How to get your site cited by ChatGPT and Perplexity

How answer engines choose what to cite

Both Perplexity and ChatGPT’s search mode work roughly the same way. They turn the user’s question into one or more web searches, fetch a handful of pages, read them, and write an answer that points back to the pages it used. The citation is a byproduct of retrieval. A page gets cited when it was retrieved and when it contained a sentence the model leaned on.

That gives you two jobs. First, be retrievable for the searches the engine runs, which is mostly classic SEO: crawlable pages, clear titles, and relevance to the question. Second, be quotable once you are retrieved, which means the answer to the question has to be easy to find and easy to lift from your page.

Step 1: Find the questions that matter

Start from the questions your buyers ask, not from your keyword list. Answer engines get full sentences such as “best bookkeeping software for a two-person agency” or “how do I audit technical SEO on a Shopify store.” Write down ten to twenty of these. Sales calls, support tickets, and the People Also Ask boxes in Google are good sources.

Run each question through Perplexity and ChatGPT and note who gets cited. You will usually see a pattern: a few review sites, one or two vendors, a forum thread, and sometimes a documentation page. That list is your real competition for that question.

Step 2: Give each question a page with a direct answer

Engines cite pages that answer the question in plain words near the top. If your page opens with three paragraphs of brand story, the model has to dig, and it will often pick a page that doesn’t make it dig.

• Put a two-to-four sentence answer directly under the H1. State the answer, then the main condition or caveat.
• Use the question, or a close version of it, as the H1 or an H2.
• Follow with specifics: steps, a short comparison table, prices, or definitions. Concrete nouns and numbers you can stand behind are easier to quote than adjectives.
• Keep one main question per page. A page that answers twelve loosely related questions is harder to retrieve for any single one.

Step 3: Make the page easy to fetch and read

Answer engines fetch pages quickly and often without running much JavaScript. If your main content only appears after client-side rendering, or sits behind a cookie wall or a login, the engine may see an empty page.

• Serve the main text in the initial HTML. Check with “view source,” not the browser inspector.
• Don’t block the crawlers you want citations from in robots.txt. OpenAI and Perplexity document their crawler user agents. Check that your CDN or firewall isn’t blocking them either.
• Use real headings (H1, H2, H3) in order, short paragraphs, and lists for steps.
• Keep the page fast and free of interstitials that cover the content.

Step 4: Earn mentions on the sites engines already cite

Look again at your list of cited domains for each question. Often the cited page is a roundup, a comparison, or a community thread that doesn’t mention you. Getting included there can be faster than outranking it.

Practical ways in: pitch the author of a roundup with a short, factual note on what you do differently; answer the forum or Reddit thread honestly and disclose who you are; make sure your listings on review and directory sites are complete and current. A mention on a page the engine already trusts is often what gets you named in the answer, even before your own page is cited.

Step 5: Be consistent about who you are

Models connect facts about your brand from many pages. If your homepage, your LinkedIn page, your directory listings, and your pricing page all describe you differently, the model has less to go on. Use the same name, the same one-line description, and the same core facts everywhere. Add Organization structured data to your site so the basics are machine-readable.

Step 6: Re-check, because answers change

AI answers aren’t stable. The same question can return different sources on different days as the index, the model, and the search queries change. Treat a single check as a snapshot. Re-run your key questions after you ship changes, and keep a simple record of who was cited each time so you can tell whether a change helped.

Common mistakes

• Writing for a keyword instead of answering a question.
• Hiding the answer below a long intro or behind a form.
• Blocking AI crawlers by accident through a CDN bot setting.
• Publishing many thin pages instead of one strong page per question.
• Measuring once and assuming the result holds.