How to do a Google Ads audit for Search and Performance Max

A Google Ads audit is an ordered set of checks on whether the account turns spend into correctly measured results. The order matters because every later check assumes the measurement is right. A bid strategy bids toward whatever conversions it counts. If those conversions are wrong, every figure you use to judge search terms, budgets or ads is wrong too.

This guide covers Search and Performance Max campaigns and splits their audit into seven phases. App, Video and Demand Gen campaigns need their own checks, which this guide does not cover. For Shopping feeds it gives only a short product check.

Each check covers where to look in the interface and what a problem looks like. It also covers how to tell the problem from a normal pattern and what to do. Menu names come from Google Ads Help pages, and the interface changes from time to time. If a menu has moved, check the current version of the linked page.

Some guides call the same work a Google Ads account audit or a pay-per-click (PPC) audit.

Do not change the account while you audit it. Collect findings first, note how much monthly spend each one affects and what evidence shows it, then rank them. The last section shows that ranking with a hypothetical example. The phases are:

1. Measurement: conversion actions, counting and values
2. Campaign settings: location, networks, auto-applied recommendations, ad schedule
3. Search terms, match types and negative keywords
4. Bidding and budget: statuses and a short checklist
5. Ads, assets and landing pages
6. Checks specific to Performance Max, including products and channels
7. Change history, prioritisation and reporting

The measurement audit checks what Google Ads counts as a conversion and which signal it uses for bidding. An error here changes the meaning of everything else in the account. That is why it comes before the other phases.

Primary and secondary conversion actions

Where to look: the Summary page under Conversions in the Goals menu. Google's page on primary and secondary conversion actions says primary actions are reported in the "Conversions" column. They are used for bidding as long as their standard goal is used for bidding.

Secondary actions are for observation only and appear in the "All conversions" column. The one exception is a secondary action inside a custom goal. That action is then used for bidding.

What a problem looks like: add-to-cart, page views or form opens sit as primary next to the purchase. The bid strategy then buys more of the cheap, frequent steps instead of traffic close to a sale.

To tell this from a normal pattern, go through each primary action and ask one question. If this action goes up, does the business make money? If not, it should be secondary.

Keeping a micro-step primary can be a deliberate choice, for example when sales are too few for the bid strategy. That choice then needs a written reason.

The same sale counted twice

Where to look: the purchase actions on the same Summary page. A common setup measures one sale twice: with the Google Ads tag and with a conversion imported from Google Analytics. Google sets Analytics-based conversions to secondary by default. The stated reason is "to prevent counting the same event twice for bidding." If someone later switches the import to primary, two actions count the same sale.

A transaction ID does not fix this. Google's transaction ID page only treats two conversions with the same ID as duplicates within the same conversion action. The page on combining several data sources in one action also says deduplication does not happen across two different actions.

A first signal comes from the numbers. Pick a period and compare both purchase actions with your order system. If each action is close to the order count, a double count is likely.

That comparison alone is not proof. Google Ads counts only the orders it attributes to ads, and it dates them by the click. Your order system counts every order on its own date. So match the date range and the scope before you compare, for example orders from paid search traffic only.

Why the two sources rarely agree is covered in the post on GA4 and Google Ads conversion gaps.

To confirm, place a test order. Google's transaction ID page suggests a test purchase checked with Tag Assistant. Look at whether the Google Ads conversion tag and the Analytics purchase event both fire on that one purchase. Check the transaction records too, if your system logs which tags fired.

If both fire and both actions are primary, the same sale enters bidding twice. The fix is to leave only one purchase action as primary.

Counting setting and conversion value

Where to look: the "Count" option and value setting of each conversion action. Google's page on conversion counting options defines two settings. "Every" counts each conversion after an interaction, and "One" counts one conversion per ad click. Google suggests "Every" for sales and "One" for leads.

A change to the count setting only applies to future reporting. It does not correct past data.

What a problem looks like: a lead form counts "Every", so one person who submits twice looks like two leads. Or a store selling products at different prices sends the same fixed value for every sale. Google recommends a different value for each conversion when prices vary.

A fixed value is not always an accident. An expected value set on purpose, for example for a lead, is a different case from a revenue figure nobody updated. Ask which one it is and note the answer.

A fixed value close to your average order value (AOV) does not prove the measurement is right. The average can match while the differences between products are lost. The AOV calculator gives you that average as a starting point. Then compare the values sent for a few orders with the order amounts.

Value is the main input for value-based strategies such as target ROAS (return on ad spend). A value that hides real differences therefore feeds them a weaker signal. To check the reported ROAS against your own figures, use the ROAS calculator.

Enhanced conversions, consent mode and action status

Enhanced conversions hash first-party data, such as an email address, with a one-way algorithm (SHA256). The hashed data then supplements your existing conversion measurement. Consent mode passes the choice from your cookie banner to Google tags, and it does not provide a banner itself. In an audit, check whether both features are on and whether consent mode receives signals from the site's banner.

The status column on the same Summary page is part of the checklist. For an action marked "Inactive", "Unverified" or "Needs attention", go to Google's troubleshooting page. Google notes that missing conversions can come from a broken tag or from too little volume. To separate the two, start with a test conversion in Tag Assistant.

A tag that fires does not yet prove the measurement is right. Check that the hit carries the right conversion ID and label, and the value, currency and transaction ID you expect. Check that it fires at the right moment, such as the order confirmation, and whether consent mode passes the banner choice to it. Then open the Diagnostics tab on the Summary page for the action's own alerts.

Only after these checks look at volume and delay. Google says reporting can take 3 to 24 hours and that conversions are dated by the click.

The settings audit checks the defaults that decide who sees your ads and where. These settings are chosen once when a campaign is built and are then often forgotten. That makes them one of the first places where spend quietly goes to the wrong place.

Location options: presence or interest?

Where to look: the Locations section of the campaign settings and the location options below it. Google's page on advanced location options says targeting includes physical locations and locations of interest by default. The options are "Presence or Interest" and "Presence". The second only reaches people in or regularly in your locations.

Excluded locations use "Presence" by default. So someone outside the excluded area who shows interest in it can still see your ads.

What a problem looks like: a store that only ships within the US targets the US with "Presence or Interest". It then pays for clicks from abroad. To check, open "When and where your ads showed" under Insights and reports in the Campaigns menu. Then open the "Matched locations" tab.

This view shows the locations that matched your ads. Google says these can be users' physical locations or locations of interest. The report alone therefore does not prove a click came from abroad.

Delivery addresses do not prove it either. Someone abroad can order to a US address, and clicks without an order have no address at all. So read two questions separately:

Where do clicks physically come from? An analytics tool gives an estimate. Google Analytics approximates city and region from the IP address of the traffic. Compare that country split for paid traffic with the Matched locations rows.

Can you serve the orders you get? Here the delivery addresses in your order system help. They show where you ship, not where the click happened.

In interest-driven businesses such as travel, real estate or education, showing ads to people abroad is normal. So "Presence or Interest" is not a mistake in every account.

Networks: search partners and the Display Network

Google states that search partners are included by default when you create a Search campaign. The setting is the "Include Google search partners" box under "Networks" in the campaign settings. The same page suggests checking your Display Network settings if you have performance issues.

In the audit, segment the search terms report by network and read search partners and Google Search separately. Keeping partner traffic that converts is not a mistake. The setting is only a problem if that traffic misses your target.

Auto-applied recommendations

Where to look: the "Automatically apply recommendations" setting in the top bar of the Recommendations page, and its "History" tab. The feature works only at account level, and the chosen recommendation types are applied regularly. Google's list includes "Add keywords", "Add broad match keywords" and bid strategy changes, but not budget increases.

A problem looks like keyword or bid strategy changes in change history that nobody remembers making. Not every auto-applied recommendation is bad. The audit question is whether each enabled type fits the account's goal.

Ad schedule and budget pacing

Google's ad scheduling page gives an important detail. Campaigns using ad schedules pace toward a monthly spend of 30.4 times the average daily budget. This holds regardless of how many days the campaign is active in a month. A weekday-only campaign can therefore spend above its daily budget on active days.

That is how pacing is designed, not an error. On any single day, spend can also reach twice the average daily budget. In the audit, compare monthly spend with 30.4 times the daily budget, not with the daily budget.

The search terms audit checks which searches your ads actually showed on. The keyword list shows intent, and the search terms report shows the outcome. The gap between the two is what this phase is about.

The search terms report

Where to look: Search terms under Insights and reports in the Campaigns menu. Sort by cost and add the "Match type" column. Google says the report lists terms "used by a significant number of people". Low-volume terms are left out for privacy reasons, so the report total does not equal campaign cost.

What a problem looks like: most of the cost sits on terms unrelated to the business, free-product searches or job searches. To tell this from a normal pattern, do not look only at spend without conversions. With a long conversion cycle, a relevant term may not have converted yet in this period.

Before adding a negative, answer two questions. Is the term related to what you sell, and has it had enough clicks that a conversion would be expected? Work out the expected rate from your own data with the conversion rate calculator. You can check the term's cost per click (CPC) with the CPC calculator.

Match types

According to Google's page on keyword matching options, broad match can show ads on related searches. These can lack the keyword's direct meaning. Phrase match shows on searches that include the meaning, and exact match on searches with the same meaning or intent. Google calls it "critical" to use Smart Bidding with broad match.

Broad match keywords in a campaign on manual bidding are a common candidate finding. Google strongly recommends the pairing with Smart Bidding, but seeing the combination is not enough to call it an error. Check the search terms and the results of those keywords first. If they bring irrelevant queries or miss the target, record the finding.

The limits of negative keywords

Negative keywords do not match close variants. In Google's example, the negative "flowers" blocks "red flowers" but not "red flower". To block a term fully, add its singular, plural and synonym forms separately.

Check the opposite direction too, because a negative may be blocking a keyword that converts. Treat negative lists as lists you read together with the search terms report, not files you write once.

The bidding and budget audit reads the status labels the account shows and the reasons behind them. A label alone does not tell you whether bidding is set up well, so the section ends with a short checklist. Which bid strategy fits which goal is the subject of a separate post in this series.

The "Limited by budget" status

According to Google's campaign status definition, such a campaign is active and can show ads, but not as often as it could. Find these campaigns with the status filter in the Campaigns table. Then add the "Search lost IS (budget)" column. It is the share of time your ads did not show on Search because of insufficient budget, at campaign level only.

Being limited by budget is not always a problem. If the campaign's return is below target, more budget only means more loss. Confirm the campaign's real return first, then look at budget. For a lead generation account, that means knowing your customer acquisition cost (CAC), which the CAC calculator works out.

Google also states that from August 17, 2026 it is updating bidding for budget-limited campaigns on target CPA or target ROAS. It says this may cause temporary fluctuations. Keep this change in mind when reading a swing that starts after that date.

The "Learning" status

The bid strategy status appears in the "Status" column on the Bid strategies page in the Tools menu. On November 5, 2026, Google's help page listed three reasons for it.

They are a new strategy, a composition change (campaigns, ad groups or keywords added or removed) and a goal change. The same page says the status never lasts more than 7 days. Changes to budgets, targets, creatives or targeting never trigger it.

The Turkish version of that page still lists four reasons, including a settings change, and appears to lag the English one.

In the audit, the problem is not the status but its frequency. If a strategy is learning again every week, change history will show campaigns added and removed or goals changed. Google suggests giving the system one to two conversion cycles after a significant change in conversion behaviour.

Fixing a measurement error from phase one is also a goal change. So do not base performance decisions on the weeks right after the fix.

A short bidding and budget checklist

Beyond the labels, six questions show whether bidding and budget fit the business. The list is for the audit, not a guide to choosing a strategy.

• Does the strategy match the business goal, such as sales value, sales count or qualified leads?
• Do the target CPA and target ROAS sit within your economic limits? A target ROAS below 1 divided by your contribution margin aims at a loss. Compare a target CPA with what a customer is worth over time, its customer lifetime value (LTV). The LTV calculator estimates it.
• Are the conversions in the evaluation window complete? Google suggests leaving out the most recent days when conversions take that long to arrive. The "Days to conversion" segment shows your delay.
• Are brand and non-brand results read separately? Brand searches come from people who already know you, so mixed figures hide how non-brand traffic performs.
• Is there a written reason for how budget is split across campaigns?
• In a lead account, is lead quality tracked beyond the form count? Google recommends bidding toward qualified or converted leads. That means bringing sales outcomes back into Google Ads, for example with offline conversion imports. The funnel analysis calculator shows where leads drop out before a sale.

The ads and landing page audit checks the experience after the click and whether the page delivers what the ad promises. Two indicators in this phase are often misread: Ad Strength and Quality Score.

Ad Strength and Quality Score are diagnostics

Google describes Ad Strength as feedback on asset diversity in responsive search ads. It "isn't used to calculate Ad Rank, Quality Score, or auction wins." An "Average" rating is therefore not a finding on its own.

Google says the same about Quality Score. It calls it "a helpful diagnostic tool, not a key performance indicator." Its components will be covered in a separate post in this series. In the audit, look at which headline and description combinations serve and whether pinned assets cut those combinations without reason.

Landing pages

Every ad's final URL should be a working page that leads to the right product or service. Ads pointing to a removed product, or redirecting to the home page, are quick to find and often missed.

Load time needs its own check, covered in the post on landing page speed. Whether the page converts its visitors is a separate question. The page-level checks in the conversion rate optimization guide help with it.

If you want to measure the effect of a change instead of guessing, run it as an experiment. The logic is explained in the A/B testing guide. To see how long the experiment needs to run, use the sample size calculator.

The Performance Max audit checks where a campaign without keywords sends traffic, and to which pages. Some of what you see directly in a Search campaign sits in separate reports or settings here.

Search terms report and negative keywords

Performance Max has its own search terms report, under Insights and reports in the Campaigns menu. You can add irrelevant terms from it as negative keywords.

According to Google, Performance Max negatives apply only to Search and Shopping inventory. Account-level negatives apply to these campaigns automatically. As in Search campaigns, negatives here do not match close variants.

Brand exclusions

Brand lists stop Performance Max from serving on queries that contain specific brands. Google notes that lists save you from entering misspellings, variants and versions in other languages one by one.

If you run a Search campaign for your own brand, check how much Performance Max spends on the same brand queries. The search terms report shows it. Google states that when a query matches an exact match keyword, the Search campaign holding it is prioritised over Performance Max. Search themes carry the same priority as phrase and broad match keywords.

Final URL expansion

Final URL expansion is on by default in Performance Max. It can send traffic to any page that has not been excluded. If the landing page report shows careers pages, blog posts or out-of-stock products, you have two options.

Open Text customization in the campaign settings and exclude those pages with "Exclude some URLs" under "Final URL". Or turn the feature off. Which pages should receive traffic is a business decision, so expansion being on is not a finding by itself.

Products and channel eligibility

If the campaign uses a product feed, a short product check belongs in the audit. Look for disapproved products in Merchant Center.

Check that price and availability in the feed match the landing page. Google's Merchant Center page says a product with a price mismatch may be disapproved. Then check coverage: are the products you want to advertise in the campaign's listing groups, or excluded?

The channel performance report sits under Insights and reports in the Campaigns menu. Its status column marks each channel as "Eligible", "Eligible (Limited)", "Not eligible" or "Missing required assets". Its diagnostics list product feed issues, such as products not approved or out of stock, and missing or disapproved assets. Add the "Ads using product data" segment to see how product ads perform.

A channel with no spend is not a finding by itself. Google says that if no issue shows, the campaign is likely bidding on channels with a higher expected return.

The last phase places findings on a timeline and ranks them by how much they matter. An audit without that ranking becomes a long list where every item looks equally important.

Change history

Change history in the Campaigns menu lists the changes made in the account over the past two years, and who made them. For each finding, answer one question on this page: when did it start?

If a break in a metric falls on the same date as a settings change, that change is a strong candidate to investigate. It is not yet the cause. Check whether the metric moved where the change applied, such as in that campaign only, and not across the account.

If no change lines up, the cause may sit outside the account, such as seasonality, a competitor or a site change. Auto-applied recommendations also appear here one by one.

Hypothetical example: ranking two findings

The account below is hypothetical and its numbers were created for illustration. It spends $30,000 a month. The audit turns up two findings.

The first is a location finding. A Search campaign for a store that only ships within the US runs on "Presence or Interest". In Matched locations, $1,800 of monthly cost falls on locations outside the US, and those rows show no conversions. That is 6% of account spend.

The second is a double count. A target ROAS campaign spending $20,000 a month has both the Google Ads tag and the imported Analytics conversion as primary. A test order shows both firing on the same purchase.

Reported conversion value is $80,000, and with one action it is $40,000. Reported ROAS is 4 and the deduplicated ROAS is 2.

Contribution here means conversion value times contribution margin, minus ad spend. Contribution margin is what remains after product costs and costs that rise with each sale. To find your own margin, start from the gross margin calculator and subtract the other costs that grow with sales.

The example uses a 40% contribution margin, which puts break-even ROAS at 2.5. On reported value, the campaign seems to leave $12,000 of contribution a month. Counted with one action, it is $4,000 below break-even.

Three numbers to keep apart

Each finding has three different numbers, and only the first two are known at audit time.

The first is affected spend: $1,800 for the location finding and $20,000 for the double count.

The second is the contribution gap visible with current numbers. For the double count, the campaign sits $4,000 below break-even, and the report overstates contribution by $16,000. For the location finding, $1,800 went to rows with no recorded conversions.

The third is the gain from the fix, and it is not yet measured. Fixing the double count corrects the report and the bidding signal. It does not by itself add $4,000.

What happens next depends on how the bid strategy responds, and on what you then decide about the campaign. Switching to "Presence" saves at most $1,800, less any part of it that came from people in the US.

Ranking by budget, evidence and decisions

A rule many audit guides suggest is to start with "wasted spend", meaning spend with no conversions. Applied to this example, the rule ranks the location finding first at $1,800. The double count scores zero, because the campaign does get conversions.

A sounder ranking looks at four things together: affected budget, strength of evidence, effect of the fix on decisions and uncertainty of the gain. The table below compares the two findings on these criteria.

The double count comes first without any promised saving. It affects about eleven times more spend, its evidence is stronger, and every later decision on that campaign reads its numbers. The wasted-spend rule misses it because it only sees spend without conversions, not conversions counted wrongly.

In the hypothetical example, does the ranking change with the assumptions?

The margin decides what the corrected report will say, so it is the input to vary. Below, spend, reported value and deduplicated value stay fixed. Only the margin changes.

At 30% and 40% the campaign falls below break-even once counted with one action. At 50% it breaks even, but the report still overstates contribution by $20,000. The ranking holds in all three rows, because affected spend and evidence do not depend on the margin.

The location amount does not flip it either. At any amount from $1,000 to $3,000, the location finding affects less than a sixth of the double count's spend. It would move ahead only if it affected more spend with evidence as strong. The result holds under these assumptions and is not a general rule.

To repeat it on your account, take each finding's affected spend, its evidence and the deduplicated conversion value from one period. The post on performance marketing metrics explains which metric supports which decision.

Automating recurring checks

Some checks are not done once. Broken final URLs, sudden cost spikes or spend without conversions can come back every week. You can schedule these checks with a script, and Google's ready-made solutions include a link checker and an account anomaly detector.

When a script beats a rule, and how to test one, is covered in the Google Ads scripts guide. For each finding, the audit report should hold seven fields. They are what was seen, which data showed it and how much spend it affects. Then come the recommended action, the responsible person, how the fix will be measured and the verification date.