Leila Fielding
Use this when: reviewing two AI responses against shared criteria, evidence and notes—not verifying an answer automatically.
WHAT YOU'LL ACHIEVE
Keep the task, responses, human ratings and source-checking notes together, without turning a subjective review into a leaderboard.
COMPARE WITH CARE
Make your judgement visible.
Use the same task and criteria for both responses. Ratings are your own assessment; they do not establish accuracy or prove a citation.
Browser-local: text is held only in this page's React state. It is not sent to a service or saved when you leave or refresh.
01 / SET THE TASK
What should the responses do?
02 / RESPONSE
Response A
03 / RESPONSE
Response B
04 / TAKE STOCK
Review summary
Add a task or response to build a transparent summary. Nothing is scored or assessed until you choose a rating.
Scores are subjective self-assessments against your task and criteria, not objective measures of quality or accuracy, and not fact-checks. Evidence status is user-reported. No winner is declared.
Preview export text
# Compare Two Drafts > Human review notes only. Ratings are subjective self-assessments, not fact-checks. This tool does not independently validate facts or citations. ## Task / prompt Not provided ## Success criteria Not provided ## Response A ### Response Not provided ### Self-ratings - Relevance: Not assessed - Completeness: Not assessed - Clarity: Not assessed - Instruction following: Not assessed - Usefulness: Not assessed - Evidence / source support: Not assessed ### Evidence status: Not assessed ### Evidence / reference None added ### Notes None added ## Response B ### Response Not provided ### Self-ratings - Relevance: Not assessed - Completeness: Not assessed - Clarity: Not assessed - Instruction following: Not assessed - Usefulness: Not assessed - Evidence / source support: Not assessed ### Evidence status: Not assessed ### Evidence / reference None added ### Notes None added --- Scores reflect the user's own judgement against the stated task and criteria. Unassessed items are not scored. Evidence status is user-reported; this tool does not check claims, sources, or citations and does not identify a winner.
KEEP THE REVIEW IN CONTEXT
The score is only one part of the decision.
Consider the task's stakes, the information shared, and whether you can independently check anything important before relying on an output.