Score Ad Copy with Copy Judge in Claude | Polaris

To score ad copy with Copy Judge, paste your script or hook variants into the same Claude chat where Polaris is connected and ask Claude to run Copy Judge on them - you get feedback on the copy directly in the conversation, before you spend any render credits turning it into video.

Why score copy before rendering

Rendering video or image ads costs credits, and a script with a weak hook or a buried offer will underperform no matter how good the visuals are. Copy Judge lets you catch that earlier, at the text stage, so the render credits you do spend go toward scripts that already have a reasonable shot. This matters most when you are testing several hook variants at once - scoring first narrows the field before you commit to a full parallel batch.

Step-by-step: using Copy Judge in Claude

1. With the Polaris MCP connector active in Claude, paste your draft script, or a set of hook variants, into the chat.
2. Ask Claude to run Copy Judge on the copy - you can ask for a general read or point it at something specific, like whether the hook lands or whether a claim is risky.
3. Review the feedback Copy Judge returns in the conversation.
4. Revise the script based on the feedback, and re-run Copy Judge on the updated version if you made significant changes.
5. Once the copy scores well, move directly into generation in the same chat - attach a product photo or confirm the script, and pick render settings from the tappable option cards.
6. Let the ad render, with the live Polaris panel showing progress and the finished result.

Pro tips

• Run all your hook candidates through Copy Judge in one pass before rendering any of them, so you can compare feedback side by side and choose which ones earn a render.
• Ask Copy Judge to specifically flag compliance risk if you operate in a regulated niche - a general score pass may not surface a borderline claim on its own.
• Use Copy Judge on recreated or edited scripts too, not just brand-new drafts - a small rewrite can introduce a new weak line just as easily as a first draft can.
• Treat the score as a signal, not a verdict - a script that scores well should still be tested against real audience results once it is running as an ad.

Honest caveats

Copy Judge evaluates copy quality and can flag obvious risk, but it is not a substitute for a real legal or compliance review in a regulated niche - if your category has strict rules around claims, get that copy checked properly before it goes live regardless of its score. And a high score is not a guarantee of ad performance: Copy Judge tells you the copy reads well, not that it will beat a specific competitor's ad in market, so real testing still decides the winner.

For the connector setup, see the MCP guide, and AI UGC ads for how copy fits the wider ad format. Check pricing for credit details, or open the app to try Copy Judge yourself.