Two New Studies Show AI-Cited Sources Often Differ From Google's Top Organic Results

A 720-question benchmark across medical and legal niches finds large gaps between Google rankings and sources cited by ChatGPT and Gemini.
By:
 
CALPE/BENISSA, Spain - Aug. 17, 2026 - PRLog -- Two open-data studies from Rotgar Research found that the pages cited by generative AI systems often differ from the pages appearing in Google's organic Top 20 for the same question.
The benchmarks covered 720 US-English questions: 432 across 12 medical niches and 288 across eight legal practice areas. For each eligible question, researchers compared explicit source links returned through configured DataForSEO ChatGPT and Gemini surfaces with the first 20 unique eligible Google organic URLs.
In the medical benchmark, the mean share of AI-cited pages also found in Google's Top 20 was 12.9% for ChatGPT and 53.9% for Gemini. When pages were grouped by registrable domain, overlap rose to 40.1% and 68.6%, respectively.
The legal benchmark showed a similar pattern. Exact-page overlap was 10.6% for ChatGPT and 51.0% for Gemini, while domain overlap was 36.7% and 65.9%.
"A domain can be visible in both Google and an AI answer while the exact cited pages are completely different," said Evgeniy Yudin, founder of Rotgar and author of the studies. "That means conventional rankings are an incomplete proxy for AI-search visibility. Teams need to record the actual cited URLs for a fixed set of questions and measure them separately over time."
The studies do not rank ChatGPT against Gemini and do not claim that appearing in Google causes an AI citation. The collection endpoints can expose different models and citation behavior, and both rankings and citations are dynamic. The analysis also did not evaluate the medical or legal accuracy of answers or sources.
The practical recommendation is to maintain separate inventories for Google organic results, AI citations, and Google AI features such as AI Overviews, then test changes prospectively against a frozen question panel.
Both versioned datasets are released under CC BY 4.0. The public study pages provide the protocols, datasets, analysis outputs, code, checksums, and documented limitations.
Medical study:
https://rotgar.com/medical/resources/google-top-20-chatgp...
Legal study:
https://rotgar.com/law-firms/resources/google-top-20-chat...
About Rotgar Research
Rotgar Research publishes reproducible studies on search visibility, AI citations, and the relationship between conventional search results and generative-answer systems. Rotgar is based in Calpe, Spain.

Contact
Evgeniy Yudin
Founder, Rotgar
***@rotgar.com
End
Source: » Follow
Email:***@rotgar.com Email Verified
Tags:AI search
Industry:Technology
Location:Calpe/Benissa - Alicante - Spain
Subject:Reports
Account Email Address Verified     Account Phone Number Verified     Disclaimer     Report Abuse
Trending
Most Viewed
Daily News



Like PRLog?
9K2K1K
Click to Share