The AI Citation Driver

ChatGPT runs dozens of searches per query. Most brands only optimize for one.

New research from 15,000 queries and 548,534 retrieved pages reveals what actually gets your content cited by ChatGPT, and what doesn't. Google rank, title alignment, and fan-out coverage drive citation likelihood.
The brands you love are winning on AirOps
What you’ll learn inside the playbook

What you'll walk away with

Retrieval rank

Discovery is not visibility

Across 15,000 prompts spanning 8 query types split evenly between informational and commercial intent, ChatGPT retrieved 548,534 pages. Only 15% of the pages retrieved, earned a citation in the final answer. Being retrieved is only the first step. Pages ranking #1 in Google SERPs were cited 3.5x more often than pages outside the top 20 search results. The report breaks down where citation drops off after retrieval, how Google rank influences which pages make the cut, and which query types give pages a stronger chance of carrying through.
The fan-out surface

A third of citations come from internal searches you can't see

On 89.6% of prompts, ChatGPT generated two or more fan-out queries, expanding the 15,000 starting queries into 43,233. 32.9% of cited pages appeared only in results for a fan-out query, not the original prompt. If you're only tracking primary keywords, you're missing where a third of citations are actually won.
Google rank still matters

Pages ranking #1 in Google Search Results are cited 3.5x more often

Pages ranking #1 in Google were cited by ChatGPT 43.2% of the time -- 3.5x higher than pages outside the top 20. 55.8% of all cited pages held a top-20 Google position for at least one query. Ranking strength in traditional search isn’t just traffic metric anymore, It's a citation advantage.
Domain authority is not the gatekeeper

Mid-authority sites are outperforming the top tier

Nearly three-quarters of all citations went to sites with DA under 80. The DA 20-40 tier earned a larger citation share (26.0%) than the DA 80-100 tier (25.4%). And high-authority sites, despite being retrieved more often than any other group, were cited at the lowest rate: 15.0%. The playing field is more open than most brands assume.
Title-query alignment

Pages whose titles aligned with ChatGPT's search query saw a 2.2x increase in citation likelihood.

Pages whose titles overlapped 50% or more with ChatGPT's search query were cited 20.1% of the time, compared with 9.3% for pages with less than 10% overlap. That is a 2.2x lift from a single alignment signal and one of the clearest levers teams can act on today.

"We were able to focus our energy on the highest-value initiatives for the business in a matter of weeks."

Nick Fairbairn
SVP of Growth, Chime
Nick Fairbain

“We set out to change the narrative around Sanity in AI search, not just show up more often. Staying the brand the models name most, and climbing in the categories that matter, is exactly the signal we were after, and we can see it happening week to week.”

Richard Lawrence
Principal, Agentic Operations & Growth Marketing, Sanity

“If we wanted to 10x the growth we were starting to see, we needed a system that handled SEO, AEO, content production, all of it. AirOps was the first to actually deliver results.”

Shani Mithani
Head of Growth, Venn
who is this for

Built for the teams optimizing for AI search

SEO and content leaders adapting strategy for AI search

Growth and demand gen teams building pipeline from AI channels

Marketing agencies advising clients on AEO

CMOs and VPs evaluating how AI search changes their content investment

AI Search Playbook

Your competitors
are already adapting.
Are you?

Join the brands using this playbook to gain 
AI search visibility. Get the framework, the data, and the 90-day action plan.

Frequently Asked Questions

What is the fan-out query?

When a user sends a query to ChatGPT, the model doesn't answer from memory alone. It issues its own web searches, called fan-out queries, to gather sources, then selects a small fraction to cite. On 89.6% of prompts in this study, ChatGPT generated two or more follow-up searches, expanding 15,000 starting queries into 43,233. This report maps the full pipeline from prompt to fan-out to retrieval to citation, across 548,534 retrieved pages.

What are the biggest findings?

85% of pages ChatGPT retrieves are never cited. Discovery alone is not enough. 32.9% of cited pages were found only through fan-out queries, not the original prompt. Google rank is the strongest citation predictor. Pages at #1 are cited 43.2% of the time, 3.5x more than pages outside the top 20. Domain authority is not the gatekeeper most brands assume: sites with DA 20–40 earned more citations than sites with DA 80–100.

Who should read this?

Content, SEO, and growth leaders who need to understand what drives visibility in AI search. If you're making decisions about content strategy, page structure, or refresh cadence, this research gives you the data to prioritize what actually moves citation rates.

Does traditional SEO still matter?

Google rankings carry a direct citation advantage -- pages at #1 are cited 3.5x more often than those outside the top 20.

How is this different from other AI search research?

Scale and depth of pipeline coverage. This study analyzed 15,000 queries across 8 intent types, evenly split between informational and commercial. ChatGPT expanded those into 43,233 total searches and retrieved 548,534 pages. Instead of stopping at "what gets cited," the report maps every stage: fan-out generation, retrieval selection, Google SERP influence, title alignment, domain authority, readability, and query intent, with controlled comparisons at each step.

What should teams do with this data?

Three things. Track fan-out coverage, not just primary keywords. A meaningful share of citations comes from follow-up searches. Audit pages ranking in Google positions 10-20 for high-intent queries, where small ranking improvements translate into outsized citation gains. And use content refreshes to expand coverage around the adjacent questions ChatGPT explores while building an answer, rather than creating entirely new pages.