How to get cited by ChatGPT
To get cited by ChatGPT you need three things in order: its crawlers must be able to read your site, your pages must survive a selection round that happens before they are even opened, and your content must state answers plainly enough to lift. Most sites fail at step one without knowing it. The good news is that citation behavior has now been studied at scale, so this playbook can cite its sources instead of guessing.
Can you really influence ChatGPT’s answers?
Yes, within honest limits. You cannot buy a mention, and anyone promising guaranteed placement is selling something other than results.
But ChatGPT with search answers from pages it retrieves in real time, and Ahrefs’ study of 1.4 million prompts mapped how those retrievals become citations: the model pulls dozens of URLs per answer, cites about half of what it retrieves, and the pages that win look measurably different from the pages that don’t. Making your site the clearest, most retrievable, most quotable source on your topic raises your odds in every one of those measurable ways.
What’s step one?
Unblock the crawlers, at both layers. Check that robots.txt allows GPTBot, and check that your firewall isn’t overriding it, because Cloudflare blocks AI crawlers by default on new domains and WAF bot lists catch them too. Our scanner tests both in one pass, fetching your page with an AI-crawler user agent to see what GPTBot actually sees.
While you’re there, confirm your content exists in the raw HTML. ChatGPT’s retrieval doesn’t run your JavaScript, so a client-rendered page is a blank page. The firewall guide and JavaScript guide cover the two failure modes in detail.
How does ChatGPT pick which pages to cite?
There’s a gate before the gate. The Ahrefs data shows ChatGPT gathers title, snippet, and URL for each retrieved result and filters on those signals before opening pages, which means your title tag and URL slug are doing selection work most owners never think about. Two findings are directly actionable.
Cited pages showed titles more relevant to the sub-queries ChatGPT generated internally than to the user’s original prompt, so write titles that answer the specific question a section of your market asks, not your brand tagline. And natural-language slugs were cited 89.8% of the time versus 81.1% for opaque ones, one of the few free citation edges available, covered in our URL guide.
What content actually gets quoted?
Content shaped like answers, with the shape measured. Surfer’s citation analysis found almost 40% of AI citations come from a page’s first 100 words, with share collapsing after that. So open with the claim: state the answer in your first paragraph, declaratively enough to be lifted verbatim, with the specific numbers and names that make a quote worth taking.
Then structure the rest as question-phrased headings with a concise answer under each, since retrieval works at passage level and each heading-plus-answer pair is a liftable chunk. The guides that operationalize this are the first-100-words guide and question headings, each of which covers the passage-level version too.
Does structured data help?
Barely, and it’s worth being precise because schema is the second-most oversold item in AI SEO. LLM parsers strip markup and keep text, which Mark Williams-Cook’s fake-address experiment demonstrated vividly when ChatGPT and Perplexity repeated an address that existed only inside broken JSON-LD. One peer-reviewed controlled study found a modest ChatGPT-only lift from LocalBusiness schema, while Ahrefs’ 1,885-page experiment found adding schema didn’t increase citations.
Add the basic five-minute version because it’s free and helps Google rich results, then move on. llms.txt helps even less; server-log studies across 137,000 domains found the file essentially unread.
What about your presence beyond your site?
Models cross-reference before they recommend. A business with a consistent name, description, and story across its site, profiles, and directories reads as a real entity; one with contradictions reads as noise, and assistants hedge on noise. Reviews function as a second check before recommendation, so a strong genuine review base is citation infrastructure, not just social proof.
Fix contradictions wherever they live, including old addresses and renamed services, because a model that can’t tell which version of you is current will often cite someone else entirely.
How do you know it’s working?
Measure the foundation now and the answers over time. Run the free scan to fix everything measurable today, then periodically ask ChatGPT the questions your customers ask and note whether you appear. Keep notes by month rather than by day, since answers vary run to run.
Ranking movement for the question-shaped queries you targeted is the leading indicator, and citations tend to follow it, because retrieval flows through search.
See where your site stands. The free scan takes about fifteen seconds and shows every fix.
Run a free AI visibility scan