Scoring a Draft Against a Rubric

What this is for

A read tells you how a talk feels. A rubric tells you what it's missing. This guide builds a short scorecard from what you actually want the talk to accomplish, then uses it to find gaps a general "what do you think" pass would miss.

When to use it

Use this after you've had at least one honest audience read (guide 30) and you want a more structured pass before a rehearsal or a deadline. It's most useful when you have a specific outcome in mind for the talk - a sale, a change in how people think, a call to action - and need to check the draft actually earns it.

The prompt kit

I want to build a rubric for my talk before you score anything. Ask me questions until we've got 5-6 criteria that matter for THIS talk specifically - not generic "was it engaging" categories, but things tied to what I actually need this talk to do.

Here's what I want the talk to accomplish: [describe the outcome you're going for]
Here's my one idea and structure: [paste from guide 20]

Here's the rubric we settled on: [paste the finalized criteria]

Now score my draft against each criterion, 1-5, with one sentence of reasoning for each score. Use only what's actually in the draft - don't score me on intent, score me on what's on the page. Here's the draft: [paste it]

For any criterion scoring 3 or below, tell me specifically what's missing or weak, and point to the section where the fix belongs. Don't rewrite anything yet - just diagnose.

Based on the low scores, what's the single highest-leverage fix - the one change that would move the most criteria up at once, rather than five small fixes that each help one score a little?

What good output looks like

A useful rubric has criteria you could disagree about the score on, which means they're specific enough to argue with. "Clarity of the one idea" is scoreable. "Was it good" is not. Good scoring output points to a location in the draft for every low score, not just a number. And a real diagnosis names what's missing, not what's wrong in general - "the close doesn't return to the opening image" is useful, "the ending felt weak" is not.

Watch out for

Building a rubric so generic it fits any talk. If your criteria would work equally well scoring someone else's talk on a different topic, they're not tied closely enough to what your talk needs to do. Push the AI to ask more questions before it locks the rubric.

Treating the score as the point. The number is a pointer to where to look, not a grade on you. Don't chase a 5 across every category - some talks are stronger for being uneven, with one loud strength instead of five adequate ones.

Scoring a draft that hasn't had a structure pass yet. A rubric applied too early just tells you the talk isn't finished, which you already know. Run guide 20 first so there's a real structure to evaluate.