The Weekly Ads Prompt Pack

A prompt helps once. A weekly set is a system.

Most people use AI like a faster intern: one clever question, one answer, then back to the dashboards. That’s why it never sticks. The operators getting leverage built a set — the same questions, in the same order, every week, so nothing gets missed on the fourth account when they’re tired.

Same Claude everyone else has. The edge isn’t the tool. It’s the operator who built the routine around it.

What’s in it: 21 prompts, 7 clusters

Three per cluster, written out below in full. No 60-prompt flex — these are the ones we actually use. Swap the bracketed bits for your numbers.

1 · Wasted spend (3)

• Search-term leak. “From this search terms report, list every term with spend above [X] and zero conversions in the last 30 days. Rank by wasted spend. For each: negative, or fix the match type / landing page — and why.”
• Placement drain. “From this placement report, find placements taking more than 5% of spend with a CPA above account average. Show spend, conversions and CPA for each.”
• Pacing vs goal. “Spend to date is [X] of a [Y] monthly budget with a [Z] ROAS target. Am I pacing over or under, and what would you change first?”

2 · Search terms & keywords (3)

• Negative miner. “Group these search terms into themes. Flag the themes clearly irrelevant to [product] and give me a negative keyword list, split exact vs phrase.”
• High-intent winners. “From these search terms, find the ones converting above account average that are NOT yet keywords. Suggest match type and a bid rationale.”
• Match-type drift. “Show which keywords are matching to terms far from their intent. Flag the broad match keywords that need tightening.”

3 · Creative & copy (3)

• Fatigue detector. “From this ad-level report over 30 days, flag creatives whose CTR dropped more than 25% versus their first week while spend stayed flat. Rank by spend at risk.”
• Winning angle. “Look at my top 5 ads by conversion rate. Extract the angle each one uses (pain, proof, offer, objection) and tell me which angle is missing from my current set.”
• RSA gaps. “Review these headlines and descriptions against [product]. Flag duplicates, missing benefits and any asset rated Low. Suggest 5 replacements.”

4 · PMax & Shopping (3)

• Input audit. “For this PMax campaign, audit the four inputs I control — audience signals, asset quality, the feed, conversion setup. Tell me which is weakest and why.”
• Asset group thinness. “Which asset groups are below the recommended asset counts? List exactly what’s missing per group.”
• Feed gaps. “From this feed sample, flag missing or weak attributes (title, GTIN, product type, images) that limit Shopping reach.”

5 · Meta & paid social (3)

• Advantage+ audit. “List every Advantage+ / automation setting enabled on this account. For each: what it changes, and what I should check weekly.”
• Audience overlap. “From these ad sets and their targeting, flag which are likely bidding against each other. Suggest how to consolidate.”
• Frequency check. “Flag ad sets with frequency above [X] and rising CPM. Show spend, frequency and CPA trend for each.”

6 · Measurement & reporting (3)

• Conversion sanity. “Review this conversion setup. Flag anything double-counted, page-view-based, or duplicated across Meta / Google / GA4 that would inflate results.”
• Discrepancy check. “Compare platform-reported conversions against GA4 for the same window. Explain the gap by attribution model and window before I call it a tracking bug.”
• The Monday read. “From these Meta and Google exports, write my Monday read: what changed vs last week, what needs a decision, and what’s fine — in that order. If a source has no data, say ‘no data’ instead of estimating.”

7 · Decision & client comms (3)

• Cut / keep / test. “Given these flagged items, triage each into cut, keep or test — one line of reasoning each, plus the risk of doing nothing.”
• Why ROAS moved. “ROAS moved from [A] to [B] this week. From this data, list possible causes ranked by likelihood and what to check for each. Don’t guess — tell me what data you’d need.”
• Client summary. “Turn this weekly read into a 5-line client update: what happened, what we did, what’s next. No jargon, no excuses.”

How to run it

• Monday, once. Export what you have (search terms, ad-level, placements, the GA4 window). Paste, run the cluster you need.
• Read-only by design. Nothing here touches a campaign. The AI reads and flags; you cut, keep or test.
• Honest gaps. If a source hasn’t settled, the prompt is written to say “no data” rather than invent a number. That’s the whole point.