How do you attribute LLM/agent use in scripts or code?

Currently, I’m doing something like this: when an LLM or coding agent was used, I add a short disclosure near the top of the script. For example, older scripts may mention the chatbot used and the model or model series; newer scripts may mention the agent(s) and/or chatbot(s) and the specific model or model series used.

For example, for an older script:

# OpenAI ChatGPT and Codex (GPT-4- and GPT-5-series models) were used in
# development.

I have also started mentioning LLM/agent use in some Git commit messages where/when appropriate, and mentioning of the LLM/agent use in at least top-level user-facing docs (e.g., top-level README.md).

I’m curious how others are handling this in their work. Do you disclose LLM/agent use in script headers, README files, acknowledgments, methods sections, commit messages?

Part of my uncertainty is terminology. “Assisted” often feels too weak when an agent makes substantial contributions. An English passive construction like “was used in development” seems a bit stronger to me, but maybe still not quite right. At the same time, an agent itself cannot take responsibility for the code, so I’m inclined to label myself, and not the LLM/agent, as the author: I review, edit, test, decide what is accepted, and I take responsibility for those decisions and the code I maintain. Still, I do not want to imply that I can personally certify the provenance of every LLM-generated suggestion.

Of course, not asking for legal advice or anything like that—just curious how others here handle this.