What are the best Claude prompts for Google Ads?

By Tom Claydon, Co-founder, Nudge · Updated 22 Oct 2026

The best Claude prompts for Google Ads ask one specific question, name the date range, and say what to rank by. Run them in order: tracking first, then wasted spend, structure, ads, and bids and budgets last, because every later answer depends on the conversion data being right. Below are 40 prompts in that order, ready to copy.

These work in two ways. With Google Ads connected to Claude through an MCP server, Claude pulls the data itself. Without a connection, export the report named under each group as a CSV and attach it to the chat.

How do you get good answers from Claude about Google Ads?

  1. Name the account and the period. "Last 30 days" beats "recently".
  2. Ask for a ranked list. "Rank by spend" or "rank by money at stake" turns a wall of text into a to-do list.
  3. Ask for the number behind every claim. Claude should show spend, clicks and conversions for each item it flags.
  4. Ask what it could not check. A good answer says which data was missing.
  5. Change nothing until you have read the evidence. Treat every answer as a proposal.

Stage 1: Is tracking right? (prompts 1 to 7)

Report to export without a connection: Goals, then Summary (conversion actions table), plus campaign settings.

  1. Conversion status. "List every conversion action in this account with its status, whether it is primary or secondary, its counting setting and its conversion window."
  2. Goals with nothing to bid on. "Which campaigns bid to a conversion goal that has no primary conversion action in it?"
  3. Double counting. "Are any two primary conversion actions likely to fire for the same conversion, such as a form submit and its thank-you page? Show the evidence."
  4. Stopped recording. "Which conversion actions recorded conversions in the 30 days before the last 14 days, but none in the last 14 days?"
  5. Real sales check. "Here is our CRM export of closed sales for last month. How many match Google's reported conversions for the same period, and where is the gap?"
  6. Value check. "Which conversion actions report a value, and does each value look plausible for this business?"
  7. Tracking verdict. "In three lines: can we trust this account's conversion data enough to change bids? If not, what must be fixed first?"

Stage 2: Where is money wasted? (prompts 8 to 16)

Reports to export without a connection: Search terms, Locations (user location), Devices, Day and hour.

  1. Search terms with no return. "List search terms with spend and zero conversions in the last 30 days, ranked by spend."
  2. Negative keyword ideas. "From those search terms, suggest negative keywords grouped by theme, with the spend each would have saved. Flag any that could block good traffic."
  3. Negative conflicts. "Do any negative keywords block keywords we are bidding on?"
  4. Outside the target area. "How much spend came from people outside the target locations? Which campaigns use Presence or interest?"
  5. Device drag. "Compare cost per conversion by device for each campaign. Where is one device more than twice the account average?"
  6. Hours and days. "Which hours and days spend money without converting, over the last 90 days?"
  7. Display and partners. "Is any Search campaign also showing on the Display Network or search partners, and how do those placements perform?"
  8. Auto-applied changes. "What has Google auto-applied in the last 30 days, and did performance change after?"
  9. Waste summary. "Add up the wasted spend found above and rank the fixes by money saved per month."

Stage 3: Is the structure sound? (prompts 17 to 22)

Reports to export without a connection: Campaigns, Keywords, Search terms.

  1. Overlapping campaigns. "Which search terms triggered ads from more than one campaign in the last 30 days?"
  2. Brand mixed in. "Are brand searches mixed into non-brand campaigns? How much of each campaign's conversions came from brand terms?"
  3. Budget versus job. "List each campaign's budget, spend and conversions. Is any campaign that brings sales limited while a weaker one spends freely?"
  4. Duplicate keywords. "Which keywords appear in more than one ad group or campaign?"
  5. Empty ad groups. "Which enabled ad groups have no active ads, or no active keywords?"
  6. Structure summary. "In five lines, what would you change about this account's structure, and why?"

Stage 4: Are ads and pages relevant? (prompts 23 to 30)

Reports to export without a connection: Keywords (with Quality Score columns), Ads, Landing pages.

  1. Weak keywords. "Which keywords have spend and a conversion rate below half the account average over 90 days?"
  2. Quality Score. "List keywords with Quality Score of 5 or below, ranked by spend, with their expected click-through rate, ad relevance and landing page ratings."
  3. Ad copy gaps. "For the top 10 search terms by spend, does the ad text mention what the person searched for?"
  4. Landing pages. "List final URLs by spend. Flag any that look off-topic for their keywords, or that return an error."
  5. Ad variety. "Which ad groups have only one responsive search ad, or ads with fewer than 8 headlines?"
  6. Message test. "Write 5 new headlines for this ad group based on the search terms that convert best. Keep each to 30 characters or fewer."
  7. Competitor terms. "Are we showing for competitor names? What does it cost and convert?"
  8. Relevance summary. "Which three relevance fixes would move the most conversions?"

Stage 5: Are bids and budgets right? (prompts 31 to 36)

Reports to export without a connection: Campaigns with Competitive metrics columns, Change history.

  1. Budget or rank. "For each campaign, show impression share lost to budget and lost to rank over the last 30 days."
  2. Where more budget goes furthest. "Which campaigns convert at or below target cost and are limited by budget?"
  3. Strategy fit. "For each campaign, list the bid strategy and the conversions in the last 30 days. Which have too little data for automated bidding to learn?"
  4. Recent changes. "What changed in this account in the last 30 days, and what happened to cost per conversion after each change?"
  5. Learning. "Which campaigns changed bid strategy or target in the last two weeks and may still be settling?"
  6. Bid summary. "If I could make three bid or budget changes this week, which would you make, and what is the worst case for each?"

Client reporting (prompts 37 to 40)

  1. Monthly summary. "Write a five-line summary of this account's last month for the client: what happened, why, and what we will do next. No jargon."
  2. Change log. "List every change we made last month with the reason and the result so far."
  3. Across all clients. "Across all my client accounts, which three need attention first this week, ranked by money at stake?"
  4. Next month. "What one test should we run next month in this account, and how will we know if it worked?"

Do these prompts work without connecting Google Ads?

Yes, if you attach the matching report. Export it from Google Ads as a CSV and upload it to the chat. A connection saves the export step, lets Claude combine several reports, and works across many accounts at once.

To connect: How to connect Google Ads to Claude.

Can Claude make the changes these prompts suggest?

Only if the connector you use can change accounts. Google's own Google Ads MCP server is read-only. Some hosted servers can make changes. For client accounts, choose one that asks for approval on every change. See six checks before an AI agent touches client accounts.

Where does Nudge fit?

Nudge runs these stages for you when you ask in Claude: it works through any connected client account in this order, ranks what it finds by money at stake, and proposes the fixes. You approve each one; Nudge checks it against the live account, applies it, and can undo it in one step.

Watch out for: Nudge covers Google Ads only, works inside Claude only, and is in private access.

Questions

What is the best prompt to audit a Google Ads account in Claude?

Start with: "Audit this account from tracking up: list the problems in order of money at stake, with the numbers behind each one, and say what you could not check."

Do Claude prompts for Google Ads also work in ChatGPT?

Most do, if the tool has access to the same data. The prompts above are plain questions, not Claude-specific syntax.

Should I let Claude change bids from a prompt?

Only after reading the evidence, and only with a tool that asks you to approve each change and lets you undo it.

How do I stop Claude making things up about my account?

Ask for the number behind every claim, and ask what it could not check. If it gives a figure you cannot find in the account, ask where it came from.

Related: Google Ads audit checklist · Connect Google Ads to Claude · ChatGPT for Google Ads · Nudge FAQ