Option A: Claude: whole-document matrix, then gap report
Best when: You want one assistant carrying the full analysis
What it does: AI builds matrix and report; you check both
Strong at: Both files fit one chat; report exports as Word
Limit: Classification correctness is not stated; check every matrix row
Free plan: Yes
Option B: Gemini Notebook: per-requirement questions with citations
Best when: Every matrix row must trace to a quotable passage
What it does: AI answers per requirement; you track coverage
Strong at: Citations click through to the quoted location
Limit: A requirement nobody asks about gets no row
Free plan: Yes
Option C: You build the matrix; ChatGPT writes the report
Best when: Classification judgement must stay entirely with you
What it does: You classify; AI only writes up your matrix
Strong at: Accuracy risk sits in your step, not the model's
Limit: You carry the full reading and classification workload
Free plan: Yes, limited (upload and feature caps)