How to Get Cited by ChatGPT, Claude, and Perplexity in 2026

AI search is finally a real channel

Two years ago, getting cited by ChatGPT or Perplexity was a vanity flex. In 2026, for a real subset of founder-targeted searches, it's the channel. When someone asks ChatGPT 'what's the best AI cold email tool for indie founders', the three to five products that get named in the answer capture most of the resulting clicks. The rest don't exist for that user.

Industry benchmarks suggest that a meaningful chunk of Google searches now end without a click — they end with an AI Overview that quotes a handful of sources. The traffic isn't gone. It's just consolidated into citations. If you're not optimizing to be one of those citations, you're invisible on a growing share of intent.

How LLMs actually pick sources

There's a lot of mysticism around generative search optimization. The mechanism is actually pretty boring. When an LLM answers a query, it (or its retrieval layer) fetches a handful of candidate pages, ranks them by some combination of authority, relevance, and clarity, and then quotes from whichever sources are easiest to extract clean facts from.

That last part is the leverage point. If your page is well-structured, makes clear factual claims, has obvious attribution, and is easy to chunk into citable pieces, you get cited more. If your page is a 2,000-word ramble with the actual answer buried in paragraph nine, you get skipped — even if you'd rank on traditional Google.

What 'citable' looks like in practice

Concrete things you can do this week to make your pages easier for an LLM to pull from.

• Lead with the answer. The first 100 words should contain the direct claim, then the page can expand into nuance
• Use real headings (H2/H3) that mirror how someone would phrase the question
• Include short, scannable lists and tables — LLMs love structured chunks they can quote
• Make claims specific. 'Most founders set up GA4 wrong' is citable. 'Analytics is important' is not
• Date your content visibly and update it. LLMs prefer recent sources for time-sensitive topics
• Name yourself. Author bylines, company name, and 'About' links help models attribute correctly

Structured data is the boring superpower

Schema markup feels like 2015 SEO advice, but in 2026 it's quietly become more important, not less. AI Overviews and retrieval systems lean on structured data to understand what your page is about: Article, FAQ, HowTo, Product, Organization, Person, SoftwareApplication. Each one gives the model a clean shortcut to your facts.

If you do nothing else, add Organization schema to your homepage with a real description, founder name, and social profiles. Add Article schema to every blog post with author, datePublished, and dateModified. Add FAQ schema where you have actual Q&A. Run your domain through /tools/instant-seo-audit to see what's missing — it's a 10-minute fix that compounds across every future post.

llms.txt and the new 'robots.txt for AI'

A small but growing convention in 2026 is the llms.txt file — a markdown summary of your site at /llms.txt that gives crawlers a clean, hierarchical overview of what's there. The major model providers are increasingly using these as a hint about which URLs are canonical for which topics.

It costs nothing and takes 20 minutes. Create /llms.txt with a short site description, then a categorized list of your most important URLs (tools, pillar articles, pricing). Add /llms-full.txt with the full text of your key pages if you want to make their job even easier. Early adopters are reporting better citation rates; even if the effect is modest, the cost is near zero.

Becoming the cleanest source on a topic

The deeper play, and the one that actually moves the needle, is becoming the unambiguous best source on a narrow topic. LLMs prefer to cite specialists. A page titled 'Cold Email Subject Line Length: Data From 12,400 Sends' will outcompete a generic 'Cold Email Best Practices' guide every time, because it's specific, claimable, and uniquely sourced.

Pick three to five topics that are genuinely in your wheelhouse. Write the most specific, most useful page on each one. Add original data, screenshots, or examples nobody else has. Then earn a few citations from elsewhere on the web (podcasts, sources, expert quotes) so models see external corroboration. This is roughly the same loop as traditional SEO, just judged by a different audience.

How to tell if it's working

Tracking GSO performance is harder than traditional SEO because the platforms don't give you a Search Console. The practical approach: every two weeks, run your top 10 target queries through ChatGPT, Claude, Perplexity, and Google AI Overviews. Note whether you're cited, where, and what they're quoting. Keep a simple spreadsheet.

Over a few months you'll see patterns — which pages get pulled most, which queries you're losing to which competitors, and which structural changes correlate with more citations. It's not a precise science yet, but the founders treating GSO as a real channel today are going to compound a real advantage by the time the rest of the world catches up.