Google Ads Audit: A 10-Point Checklist You Can Run with AI

Set up your account and audit period

Connect your Google Ads account to SourcetoMCP, then connect your AI assistant. Follow the setup guide for Claude or ChatGPT.

Start with two comparable periods. Twenty-eight complete days compared with the previous twenty-eight gives each period the same mix of weekdays. For accounts with few conversions, use a longer period and account for changes in season, promotions and budget.

Recent conversions can arrive after the original ad interaction. Choose an end date that accounts for your usual conversion delay, and record the exact dates in the report.

Give your assistant this opening instruction:

Audit Google Ads account [account name and ID] for [dates], compared with [comparison dates]. Our main goal is [purchases or qualified leads], our target is [CPA or ROAS], and we serve [locations]. Use the account's reporting currency and time zone. Analyse only. Make no account changes. For every finding, show the supporting figures, missing information and recommended next step. Flag incomplete reports and row limits.

CPA means cost per acquisition, or cost per recorded conversion in your report. ROAS means return on ad spend, calculated as conversion value divided by advertising cost. Your targets should reflect your margins and the value of a customer.

1. Check which conversions the account records

Review conversion settings before judging campaign efficiency. A campaign reporting inexpensive conversions deserves a different assessment if those conversions are newsletter signups rather than qualified enquiries.

SourcetoMCP can retrieve conversion action settings, including category, status, counting method, attribution model, primary status and default values.

List the account's conversion actions with their category, status, primary setting, counting method, attribution model and default value configuration. Compare them with our goal of [goal]. Flag actions that need review, including possible duplicate measurement or values that do not match the business outcome. Explain what needs checking directly in Google Ads or on the website.

Primary and secondary settings affect reporting and bidding, but campaign goals also matter. Google documents an exception for secondary actions included in custom goals, so check the goals used by each campaign before drawing conclusions. See Google's conversion action guidance.

A settings report cannot prove that a tag fires correctly or that a submitted lead is genuine. Complete a test conversion and compare recorded outcomes with your order system or CRM.

2. Review campaign structure and current status

List your campaigns before comparing them. Identify their channel types, current status, bidding strategies and budgets.

Separate campaigns serving different purposes. Searches for your company name, searches for a product category and campaigns reaching previous visitors can produce different costs and conversion rates.

List campaigns with their IDs, channel types, current status, serving status, bidding strategy and daily budget. Group them by business purpose using the context I provide. Mark uncertain classifications. Identify campaigns that appear active but are not serving, and campaigns whose purpose or structure needs clarification.

Keep current settings separate from historical performance. A campaign paused today can still appear in the previous month's spending report.

Avoid treating every difference in structure as a problem. Investigate whether each campaign has a clear purpose, an appropriate budget and results you can measure against that purpose.

3. Establish the performance baseline

Compare campaign results before examining individual keywords or ads. Start with spend, clicks, conversions, conversion value, CPA and ROAS.

Ask for changes in both amounts and percentages. A large percentage change on a tiny spend deserves different attention from a smaller change affecting your largest campaign.

Compare the two audit periods by campaign. Show spend, clicks, conversions, conversion value, CPA and ROAS. Include absolute and percentage changes. Calculate account-level CPA and ROAS from the totals. Rank the largest changes by their effect on spend and business outcomes, and flag small samples.

Consider a hypothetical campaign that spent $2,000 for 40 conversions, then $2,400 for 30 conversions. Its CPA rose from $50 to $80, an increase of 60%. Spend increased 20%, while conversions fell 25%.

That calculation identifies the size of the problem. The remaining checks help you investigate its cause.

When conversions or spend are zero, ask the assistant to mark the relevant ratio as unavailable. An undefined CPA should not appear as zero.

4. Review the search terms receiving your budget

Search terms show the queries that triggered your ads. Review them against what you sell, who you serve and which locations you cover.

Separate irrelevant searches from relevant searches that have not converted. The first raises a targeting question. The second needs investigation into volume, conversion delay, ads and landing pages.

Review the returned search terms for [campaigns] during the audit period. Group them into relevant, irrelevant and uncertain using this description of our offer: [description]. Show clicks, cost and conversions. Rank irrelevant terms by spend and explain each classification. Report the total cost of those returned terms without assuming it represents all wasted spend.

Google does not expose every query in the search terms report. Privacy restrictions leave some searches out, so the visible rows cannot provide a complete account of every query that received spend. See Google's search terms guidance.

For example, $300 spent on clearly irrelevant returned queries is $300 of identified spend to review. It is not proof that the account's total waste was exactly $300.

5. Audit negative keywords and their scope

Check existing exclusions before proposing new ones. A negative keyword applied to one ad group has a different scope from one applied to a campaign or shared list.

Review whether exclusions still match the business. An old product exclusion can become a problem after you start selling that product.

Review campaign-level and ad-group-level negative keywords, plus shared negative keyword lists and their campaign associations. Compare them with the irrelevant queries identified earlier. Propose additions with the keyword, match type and intended scope. Flag existing exclusions that could block relevant searches. Make no changes.

Use the narrowest scope that fits the reason for exclusion. A term irrelevant to one product campaign can still be useful elsewhere in the account.

Review the proposed match types as well as the words. Excluding a broad concept can remove more searches than you intended.

6. Compare budgets and bidding with your business targets

Review how spending and results relate to your target CPA or ROAS. Identify shared budgets before recommending changes, since several campaigns can draw from the same budget.

For Search campaigns, include available impression share lost to budget and rank. These figures help distinguish different delivery constraints, but they do not establish how profitable extra traffic would be.

Compare campaign spend, conversions, conversion value, CPA and ROAS with our targets. Retrieve current budgets and bidding strategies, and identify shared budgets. For Search campaigns, include available impression share lost to budget and rank. Flag unavailable bidding targets. Propose changes for review while keeping the overall daily budget allocation unchanged.

A campaign below your target CPA deserves investigation for additional spending capacity. Check its conversion quality, available demand and current delivery constraints before increasing its budget.

Avoid asking AI to guarantee the return from a proposed increase. Historical results support a test, but they cannot establish the exact number of additional conversions a larger budget will produce.

7. Check locations, devices and ad schedules

Compare where your ads serve with where your business can fulfil orders or provide services.

Then review device and time patterns. A weak result on mobile can reflect the landing page, form or checkout experience. Switching off mobile traffic without investigating those issues can hide a fixable problem.

Review campaign targeting criteria and country-level user-location performance. Compare the returned locations with our service area of [locations]. Separately compare performance by device, day of week and hour using the account time zone. Show spend and conversion volume alongside efficiency metrics. Flag patterns for investigation without automatically excluding any segment.

Keep the report at the level the connector actually returns. Country-level location data does not establish which suburbs received your budget.

Treat small samples carefully. Two conversions from one device and three from another are rarely enough to justify a confident claim about which device performs better.

8. Use Quality Score to investigate keyword problems

For Search keywords, inspect Quality Score alongside its three components: expected click-through rate, ad relevance and landing page experience.

Google describes Quality Score as a diagnostic tool. It should help you choose what to investigate rather than become the account's performance target. See Google's Quality Score guidance.

Retrieve current keyword Quality Scores and component ratings. Match them to keyword performance using campaign, ad group and keyword identifiers. Prioritise keywords with meaningful spend and below-average component ratings. Recommend whether to investigate the ad, landing page or keyword fit. Keep unavailable scores separate from low scores.

SourcetoMCP's Quality Score report returns the current snapshot. Use it to identify present diagnostic issues.

A below-average landing page rating gives you a reason to review the destination page. It does not prove that a particular headline, image or form field caused the rating.

9. Review ads and landing pages together

Compare ads within similar campaign and ad-group contexts. Look at spend, clicks, conversions and conversion value before deciding which ads deserve a test.

SourcetoMCP's ad performance report provides metrics. Supply the current ad copy and landing page content when you want the assistant to assess the message itself.

Compare ad performance within each relevant ad group. Identify ads needing further investigation, showing spend, clicks and conversions. Using the ad copy and landing page content I provide, check whether the offer, claims and requested action agree. Propose a specific test for each issue and identify the result we should measure.

Check whether the landing page carries through the offer made in the ad. An ad promoting a particular service should lead to a page where visitors can find that service and take the promised action.

Open the page on a phone and desktop. Test its form or checkout yourself. Text supplied to an AI assistant cannot demonstrate whether a button works or a page loads correctly.

Keep ad comparisons observational unless you ran a controlled experiment. Different delivery conditions can affect the results.

10. Review recent changes and prioritise the findings

Check whether recent account edits coincide with performance changes. Budget, bidding, keyword and status changes can identify useful starting points for investigation.

SourcetoMCP's Google Ads change-history operation covers approximately the latest 30 days. It returns which fields changed, but does not provide their complete before-and-after values.

Review available change history within the latest 28 days. Identify changes that coincide with the performance shifts found in this audit. Separate timing associations from confirmed explanations. Then produce a prioritised findings table with the affected campaign, supporting figures, recommended action, missing evidence and method for checking the result.

An edit followed by weaker performance does not prove the edit caused it. Demand, competition, tracking changes and conversion delays can also affect the comparison.

Prioritise confirmed measurement and delivery problems, then investigate spending that conflicts with your offer or targets. Label proposed experiments separately from fixes supported by direct evidence.

Turn the findings into an action plan

Give each finding an owner, a proposed action and a review date. Record the evidence before changing the account so you can compare results afterwards.

Use a report with these columns:

Ask your assistant to list the exact edits for approval before applying them. SourcetoMCP supports Google Ads changes including budgets, keywords and negative keywords, so a separate implementation request can follow the audit once you have reviewed the findings.

You can also save the reviewed report to a connected Google Sheet or Google Doc. Specify the destination file and approve the proposed content first.

Connect Google Ads to SourcetoMCP, choose one account and run the conversion and campaign checks before working through the remaining prompts.