Option A: Gemini Notebook: cited claim matrix, then the debate pack
Best when: You want to audit fairness by following citations
What it does: Sources uploaded once; Gemini Notebook maps and drafts
Strong at: Every claim cites the source span behind it
Limit: Retrieval may not draw on every source; you add counterclaims
Free plan: Yes, limited (notebook and source caps)
Option B: ChatGPT: argues each side separately, then builds the pack
Best when: You want each side argued at full strength
What it does: ChatGPT argues and builds; you check each stage
Strong at: Each side is argued without seeing the other
Limit: Needs several separate chats and checks
Free plan: Yes, limited (uploads and data analysis)
Option C: Claude: drafts the pack from your matrix
Best when: You want full control over fairness
What it does: You build the matrix; Claude drafts the pack
Strong at: Fairness judgements stay yours; the assistant sees only the approved matrix
Limit: You build the whole matrix yourself
Free plan: Yes, limited (5-hour and weekly usage caps)