ChatGPT citations favor pages that rank well, match the query in their titles, and stay focused, according to an AirOps study of 16,851 queries. The high-retrieval result was cited 58% of the time, and pages that answered the main question performed significantly less than the broad, comprehensive guidelines.
Why do we care. This study explains how to get ChatGPT quotes: win a return, show a question in your articles, and answer one question very well. In this study, that was more important than scope.
Findings. The return rate was a very strong signal. Pages in the top search position are cited 58.4% of the time, compared to 14.2% of pages in the 10th position.
- The relevance of the title was the strongest feature of the page. Pages with the same strong topic query were cited 41.0% of the time, compared to about 30% of weak matches.
- Focused pages also beat perfect. The pages that answered the main question worked a little better than the comprehensive, comprehensive guides, reducing the traditional “ultimate guide” approach.
What drove ChatGPT quotes. In this study, the pages that won the citations generally ranked well, used topics closely related to the question, and were always focused on answering it.
- The structure helped, but only slightly: Pages with JSON-LD markup posted a citation rate of 38.5% versus 32.0% for incoming pages, and articles with 4 to 10 subtopics performed best.
- Beyond a certain point, length hurts performance: Pages between 500 and 2,000 words performed best, but pages longer than 5,000 words were cited less often than pages under 500 words.
Being fresh helps, up to a point. Pages published 30 to 89 days ago performed best, while pages newer than 30 days performed worst. This suggests that new content may need time to generate retrieval signals.
- Pages more than 2 years old were not cited as often, suggesting that updating content can help if you are already at the level of relevant queries.
About the data. AirOps said it removed the ChatGPT interface, not the API, and analyzed 50,553 responses generated from 16,851 unique questions in triplicate. The dataset included 353,799 pages and rows of information for over 1.5 million followers across 10 verticals and four query types.
Lesson. The Fan-Out Effect: What Happens Between a Question and a Quote
Search Engine Land is owned by Semrush. We are committed to providing the highest quality of marketing articles. Unless otherwise stated, the content of this page is written by an employee or paid contractor of Semrush Inc.
