Search epoch.ai
Polling on AI Usage
We collaborated with the polling firm Blue Rose Research to survey 5,660 Americans about their AI usage habits in late 2025.
Last updated December 19, 2025
The relative subscription rates for different tiers of the same product (ChatGPT Plus vs Pro) may not be accurate, given the similar names of the subscriptions and low results. These relative rates also may not align with the publicly available information on revenue and subscription numbers. Additionally, companies may have other sources of AI revenue not accounted for here, such as corporate subscriptions or API usage. Also note that some of subscription plans, including those from Microsoft and Google, bundle together both AI and non-AI services.
This question was asked of the 3,382 participants who said they had used an AI service in the past week.
This question was asked of the 3,382 participants who said they had used an AI service in the past week.
This question was asked of the 3,382 participants who said they had used an AI service in the past week.
This question was asked of the 3,382 participants who said they had used an AI service in the past week.
Methodology
This polling was conducted by Blue Rose Research from November 25th to December 3rd, with a sample size of 5,660. Some questions were asked only of the 3,382 participants who said they had used an AI service in the past week. The results have been corrected by Blue Rose Research to represent the behavior of registered voters, who make up more than 70% of US citizens. The full text of each option is available by clicking on the bars in the graphs above.
The uncertainty intervals in the graphs are Wald intervals at 90% confidence. This means that, for each percentage \(p\) based on a sample of size \(n\), we compute an interval of
\[p\pm 1.645\cdot \sqrt{\frac{p(1−p)}{n}}\]where 1.645 is the 90% two-sided critical value from the standard normal distribution. Informally: if we repeated the same survey with many more people under the exact same conditions, selected in the exact same way, there is at least a 90% chance that the resulting score would fall within the interval we provide.
We include additional polling questions and results from Blue Rose Research done prior to our collaboration in the data available for download, also corrected by Blue Rose to represent the behavior of registered voters. This data is presented chronologically and is separate from the charts above, with polling conducted in two phases: from June to July 2025 and again from November 25th to December 3rd.
Use this work
Licensing
This data is free to use, distribute, and reproduce provided the source and authors are credited under the Creative Commons Attribution license.
Citation
Epoch AI, Blue Rose Research, ‘Polling on AI Usage’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/data/polling’ [online resource]. Accessed .BibTeX Citation
@misc{EpochAIPolling2025,
title = {Polling on AI Usage},
author = {{Epoch AI, Blue Rose Research}},
year = {2025},
month = {12},
url = {https://epoch.ai/data/polling},
note = {Accessed: }
}Python Import
import pandas as pd
data_url = 'https://epoch.ai/data/polling_on_ai_usage_dec_2025.csv'
# "keep_default_na=False" prevents "None" from being parsed as NaN
df = pd.read_csv(data_url, keep_default_na=False)
Download this data
Polling results
CSV, Updated December 18, 2025
Questions
PDF, Updated December 18, 2025
Additional Blue Rose data
CSV, Updated December 18, 2025
Additional Blue Rose questions
PDF, Updated December 18, 2025
Explore other work
The changing drivers of LLM adoption
Public data as well as our original polling suggest LLM adoption is roughly on trend, but the underlying drivers are shifting.
Updated Dec 19, 2025
Data on AI Companies
Our database of AI company data, with data on revenue, funding, staff, and compute for many of the key players in frontier AI.
Updated Dec 18, 2025