Disclosing Artificial Intelligence Use in USGS Scientific Information Products

Transparency regarding the use of artificial intelligence (AI) in scientific information products and the analyses supporting those products is essential for maintaining scientific integrity, reproducibility, and public trust. Example statements for disclosing AI use are provided below and are tailored for various AI applications that might be used in assisting the generation of USGS scientific information products. These statements are consistent with USGS guidance (in USGS Fundamental Science Practices [FSP], Survey Manual [SM] chapters 502.4 and 502.6, and FSP FAQs 209, 210, 211, 212), and best practices from the broader scientific community. These examples guide disclosure within USGS scientific information products and are separate from the Department of the Interior Office of the Chief Information Officer (OCIO)’s required annual AI use‑case inventory and other DOI/USGS policies. 

General Principles

• Generative AI poses risks to the quality and integrity of research and information products. USGS scientists need to evaluate, monitor, and disclose these risks and take necessary measures to eliminate errors and inaccuracies, including those introduced into information products as a result of using AI throughout the duration of the study and during the development of scientific products. 

• AI tools cannot be listed as authors; all responsibility for content and accuracy remains with human authors (210). 

• Disclosures should be tailored to specify the type of AI tool used, version information, its purpose, and the extent of its use. 

• All AI-assisted outputs must be reviewed and verified by the author(s) before publication (210). 

• Disclosure is required for most uses of AI tools, including multi-purpose generative AI tools such as large language models. For machine learning models developed for specific scientific tasks (for example, neural networks or random forests), disclosure is not necessary as long as USGS modeling and data management best practices and FSP guidance are followed. These include describing and citing techniques in data release metadata (SM 502.6) or in the scientific publication (SM 502.4). 

• Authors should tailor these disclosures as needed to accurately describe what was done. If multiple AI use cases apply, authors may combine relevant disclosures or use the comprehensive examples provided below. Include tool names and versions, as well as the date of AI use for transparency. Whenever trade, firm, or product names are mentioned, include the standard USGS non-endorsement disclaimer: "Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government." 

   

Example Disclosure Statements by AI Use Case:

1. Generative AI for grammar and language improvement
2. Generative AI for coding and scripting
3. Generative AI for literature summarization
4. Generative AI for idea generation
5. Generative AI for image creation
6. Generative AI for data analysis or visualization
7. Comprehensive example for multi-tool use

1. Generative AI for grammar and language improvement

"During manuscript preparation, generative AI tools [name and version of tools used] were used to improve grammar, clarity, and coherence of the text. All AI-generated suggestions were reviewed, revised as needed, and approved by the authors to ensure accuracy and maintain the integrity of the scientific content." 

2. Generative AI for coding and scripting

"AI-assisted coding tools [name and version of tools used] were employed to generate code used in [describe usage, such as data processing, analysis, application development]. All code was reviewed, tested, and validated by the authors to ensure correctness and reproducibility." 

3. Generative AI for literature summarization

"Generative AI tools [name and version of tools used] were used to assist in summarizing relevant literature. The outputs from these tools were evaluated, expanded, and integrated by the authors to ensure consistency with the final integrated content and conclusions. All citations were verified to be accurate to the original source." 

4. Generative AI for idea generation

"Generative AI tools [name and version of tools used] were used to support brainstorming and the development of research ideas. All AI-generated suggestions were reviewed, refined, and, when appropriate, incorporated by the authors to ensure accuracy of the final work." 

5. Generative AI for image creation 

“Generative AI tools [insert tool name(s) and version(s)] were used to create the graphics in [Figure #]. The generated image was reviewed and verified by the authors to ensure accuracy and appropriateness. Images generated comply with all applicable laws and regulations. This includes, but is not limited to, intellectual property rights, privacy rights, and any other relevant legal considerations. Responsibility for the accuracy and suitability of the image rests solely with the authors.”   

6. Generative AI for data analysis or visualization

“Generative AI tools [insert tool name(s) and version(s)] were used to support data analysis and visualization. These tools were applied under the supervision of the authors, and all outputs were reviewed and validated against established scientific methods to ensure reproducibility and transparency.” 

 7. Comprehensive example for multi-tool AI use

"This research employed a variety of AI tools: generative AI for manuscript editing, and AI-assisted coding for data processing. All AI-generated outputs, including text, code, and analytical results, were reviewed, validated, and, where necessary, modified by the authors to ensure scientific rigor, integrity, accuracy, and compliance with USGS standards." 
 

Disclosure Placement and Formatting Guidance

• Methods Section: Disclose AI use in research design, data analysis, or coding. 

• Acknowledgments: Include AI assistance in writing, editing, image creation, or idea generation. 

• Supplementary Materials: Provide details such as tool names, versions, and prompts and dates of use, if publishable (212). 

Key Points for USGS Authors

• Specify the AI tools used, including names, versions, and dates of use. 

• Describe the purpose and context of AI use. 

• Ensure author oversight and final responsibility. 

• Do not list AI as an author (210). 

• Adhere to all IT security and privacy requirements (212) for the use of AI tools.