Optyino.ai is a GEO/AIO/LLMO platform that visualizes and optimizes brand visibility in the era of AI search. Leveraging accumulated AI response logs, it analyzed how much the composition of websites (domains) cited in generative AI responses changes between the beginning and end of a month.
This study analyzed 13,982 AI responses collected between August 11, 2025, and July 10, 2026, based on 40 diverse prompts spanning industries such as education, housing, finance, B2B services, and product e-commerce, including queries like "Recommend AI seminars for working professionals," "Which company is recommended for apartment construction and management?" and "Recommend FP consulting services."
Survey Summary
From the 13,982 analyzed AI responses (7,093 at month-start, 6,889 at month-end), 133,913 cited URLs were extracted. After removing duplicates within individual responses, the total number of unique cited URLs was 129,628.
Among these, 5,213 unique cited domains were identified at month-start and 5,147 at month-end. Only 3,250 domains were cited in both periods.
The "domain overlap rate," which measures the proportion of common domains among all domains cited in either period (union), was 45.71%. This means more than half of the domains were cited in only one of the two periods.
On the other hand, the "composition change index," which measures the difference in distribution of citation counts by domain between month-start and month-end, was only 13.94% (indicating 86.06% overlap in composition). The most frequently cited domain in both periods was my-best.com.
The combined share of the top 10 domains by citation composition was 14.84% at month-end, and the lineup of top sites remained largely unchanged.
When analyzed by AI model, the domain overlap rate ranged from 37.44% to 69.01%, and the composition change index ranged from 11.02% to 29.20%, showing significant variation across models. Similarly, by prompt, the composition change index varied from 11.72% to 35.29%, a nearly threefold difference.
Background of the Survey
Websites cited in generative AI responses are increasingly becoming a key factor determining how visible a brand or company is in the age of AI search.
For companies implementing strategies such as GEO (Generative Engine Optimization), LLMO, or AIO, understanding how much the set of cited websites changes over short periods is crucial for prioritizing actions and determining the frequency of performance evaluation.
If the composition of cited sites fluctuates significantly on a monthly basis, performance must be measured in shorter cycles. Conversely, if it remains stable, a long-term, cumulative strategy becomes effective.
Therefore, this survey compared AI responses to the same set of prompts at the beginning (1st–5th) and end (last 5 days) of each month, analyzing how much the composition of cited domains changes from four perspectives: overall, top domain composition, by AI model, and by prompt.
Key Survey Findings
Generative AI Cited Domains Show Only 45.71% Overlap Between Month-Start and Month-End
Of the 5,213 unique cited domains at month-start, only 3,250 were also cited at month-end. The "domain overlap rate," defined as the proportion of common domains relative to the total number of domains cited in either period, was 45.71%.
The remaining 54.29% were cited in only one period: 1,963 domains (27.61% of total) only at month-start and 1,897 domains (26.68%) only at month-end.
With only 45.71% of domains common to both periods, the total number of domains cited in just one period slightly exceeds those cited in both.
At first glance, this suggests a significant turnover in websites cited by generative AI between month-start and month-end.
However, as discussed below, the lineup of top-cited sites remains stable, and the majority of turnover occurs among mid-to-lower tier domains with fewer citations.
Composition Change Index at 13.94%; Most-Cited Domain Consistent Across Both Periods
The "composition change index," which reflects weighted changes in overall composition based on citation volume, was only 13.94%, significantly lower than the 45.71% domain overlap rate when counting domains.
Top 10 Domains by Citation Share at Month-End
Month-Start Share
Month-End Share
Difference (Month-End − Month-Start)
1st Domain
2.36%
2.46%
+0.10%
2nd Domain
2.07%
2.18%
+0.10%
3rd Domain
2.02%
1.84%
-0.18%
4th Domain
1.32%
1.54%
+0.22%
5th Domain
1.24%
1.35%
+0.11%
All top 10 domains by citation share at month-end were also in the top tier at month-start, with the largest change in share being only 0.22%.
The combined share of the top 10 domains was 14.84% at month-end. The most frequently cited domain in both periods was my-best.com, with its share increasing slightly from 2.36% at month-start to 2.46% at month-end.
However, not all high-ranking domains showed uniform stability. One domain ranked 2nd at month-start with a 2.17% share dropped to 17th place with only 0.71% at month-end.
The reason the composition change index remains low (13.94%) despite a low domain overlap rate (45.71%) is that the composition of top-cited sites changes little, and most variation occurs among domains with fewer citations.
For detailed analysis including ranks 6 to 10, see:
https://optyino.ai/press/monthly-domain-citation-turnover
Domain Overlap Rate by AI Model Ranges from 69.01% to 37.44%; ChatGPT Shows Highest Composition Change Index at 29.20%
Among eight AI models compared, Copilot had the highest domain overlap rate at 69.01%, while Google AI Overview had the lowest at 37.44%, a difference of 31.57%.
AI Model
Common Domains
Domain Overlap Rate
Composition Change Index
Copilot
472
69.01%
11.02%
Claude
309
68.67%
11.04%
ChatGPT
843
39.65%
29.20%
Google AI Overview
751
37.44%
24.89%
While a higher domain overlap rate generally correlates with a lower composition change index, the two are not perfectly aligned.
ChatGPT, with the second-lowest domain overlap rate (39.65%), had the highest composition change index among the eight models at 29.20%. In contrast, Google AI O
FACT BOX
- Source: PR TIMES
- Category: Survey
- Products / services: Optyino.ai