ChatGPT does not present a list of ranked pages. It retrieves candidates, evaluates them, and generates a synthesised answer citing only the pages that passed its selection criteria. An AirOps study analysing 548,534 pages across 15,000 prompts found that ChatGPT cites only 15% of the pages it retrieves. The other 85% are pulled into the process, scored, and discarded without ever appearing in the answer.
That 15% figure is the gap this post addresses. Not how to drive traffic, not how to rank on Google — specifically, how to move from retrieved-but-ignored to cited.
ChatGPT reaches 883 million monthly users, up significantly from 300 million weekly users in early 2026. 44% of SaaS brands with strong Google rankings have zero ChatGPT visibility, per EMGI Group's April 2026 analysis. Strong SEO is a contributing factor, not a guarantee. The two channels have diverged enough that you need to build for both explicitly.
How Does ChatGPT Search Actually Work?
ChatGPT Search does not search for the phrase a user types. It expands the query before retrieval through a process called fan-out. The AirOps dataset found that 89.6% of 15,000 original prompts triggered two or more follow-up searches. The total query set expanded from 15,000 prompts to 43,233 queries — nearly a 3x increase.
What this means in practice: 32.9% of all cited pages appeared in fan-out results only. They were never discovered through the primary keyword. Nearly one-third of citation opportunities exist entirely outside what a conventional keyword strategy tracks.
The fan-out queries that triggered citations had one notable characteristic: 95% of them had zero traditional search volume. These are not keywords brands target. They are the sub-questions ChatGPT asks itself while building an answer — questions like "NCLEX pass rates by nursing school" when a user asked "what are the best nursing programs." A page answering that specific sub-question in depth gets cited even if it has no organic traffic and ranks nowhere for the parent query.
What Is the Difference Between Retrieval and Citation?
This distinction is where most ChatGPT optimisation advice breaks down. Retrieval means ChatGPT found and considered your page. Citation means ChatGPT actually used your page in the answer. These require different interventions.
Retrieval eligibility depends on crawl access, indexability, and enough topical relevance to surface in one of ChatGPT's sub-queries. If OAI-SearchBot or ChatGPT-User cannot reach your page, or if the page is not indexed, retrieval does not happen regardless of content quality.
Citation selection depends on what happens once ChatGPT has your page in the candidate pool. At that point, the content itself determines whether it gets cited: how directly the answer appears, how verifiable the claims are, whether the writing is in active voice with specific rather than vague language, and whether the page has schema that tells ChatGPT exactly what type of content it is reading.
Pages ranking in position one on Google are cited by ChatGPT 3.5x more often than pages outside the top 20. But only 12% of URLs cited by ChatGPT also rank in Google's top ten. High Google rankings improve retrieval probability. They do not determine citation. The content quality and structure at the citation stage is what creates the gap.
Which Content Types Get Cited Most by ChatGPT?
The content types that produce the highest ChatGPT citation rates, based on the AirOps dataset and cross-referenced against EMGI Group's April 2026 analysis, share three characteristics: they answer a specific question directly in the first sentence, they include verifiable data with named sources, and they are structured so any paragraph can stand alone without surrounding context.
FAQ content performs disproportionately well because each question-answer pair is a self-contained extractable unit. ChatGPT can pull one pair without needing the page's introduction or conclusion. Definition pages perform similarly — a clean, specific definition in the first sentence of a section is one of the most reliably cited content formats.
Reddit ranks among the top-cited domains across ChatGPT and other AI platforms. An analysis of approximately 250,000 Reddit posts found that Q&A threads account for over 50% of Reddit's AI citations. The strategic implication: authentic answers in relevant subreddits, written as genuine contributions rather than promotions, get surfaced repeatedly. ChatGPT's citation algorithm has become more resistant to manipulation. Obvious promotional content in community spaces works against the brand now.
LinkedIn's domain rank on ChatGPT moved from approximately position 11 to position 5 between November 2025 and February 2026 — over a 2x increase in citation frequency, per Profound's March 2026 data. Long-form articles and posts from named individuals with genuine expertise are being surfaced alongside editorial sources. Founder commentary under real names, with category-specific points of view, is now a direct ChatGPT citation channel.
What Technical Requirements Must a Page Meet?
OAI-SearchBot and ChatGPT-User are the two crawlers involved in ChatGPT Search. OAI-SearchBot handles bulk indexing. ChatGPT-User fires in real time when a user's query needs a current page. Both must be allowed in your robots.txt. Blocking OAI-SearchBot cuts you out of ChatGPT's index. Blocking ChatGPT-User means real-time queries for current content cannot reach your pages.
User-agent: OAI-SearchBot
Allow: /
User-agent: ChatGPT-User
Allow: /Beyond access, JavaScript dependency is the most common technical failure. ChatGPT's crawlers do not execute JavaScript. Content that exists only after client-side rendering is invisible. Check what OAI-SearchBot actually sees with a plain HTML fetch on your most important pages. If the answer content is missing from the raw HTML response, SSR or static generation is the fix.
Page speed matters at the crawl stage. Slow initial server response reduces how thoroughly a page gets parsed. A page with a time-to-first-byte above 800ms is a lower-quality crawl target than an equivalent page below that threshold. Fix TTFB on key pages before writing new content.
How Does the Fan-Out Technique Change What You Write?
A conventional content strategy targets one primary keyword per page. Fan-out citation opportunities are almost entirely outside that keyword universe. 95% of the sub-queries that triggered ChatGPT citations had zero search volume.
This does not mean abandoning keyword research. It means adding a second layer to your content structure. After writing for the primary question (the keyword), identify the three or four sub-questions a buyer might ask while researching that topic. Write a full answer-block for each sub-question as a separate H2 section. Each section becomes a separate candidate for a different fan-out sub-query.
A page targeting "AEO prompt tracking" written only for that phrase is one extraction candidate. A page targeting "AEO prompt tracking" with dedicated sections answering "how many prompts should I track for AEO," "what is a good AI share of voice score," and "how long before prompt tracking changes show results" is four extraction candidates — three of them for sub-queries with no search volume that a conventional strategy would never target.
The NotioncCue Prompt Tracker runs your tracked prompts through ChatGPT on a weekly cadence and shows whether your brand appears in the cited sources. It also surfaces which competitor pages are cited instead, which is where the content brief for your next fan-out section usually comes from.
What Off-Site Signals Matter Specifically for ChatGPT?
ChatGPT's citation algorithm weights off-site signals differently from Perplexity's. Where Perplexity's layer-three reranker amplifies journalism and news coverage, ChatGPT gives more weight to structured data and directory listings alongside editorial mentions. The multi-platform strategy needs to account for this difference.
For ChatGPT specifically: review platform profiles on G2, Capterra, and Trustpilot correlate strongly with citation probability for commercial and comparison queries. Domains with complete profiles on these platforms get selected by ChatGPT as sources more often than equivalent domains without them, per Moburst's February 2026 analysis. This is the entity corroboration mechanism working through a different source set than Perplexity's journalism preference.
Claimed and complete business profiles on major platforms, founder-authored LinkedIn articles in the relevant topic area, and expert quotes in niche trade publications all strengthen the off-site entity signal ChatGPT uses to assess how much to trust a domain on a given topic.
Frequently Asked Questions
Does blocking GPTBot affect ChatGPT Search citations?
No, if you allow OAI-SearchBot and ChatGPT-User separately. GPTBot is OpenAI's training crawler — blocking it keeps your content out of model training but does not affect live search citations. OAI-SearchBot handles ChatGPT Search indexing. The two are independently configurable.
How quickly can ChatGPT citations change after a content update?
ChatGPT's retrieval layer responds faster than model memory. A structural content change on a page that gets re-crawled within days can affect citation selection within one to two weeks. Model memory updates take weeks to months. For the fastest feedback loop on content changes, watch your Perplexity citation rate first — it responds within days and signals whether the structural change is working before ChatGPT's slower cycle confirms it.
If I rank top three on Google, do I automatically get ChatGPT citations?
Not automatically. Pages ranking position one on Google are cited 3.5x more often by ChatGPT than pages outside the top 20 — but only 12% of ChatGPT-cited URLs rank in Google's top ten. Strong rankings improve retrieval probability, not citation selection. The content structure and entity signals at the citation stage determine whether a retrieved page earns a citation.
What is the single most effective structural change for ChatGPT citation?
Moving your key answer to the first sentence of each section. Not the first paragraph — the first sentence. A section that leads with context before the point gets scored lower than an identical section where the answer is sentence one. This single change affects citation selection for every section on the page simultaneously.