I hate digging through Google Ads search term reports - it eats my time.

The goal was simple: separate the queries that actually drive conversions from the ones that just burn budget, decide what to scale and what to block, and do it without spending half the day in the UI.

After a few frustrating weeks of manually slicing tables, I built a prompt that takes the campaign objective and the last seven days of search‑term data and spits out a ready‑to‑action table. The output includes:

Intent classification (high, medium, low)
Estimated conversion probability
Recommendations: keep, test, or block
Top keywords worth scaling
Negative‑keyword suggestions

In practice the prompt surfaces low‑intent leaks that I would normally miss - the cheap‑modifier searches, irrelevant competitor terms and hidden broad‑match spill‑overs that nibble at the budget.

Here's the workflow I settled on, written as a prompt for a Large Language Model:

Step 1 - Context

Ask for the primary campaign goal (lead gen, e‑commerce sales, bookings, etc.).
Capture the target CPA or ROAS if known.
Confirm the campaign type (Search, Performance Max, Shopping).

Step 2 - Data Input

Paste the last 7‑14 days of the search‑term report, at a minimum including Search Term, Clicks, Impressions, Cost and Conversions.
Optionally add Conversion Value, Campaign/Ad‑Group name, Match Type, Device or Location for richer analysis.

Step 3 - Core Analysis Table
Create a structured table with these columns:

Search Term
Intent Level (High/Medium/Low)
Intent Type (Demand Capture, Demand Shaping, Exploration, Irrelevant)
Relevance to Goal (High/Medium/Low)
Clicks, Cost, Conversions
CPA (cost per conversion)
ROAS (if revenue available)
Conversion Rate
Conversion Probability (percent estimate based on intent + performance)
Action (Scale, Optimise, Test, Add Negative, Isolate)

Step 4 - Decision Framework
Base actions on both intent and performance:

High intent + strong CPA → SCALE
High intent + weak CPA → OPTIMISE (copy, LP, bids)
Medium intent + good CPA → TEST & EXPAND
Medium intent + no conversions → LIMIT / TEST
Low intent + spend → ADD NEGATIVE
Irrelevant → NEGATIVE instantly

Step 5 - Pattern & Leakage Detection
Identify recurring low‑quality modifiers (free, cheap, jobs), competitor terms, mismatched locations, informational queries and broad‑match spill‑overs. Group them into negative‑keyword themes.

Step 6 - Top Opportunities

List the top 20 keywords to scale (high intent, proven performance or strong probability, clear commercial value).
Highlight hidden opportunities: low‑volume but high‑intent queries worth isolating into new ad groups.

Step 7 - Waste Analysis

Flag the queries that cost the most with zero or minimal conversions.
Surface budget‑leakage areas for immediate pruning.

Step 8 - Budget Allocation Guidance
Answer where the budget should be increased, where it should be cut back, and which campaign types deserve more focus.

Step 9 - Immediate Action Plan

Actions to take today: negative keywords to add, queries to scale, queries to pause.
Next 7‑day testing plan: what to test and what to monitor.
Scaling signals: indicators that a keyword is ready for higher spend.

Step 10 - Key Insights
Summarise three to five takeaways - the biggest performance blockers, the primary growth levers, and the insights most advertisers overlook in a raw search‑term dump.

A few accounts I ran this on already showed a 12‑15 % lift in ROAS after implementing the recommended negatives and scaling the highlighted high‑intent terms. The biggest surprise was a set of low‑volume, high‑intent modifiers that were previously filtered out by the default match‑type settings.

Would anyone actually plug this into their workflow, or do you still prefer the old‑school manual drill‑down? I'm keen to hear whether the structured approach saves you time or whether the hands‑on feel of the UI remains indispensable.