Provenance Labels work at any stage of a project. Answer two questions and we'll show you where to begin.
Starting fresh with Claude. Choose how you want PLGen available — the workflow after setup is the same either way.
Install PLGen in your existing project once. It's there every session — no prompt to paste, no setup each time.
See install guide →Use the standalone PLGen Claude Project. Open it and start — everything's already configured, just for you.
Open PLGen in Claude →Once you're set up via either path above:
plgen initOpens the session. PLGen records the date and tool automatically. No setup questions.
Write, research, draft — whatever you're making. PLGen observes contributions silently throughout the session.
plgen generatePLGen closes the session and produces your label. It'll ask you to confirm the contribution split. You review and approve — the final number is yours.
Registration gives your label a permanent ID and timestamped public record. Useful for academic work, publications, or anywhere verifiable disclosure matters. Become a member →
Starting fresh with ChatGPT. Choose how you want PLGen available — the workflow after setup is the same either way.
Install PLGen in your own Custom GPT once. It's there every session — no prompt to paste, no setup each time.
See install guide →Use the standalone PLGen Custom GPT. Open it and start — everything's already configured, just for you.
Open PLGen in ChatGPT →Once you're set up via either path above:
plgen initOpens the session. PLGen records the date and tool automatically. No setup questions.
Write, research, draft — whatever you're making. PLGen observes contributions silently throughout the session.
plgen generatePLGen closes the session and produces your label. It'll ask you to confirm the contribution split. You review and approve — the final number is yours.
Registration gives your label a permanent ID and timestamped public record. Useful for academic work, publications, or anywhere verifiable disclosure matters. Become a member →
You're starting fresh. PLGen's command set works natively in Claude and ChatGPT — for other tools, use the manual prompt approach.
Using Miro, Notion, Canva, or another tool with AI built in? See the embedded tools guide →
At the start, tell your AI: "We're using Provenance Label standards for this session. Track our contributions as we work."
Work normally. The AI will have context to estimate contributions at the end.
Use the prompt on the embedded tools page to generate a properly formatted label. Review the split and adjust if needed.
Registration gives your label a permanent ID and timestamped public record. Become a member →
PLGen's session tracking and commands work natively in Claude and ChatGPT. Switching means less manual work at the end.
You're already working in Claude — don't switch. Paste the prompt below at the end of your current session. That's it.
The estimation process often surfaces where your contribution actually lives: the research questions, the judgment calls, the edits. If the work doesn't feel like yours yet, this is a useful diagnostic.
Think honestly: who initiated the ideas? Who made the key decisions? What percentage of what's on the page came from you versus the AI? You're not auditing — you're making a good-faith estimate. Write it down before moving to the next step.
Fill in your name, today's date, and the split you estimated. Paste it directly into the Claude conversation you've been working in.
Based on our collaboration, generate a Provenance Label v1.2. My estimated overall split (including work before this session): [X]% human / [Y]% AI. @provenance 1.2 author: [my name] date: [today YYYY-MM-DD] human: [X] ai: [Y] tools: [AI name and version] note: [one plain sentence describing how we worked together] Requirements: - human + ai must equal 100 - Use my prior estimate as the baseline - Be honest, not flattering - Output only the label block above, ready to copy-paste
Claude may suggest adjustments — the final number is yours. Once you're happy with it, your label is ready to use.
Registration creates a timestamped, permanent record of your disclosure. For academic or professional contexts, this is how you prove you disclosed — not just that you labeled. Become a member →
Want PLGen available for every session going forward?
Install PLGen in your Claude Project once — it's there every session, no prompt to paste each time.
Install PLGen →Use the standalone PLGen Claude Project for future sessions — open it, type plgen init, and go.
Open PLGen in Claude →You're already working in ChatGPT — don't switch. Paste the prompt below at the end of your current session. That's it.
The estimation process often surfaces where your contribution actually lives: the research questions, the judgment calls, the edits. If the work doesn't feel like yours yet, this is a useful diagnostic.
Think honestly: who initiated the ideas? Who made the key decisions? What percentage of what's on the page came from you versus the AI? You're not auditing — you're making a good-faith estimate. Write it down before moving to the next step.
Fill in your name, today's date, and the split you estimated. Paste it directly into the ChatGPT conversation you've been working in.
Based on our collaboration, generate a Provenance Label v1.2. My estimated overall split (including work before this session): [X]% human / [Y]% AI. @provenance 1.2 author: [my name] date: [today YYYY-MM-DD] human: [X] ai: [Y] tools: [AI name and version] note: [one plain sentence describing how we worked together] Requirements: - human + ai must equal 100 - Use my prior estimate as the baseline - Be honest, not flattering - Output only the label block above, ready to copy-paste
ChatGPT may suggest adjustments — the final number is yours. Once you're happy with it, your label is ready to use.
Registration creates a timestamped, permanent record of your disclosure. For academic or professional contexts, this is how you prove you disclosed — not just that you labeled. Become a member →
Want PLGen available for every session going forward?
Install PLGen in your own Custom GPT once — it's there every session, no prompt to paste each time.
Install PLGen →Use the standalone PLGen Custom GPT for future sessions — open it, type plgen init, and go.
Open PLGen in ChatGPT →You've been using AI and want to start tracking. You don't need logs or history — you estimate the work done so far, then note it in your next session.
Using Miro, Notion, Canva, or another embedded tool? See the embedded tools guide →
Think honestly: who initiated the ideas? Who made the key decisions? What percentage of what's on the page came from you versus AI? You're not auditing — you're making a good-faith estimate. Write it down.
Tell your AI: "We're using Provenance Label standards. I've been working on this for [context]. My estimated split so far is [X]% human / [Y]% AI. Track our contributions in this session."
Work normally. The AI will track this session's contributions in context of your earlier estimate.
Use the prompt on the embedded tools page to generate your label, or ask your AI to generate one based on your overall split. Review and approve.
Registration creates a timestamped, permanent record of your disclosure. For academic or professional contexts, this is how you prove you disclosed. Become a member →
Your work is finished. Open any Claude conversation, paste the prompt below with your split estimate, and you're done.
Think honestly: who initiated the ideas? Who made the key decisions? Who edited, cut, and shaped the final output? You're not auditing — you're making a good-faith estimate.
Open any Claude conversation and paste the prompt below. Fill in your split estimate, name, and the date the work was completed.
Generate a Provenance Label v1.2 for completed work. My estimated split: [X]% human / [Y]% AI. @provenance 1.2 author: [my name] date: [completion date YYYY-MM-DD] human: [X] ai: [Y] tools: [AI name and version] note: [one plain sentence describing how you worked together] Requirements: - human + ai must equal 100 - Use my estimate above as the split - Be honest, not flattering - Output only the label block above, ready to copy-paste
Claude may refine the wording of the note — the split is yours to confirm. Once approved, the label is ready.
For finished academic or professional work, registration is how the disclosure becomes verifiable — a timestamped public record, not just a label you added. Become a member →
Your work is finished. Open any ChatGPT conversation, paste the prompt below with your split estimate, and you're done.
Think honestly: who initiated the ideas? Who made the key decisions? Who edited, cut, and shaped the final output? You're not auditing — you're making a good-faith estimate.
Open any ChatGPT conversation and paste the prompt below. Fill in your split estimate, name, and the date the work was completed.
Generate a Provenance Label v1.2 for completed work. My estimated split: [X]% human / [Y]% AI. @provenance 1.2 author: [my name] date: [completion date YYYY-MM-DD] human: [X] ai: [Y] tools: [AI name and version] note: [one plain sentence describing how you worked together] Requirements: - human + ai must equal 100 - Use my estimate above as the split - Be honest, not flattering - Output only the label block above, ready to copy-paste
ChatGPT may refine the wording of the note — the split is yours to confirm. Once approved, the label is ready.
For finished academic or professional work, registration is how the disclosure becomes verifiable — a timestamped public record, not just a label you added. Become a member →
Your work is finished. You can still generate an honest label — estimate your split from memory, then generate it with any AI tool.
Used Miro, Notion, Canva, or another embedded tool? See the embedded tools guide →
Think honestly: who initiated the ideas? Who made the key decisions? Who edited, cut, and shaped the final output? You're not auditing — you're making a good-faith estimate.
Use the prompt on the embedded tools page, or open any AI and ask: "Generate a Provenance Label v1.2 for this work. My split estimate is [X]% human / [Y]% AI. Tools used: [tool names]."
For finished work — especially academic or professional — registration is how the disclosure becomes verifiable. Become a member →