The 30-minute AI Google Ads audit (six steps, prompts included)
Updated July 14, 2026
A proper Google Ads audit used to cost me a full day per account: exports, pivot tables, screenshots, and a write-up nobody reads past page two. With Claude connected to the account over MCP, the same audit takes about thirty minutes, and most of that is me reading. This is the exact six-step sequence I run on new client accounts, with the prompts included so you can paste them as-is. Every step works on the free read-only tier; nothing here changes the account.
Prerequisite: Claude (or ChatGPT) connected to your Google Ads account. If you have not done that yet, the five-minute connection guide covers it, and the safety checklistexplains what happens under the hood. Come back when "list my google ads accounts" returns your accounts.
Step 1: Structure and settings scan (5 min)
Start wide. Before judging performance, you want the map: how many campaigns, what types, which bid strategies, and the settings that silently burn money.
Map this account's structure: every enabled campaign with type, bid strategy, daily budget, and network settings. Flag anything unusual: Search campaigns opted into Display, broad match on low budgets, campaigns sharing a budget, or bid strategies with too few conversions to learn.
What I look for in the answer: Search partners and Display Expansion enabled by default (the classic silent leak), Smart Bidding strategies starved below roughly 30 conversions a month, and duplicate campaigns competing with themselves.
Step 2: Search-term waste (7 min)
The highest-ROI step of any audit, which is why it goes early:
Pull the last 90 days of search terms across all enabled Search campaigns. Group them: (1) terms with $50+ spend and zero conversions, (2) terms clearly irrelevant to our offer, (3) competitor names, (4) everything converting fine. For groups 1 and 2, total the wasted spend and draft a negative keyword plan grouped by theme. Do not change anything.
On a five-figure monthly account this reliably surfaces hundreds to thousands of dollars of monthly waste. The themed grouping matters: you want negative lists you can maintain, not five hundred one-off negatives.
Step 3: Ad and asset health (5 min)
Review every enabled responsive search ad: Ad Strength, pinned assets, and CTR against its ad group average over the last 30 days. Flag RSAs below 1.5% CTR with 1,000+ impressions, ads with everything pinned, and ad groups running a single ad. Suggest which three to rewrite first and why.
The point is triage, not a rewrite session. You leave this step with a ranked shortlist instead of a vague feeling that "creative could be better".
Step 4: Quality Score and relevance (4 min)
Pull Quality Score components (expected CTR, ad relevance, landing page experience) for every keyword with impressions in the last 30 days. Show the distribution, then list ad groups where the weakest component is landing page experience versus ad relevance, as separate lists. Which ten keywords cost us the most because of a low score?
Splitting by weakest component tells you whether the fix is copy work or landing page work, which are different afternoons.
Step 5: Budget pacing and impression share (4 min)
For each enabled campaign, last 30 days: budget, actual spend, search impression share, and impression share lost to budget versus rank. Which campaigns are capped by budget while beating the account's average CPA? Where would you move the next $500 a day?
This is the reallocation conversation every account owner actually wants: not "spend more" but "spend the same, aimed better".
Step 6: Conversion tracking sanity (5 min)
Audit conversion tracking: list every conversion action with source, counting method, attribution model, last conversion date, and whether it is set as primary or secondary. Flag duplicates, stale actions with no conversions in 90 days, and anything that looks like double counting.
Assembling the report (2 min)
One final prompt turns the session into a deliverable:
Summarise this audit as a client-ready report: top five findings ranked by dollar impact, each with the evidence and the specific recommended fix. Then a table of quick wins versus structural projects. Keep it under one page.
Because the whole conversation happened against live data, the report cites real numbers instead of vibes. If a finding needs action, the same connection can execute it later with previewed writes: the negative keyword plan from step 2 is one confirmation away from applied.
Run this monthly and the thirty minutes drop to fifteen, because you are diffing against last month instead of starting cold. And if paid social is part of your mix, the same audit pattern works on LinkedIn Ads through the same setup.