Tactical guide
How to get cited by ChatGPT
To get cited by ChatGPT, structure each page so the first sentence is the literal answer to a plausible prompt, attribute every claim to a named human with title and credential, and ship FAQPage + NewsArticle + Speakable JSON-LD on every URL. Then make the page discoverable to GPTBot, ClaudeBot, and PerplexityBot via a clean sitemap and a llms.txt file. First citations typically appear within 7–14 days. The same checklist works for Gemini, Perplexity, Claude, Copilot, and Grok — the engines disagree on ranking, not on what makes a page liftable.
Key facts
- Answer-first lede: first sentence = literal answer to a real prompt.
- Named human attribution beats anonymous claims for every citation.
- FAQPage + NewsArticle + Speakable JSON-LD on every page.
- llms.txt + llms-full.txt accelerate AI engine discovery.
- First AI citations land 7–14 days after a well-structured page is crawled.
- Track citations directly — engine-by-engine — not via referral traffic.
What does ChatGPT actually look for when choosing a citation?
Across thousands of monitored citations on the PressGEO network, six structural features appear in cited pages with disproportionate frequency:
- Answer-first lede — first sentence is the literal answer to a real prompt.
- Named human quotes with full name, title, company, and credential signal.
- FAQ schema (FAQPage JSON-LD) with 3–5 Q&A pairs targeting prompt phrasing.
- Numeric specificity with inline sources — every stat carries (provider, date).
- Question H2s matching how users phrase queries.
- Speakable + Article + BreadcrumbList JSON-LD on every page.
Full per-engine rubric: how AI engines choose citations.
What's the fastest workflow to start earning citations?
- Pick your three highest-intent existing pages — the ones buyers would land on after asking ChatGPT.
- Score each with the free GEO Score tool.
- Rewrite the lede to be the literal answer to the prompt the page targets.
- Add a named human quote with title, company, and a credential signal.
- Add a 3–5 question FAQ block, then mirror it as FAQPage JSON-LD.
- Publish a
/llms.txtfile linking to the page catalog and a/llms-full.txtwith the full body text. - Wait 7–14 days. Re-probe ChatGPT/Perplexity/Gemini with the target prompts and log citations.
What are the most common mistakes that block citations?
- Burying the answer behind a marketing intro paragraph.
- Anonymous claims ("industry experts agree") instead of named attribution.
- Stats without sources — engines won't lift unverifiable numbers.
- Keyword-stuffed headings that don't match how buyers phrase prompts.
- Blocking GPTBot, ClaudeBot, or PerplexityBot in
robots.txt. - Treating GEO as SEO — see GEO vs SEO for the difference.
How do I measure whether it's working?
The right metric is citation rate: the percent of monitored prompts that return your URL or your name inside the AI answer. Referral traffic from AI engines is a lagging indicator at best — most awareness lift from a citation never resolves into a click, because the buyer got their answer.
PressGEO publishes the rolling 30-day, network-wide citation rates as open data (CC-BY-4.0) at /research/data. The live monitor lives at /proof.
“The single highest-leverage rewrite is the first sentence. Most pages bury the answer two paragraphs into a marketing intro — moving the literal answer to sentence one routinely swings the GEO score 20+ points and is usually enough to start earning citations within two weeks.”
Frequently asked questions
- Can I make ChatGPT cite my website?
- Yes, but ChatGPT cites pages that satisfy three conditions: an answer-first lede on a question users actually ask, named-human attribution for every claim, and machine-readable structure (FAQPage, NewsArticle, Speakable JSON-LD). Pages that satisfy all three earn citations within days of being crawled by GPTBot.
- What's the single biggest mistake that kills citations?
- Burying the answer. ChatGPT lifts sentences, not paragraphs. If your first sentence is a marketing intro, the engine has no liftable answer and moves on to the next source. Rewriting the first sentence to be the literal answer to a plausible prompt is usually a 20-point GEO score swing.
- How long does it take to get cited after publishing?
- Usually 7–14 days for the first AI engine pickup once a well-structured page is fetched by GPTBot, ClaudeBot, and PerplexityBot. Brand-aware prompts (where the prompt names your company) earn the first citations; non-brand category prompts (where buyers describe the problem) follow as authority compounds over weeks.
- Do I need llms.txt to get cited by ChatGPT?
- Not strictly — ChatGPT cites pages it discovered via GPTBot crawl of standard HTML. But a llms.txt + llms-full.txt feed accelerates discovery for new and updated pages, and is becoming a default expectation across Perplexity, Claude, and Gemini. The marginal cost is one static file; the marginal benefit is faster ingestion.
- How do I track whether ChatGPT is citing my pages?
- Probe the engine directly with the prompts buyers in your category ask, and log whether your URL appears in the answer's cited sources. PressGEO automates this for every release across ChatGPT, Gemini, Perplexity, Claude, Copilot, and Grok — see the live monitor at /proof and the open-data dataset at /research/data.
- Do paid backlinks or SEO tricks help get ChatGPT citations?
- No. ChatGPT and the other AI engines weight structural signals on the page itself (answer-first writing, named attribution, FAQ schema) far more than link graphs for citation selection. The fastest path is rewriting your highest-intent existing pages, not buying links.
Last updated: June 7, 2026 · By PressGEO Research