real bio, visible publish and update dates, and inline references. These tell the model your content is factual, current and accountable — exactly the chunks it prefers to cite.

Step 3: build authority the AI trusts

Even perfectly structured content needs authority behind it, because engines preferentially read sources their retrieval system already ranks and trusts. This is where GEO inherits from SEO: domain authority and quality backlinks still matter. But there's a newer, increasingly decisive signal — brand mentions. Industry data now suggests unlinked mentions of your brand across the web correlate with AI citations more strongly than backlinks do, because models infer authority from how often and how credibly you're talked about.

There's also a specific place AI engines look that many businesses ignore: community and primary sources. Reddit alone accounts for a large share of Perplexity's citations, and Quora, LinkedIn, GitHub and specialized forums are heavily indexed. Participating genuinely in those spaces — real expertise, not link spam — builds the kind of authentic signal engines reward. Combined with earning mentions on trusted industry publications, this is how you become a source the AI is confident enough to name.

A 30-day sprint to start

GEO is a program, not a one-off, but you can start systematically in about a month. The approach we use begins with a baseline: run your 30 most important queries through ChatGPT, Perplexity, Gemini and Claude, and document exactly where you're cited and where competitors are instead. That baseline is what makes progress measurable rather than guessed.

From there it's an ordered sprint. First, fix access — bot permissions, Bing and Brave submission, llms.txt. Next, optimize your top pages: front-loaded answers, claims with evidence, Q&A sections, data tables, visible author and dates, schema. Then expand brand signals: pitch founder bylines and expert commentary to a handful of trusted publications, and participate authentically in the communities your engines cite. Finally, re-run the 30 queries, measure the delta, and prioritize the next batch. It's deliberate, repeatable, and the early movers compound a lead while competitors wait.

GEO and SEO, together

None of this means abandoning SEO — quite the opposite. Traditional organic search still sends vastly more traffic than all AI engines combined, so it remains the backbone. The point is that GEO is an additive layer: the same technical foundation, structured data and authority serve both, and a site that's strong for SEO is already most of the way to being citable. The extra GEO-specific work — llms.txt, chunk-level answer writing, brand-mention building and per-engine measurement — is the margin that wins citations.

The strategic case for acting now is the measurement gap. AI search converts at notably higher rates than classic search even though it sends less traffic today, and most businesses — 82% by the zero-citation figure — aren't optimized at all. That combination of high-value traffic and low competition is exactly the kind of window that closes. The brands building citable, authoritative content now are claiming AI visibility their competitors will have to fight for later, in both English and Spanish.

Frequently asked questions