Use when you are filling a Performance Max asset group and about to accept whatever the defaults suggest.
pmax-asset-group-build.md
You are a paid search specialist filling one Performance Max asset group. You are not planning campaign structure and you are not setting budgets.
Product or service in this asset group: {{ASSET_GROUP_PRODUCT}}
Landing page text, as a visitor reads it: {{LANDING_PAGE_TEXT}}
Customer lists and search data I could use as audience signals: {{AUDIENCE_SIGNAL_INPUTS}}
Claims I can prove: {{PROVABLE_CLAIMS}}
Produce five numbered sections.
1. Five headlines at 30 characters, five at 20, and one long headline at 90. Print the count in brackets after each.
2. Four descriptions at 90 characters and one at 60, counts shown.
3. An asset to page check as a table: Asset, The line in {{LANDING_PAGE_TEXT}} it comes from, Verdict (on page / paraphrase / not on page). Rewrite anything marked not on page before listing it.
4. An audience signal spec: which of {{AUDIENCE_SIGNAL_INPUTS}} to use as a customer list, which as custom search terms, and which to leave out because it would only broaden the group.
5. Missing assets: the images, video lengths, sitelinks and business fields still needed.
Rules:
- Every claim must trace to {{PROVABLE_CLAIMS}}. Anything else is written as {{CLAIM_NEEDED}}.
- No two headlines may carry the same idea about {{ASSET_GROUP_PRODUCT}}.
- No exclamation marks, no ALL CAPS, no em dashes.
Finish with a count of how many assets carry {{CLAIM_NEEDED}}, so I know how many gaps to close before this group can run.
Replace each placeholder with your own detail. The more specific you are, the less the model invents.
When should I use this rather than the assets Google suggests?
When you are filling an asset group and about to accept the automatically generated suggestions. Those are pulled from your site by a system that cannot tell a claim you can prove from a line of marketing filler. This works to the same limits but ties every asset to your page text and your provable claims.
What do I need in front of me?
The live landing page text as a visitor reads it, not the brief and not an old draft. Section three checks each asset against that text, so feeding it a brief means it rubber stamps assets the page cannot back. Add {{PROVABLE_CLAIMS}} and the customer lists and search data you could use as audience signals.
What comes back?
Headlines at 30, 20 and 90 characters and descriptions at 90 and 60 with counts, an asset to page verdict table, an audience signal spec, a missing assets list, and a count of assets carrying {{CLAIM_NEEDED}}. That count is your readiness figure. A high one means the group would launch on claims nobody has verified.
What is the mistake?
Putting more than one product theme in a single group. The signal spec and the no duplicate ideas rule both assume one theme, so a mixed group produces headlines that fit nothing and signals that broaden the group rather than steer it. Treat the missing assets list as a designer brief instead of launching thin.