How to Get Cited by ChatGPT, Claude, and Perplexity in 2026
AI search is finally a real channel
More and more questions are answered by ChatGPT, Perplexity or Google's AI Overviews instead of a list of links. When someone asks an assistant 'what is the best AI cold email tool for indie founders', the few products named in the answer are the only ones that user hears about.
When an AI Overview answers the question on the results page, fewer people click through, and the sources it quotes get most of the attention that remains. Being one of those sources is worth working on.
How LLMs actually pick sources
Generative search optimization is surrounded by a lot of guesswork. The broad mechanism is simple: when an assistant answers with web results, a retrieval step fetches a handful of candidate pages, and the answer quotes the ones that are relevant, trustworthy and easy 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. A 2,000-word ramble with the answer buried in paragraph nine is harder to quote, even if it ranks in classic search.
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: put the direct claim in the first 100 words, then expand
- Use real headings (H2/H3) that mirror how someone would phrase the question
- Include short lists and tables: structured chunks are easy to quote
- Make claims specific and back them up. 'GA4 needs conversion events configured before it is useful' can be quoted; 'analytics is important' cannot
- Date your content visibly and update it, especially on topics that change quickly
- Say who publishes the page: company name, about page, contact details
Structured data
Schema markup is old SEO advice that still helps: it tells search engines in a standard format what a page is about (Article, Product, Organization, SoftwareApplication and other types). Google documents which types it uses and how; follow its gallery rather than adding every type you can find.
If you do nothing else, add Organization markup to your homepage with a real description and contact details, and Article markup to every post with its author, datePublished and dateModified. Check the result with Google's Rich Results Test, which reads your live page.
llms.txt and the new 'robots.txt for AI'
llms.txt is a proposed convention: a Markdown summary of your site at /llms.txt that gives AI tools a clean overview of your most important pages. It is a proposal, not a standard, and no major search engine has said it uses the file for ranking or citations.
It takes about 20 minutes: a short site description, then a categorised list of your most important URLs (tools, pillar articles, pricing). Treat it as cheap housekeeping, not as a ranking lever.
Becoming the cleanest source on a topic
The deeper work is becoming the clearest source on a narrow topic. A page built on your own data (say, subject line length measured across your own sends) is more quotable than a generic 'cold email best practices' guide, because nobody else has it.
Pick three to five topics you know well. 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 is not a precise science, and the assistants change often, so keep the effort proportionate.