A Google Ads Audit That Follows the Lead Past the Click

Most Google Ads audits start in the wrong place. They open the account, sort campaigns by spend, and work through settings until the checklist runs out. That finds real problems, but it rarely answers the question that made you want an audit in the first place: why does the account report a healthy cost per lead while sales says the leads are thin?

If you are about to raise the budget, cut it, or hand the account to someone new, the useful audit runs in the other direction. Start with a lead your business would actually want, trace it back to the search that produced it, and only then judge the account. Every setting matters less than whether the account is buying the outcome you pay for.

This guide is written for teams already spending meaningfully on Google Ads with an established lead flow. It is not a setup tutorial, and if conversions stopped suddenly, start with why Google Ads stop converting instead.

The short version:

1. Define what a qualified lead is, in writing, before you open the account.
2. Prove the conversion data is real: no duplicates, no missing calls, the right action driving bids.
3. Classify where spend goes by commercial intent, not by campaign name.
4. Check that the page keeps the promise the ad made and asks for the right action.
5. Review automation settings against current Google behavior, with dated checks.
6. Log every finding with its evidence, confidence and next step, then decide what to repair, what to test and what to leave alone.

Define the outcome before opening the account

An account can only optimize toward what it is told to count. If the business counts booked consultations and the account counts form submissions, the two will drift apart, and the account will look better than the business feels.

Lead, qualified lead, booked consult and completed outcome

Write down each stage, who confirms it and where it is recorded. This table is the first page of a useful audit.

Two practical points. First, the qualification criteria must be written, not “sales knows one when they see one.” If two people would classify the same lead differently, the audit cannot measure lead quality. Second, decide which stage is the number the account is judged on. For most lead-generation accounts that is cost per qualified lead or cost per booked consult, not cost per form.

Align lookback periods, conversion lag and business capacity

Reports disagree for boring reasons before they disagree for interesting ones. Before you conclude anything, check:

• Date logic. Google Ads generally reports conversions against the date of the ad click, while a CRM usually records the date the lead was created or qualified. A month-end comparison between the two will never match exactly.
• Conversion lag. If qualification takes two weeks, the last two weeks of any report are incomplete. Judge recent changes on a window that allows for that delay.
• Capacity. If intake can only handle a fixed number of consults per week, a jump in leads may produce a drop in booked rate that has nothing to do with lead quality. Check staffing, response times and missed calls for the same period.

Test the conversion path

This is where many accounts quietly fail. A bidding strategy will spend confidently against bad data.

Event firing, duplicates, primary and secondary actions, and values

Run this event QA checklist on every conversion action that is set to primary.

1. Submit a test lead yourself on every live landing page and confirm one conversion records, not two. Thank-you pages that reload, tags installed both directly and through Tag Manager, and forms that fire on button click and on page load are the usual sources of doubles.
2. Place a test call from the ad’s call asset and from the number on the landing page. Confirm each is tracked under the right action.
3. List every conversion action and mark which are primary. Google’s documentation explains that primary actions are reported in the Conversions column and used for bidding, while secondary actions appear only in All conversions and are not used for bidding (Google Ads Help: primary and secondary conversion actions). A page view, scroll or “clicked phone link” set to primary will teach the system to buy the wrong behavior.
4. Check counting settings. For lead generation, counting one conversion per click is usually the sensible default; counting every conversion is more common for purchases.
5. Check values. If values are assigned, confirm they reflect relative business worth (a booked consult worth more than a brochure download), not a placeholder.
6. Compare a sample. Pull 30 recent conversions and match them to CRM records by date, time and source. Note how many match, how many are duplicates and how many cannot be found.

Calls and offline outcomes

For many service businesses, calls carry a large share of real demand and are the least visible part of the account. If calls are not tracked, the account undercounts what is working and over-credits forms. Our existing article on call tracking before you scale covers the measurement side in depth.

Beyond calls, the strongest audit move is connecting qualified outcomes back to the ad click. Google’s enhanced conversions for leads uses hashed user-provided data from your lead forms and later matches imported offline conversions (such as a qualified or booked status from your CRM) against it. Done well, this lets bidding learn from qualified leads instead of raw form fills.

Treat this as a reconciliation project, not a switch. Before any upload, confirm what data you are permitted to collect and send, who approved it, what your privacy notice says, and what your contracts with vendors allow. For healthcare, finance and other sensitive categories, hashing and automated imports do not by themselves make it permissible to send a data point. Counsel and the data owner decide that, not the ad account.

Update, checked October 1, 2026: Google’s help center states that “Starting June 15, 2026, offline conversions import and enhanced conversions for leads uploads will be migrated to the Data Manager API and blocked in the Google Ads API,” and that developer tokens with no requests between January and June 2026 “will not be allowlisted for legacy access” (settings update). Google’s pages describe the combined enhanced conversions setting with different timing (April 2026 on one page, June 2026 on another). If your CRM uploads run through a custom integration or a third-party connector, confirm which route it uses today and check the diagnostics for recent uploads. Do not assume an older integration is broken, or that it still works, without looking.

Inspect where the spend actually goes

Once you trust the numbers, look at what the money buys.

Search terms and commercial relevance

Export the search terms report for a period long enough to be representative, and classify terms by intent rather than by which campaign triggered them. The categories below are a useful default. The example terms are a hypothetical private school, not a client account.

Hypothetical example: search-term classification

The adjacent informational row is where audits most often go wrong. Some educational searches cost a lot and never produce a qualified lead. Others are the first touch for buyers who convert weeks later. Do not apply a blanket rule. Check whether any qualified leads in your CRM trace back to those terms, and whether a different page or a lower bid would serve them better before you exclude them.

Geography, service eligibility, brand versus nonbrand, and time

• Geography. Compare the location report against where you can actually deliver. Check whether location settings target people in your area or people showing interest in it, and whether that matches your service model.
• Eligibility. If you only serve certain payers, ages, licensure states or service types, check whether ads and pages make that clear enough to stop unqualified clicks.
• Brand versus nonbrand. Mixed brand and nonbrand campaigns make the account look more efficient than its new-demand performance is. Report them separately.
• Time. Look at qualified leads, not just conversions, by hour and day. Calls that arrive when nobody answers are a staffing finding, not a bidding one.

Campaign segmentation versus fragmentation

Segment campaigns when the segments need different budgets, pages, bids or reporting: distinct services, distinct geographies with different economics, brand versus nonbrand. Do not split purely for tidiness. One ad group per keyword synonym spreads conversion data thin, and with today’s broad-matching behavior it often produces overlapping traffic anyway. If a campaign rarely records enough qualified conversions to inform bidding, consolidation is usually worth testing.

Check the promise after the click

Click the ads. All of them that spend meaningfully. For each, note:

• Does the page repeat what the ad promised? If the ad says “same-week appointments,” the page should say it above the fold.
• Is there one primary action? Call, form or booking, matching what intake can handle.
• Does the page state qualification requirements (service area, eligibility, starting price where appropriate) so unqualified visitors self-select out?
• Is the call number tracked and is the form short enough for the stage of intent?
• Does it load and work on a phone?

If Final URL expansion or similar automation is active, also check which pages Google is actually sending people to. A blog post or careers page receiving paid traffic is a common and fixable finding. Our article on landing pages with one job covers page principles in more depth, and our landing page audit guide shows how to check the whole journey.

Audit automation and current campaign behavior

Automation changes what the account does on your behalf, so the audit has to check what is switched on, what the controls are and what Google has recently changed.

AI Max, URL expansion and account controls

Google describes AI Max for Search as combining search term matching (broad match and keywordless technology), text customization and Final URL expansion, with controls including brand inclusions or exclusions, URL exclusions and location settings (How AI Max for Search campaigns works). For an audit, that means four checks:

1. Which campaigns have AI Max or its individual features enabled, and since when.
2. Whether brand, URL and location controls reflect real business boundaries.
3. What the search terms and landing pages reports show for AI Max-matched traffic.
4. Whether qualified-lead rates for that traffic differ from keyword-matched traffic.

Update, checked October 1, 2026: Google’s April 15, 2026 post on upgrading Dynamic Search Ads to AI Max carries a June 11, 2026 update: DSA sunset and auto-upgrade “will begin in February 2027,” while automatically created assets and campaign-level broad match settings continue to be auto-upgraded starting September 2026. The original body text of the post still refers to September for DSA (Google announcement). Check your own account’s notifications and campaign settings rather than relying on the date alone.

Also announced: an August 20, 2026 post describes A/B tests of budgets and ROI targets across multiple Search campaigns, rolling out in September 2026 (testing update), and a September 23, 2026 post introduces a search journey report connecting search terms, assets and landing pages, with “additional details and availability later this year” (reporting update). Announced is not the same as available in your account.

AI Max is not automatically good or bad for lead quality. It can find useful demand your keyword list missed, and it can find volume that converts on the platform but not in the CRM. The way to know is a controlled comparison judged on qualified outcomes. Google’s AI Max experiments split traffic within a single campaign between AI Max on and off, which is a reasonable starting design if your qualified conversion volume can support a read.

Search versus Performance Max and YouTube

A Search audit does not fully audit Performance Max or YouTube. Performance Max serves across multiple Google surfaces and reports differently, and YouTube often plays an earlier role that last-click lead reports will undervalue. If those campaigns spend meaningfully, scope them separately and say so in the findings rather than stretching one checklist across all formats.

When a reasonable audit reaches the wrong answer

Audits can be wrong too, usually when sound general rules meet an account that does not behave like a general account. Here is one we responded to in 2026, from a specialty pharmacy client we manage. The client is not named at their request, and we have kept the figures to proportions.

A third-party agency reviewed the account and made three recommendations, each reasonable on a standard checklist:

1. Remove the highest-spending keywords, because their individual cost per conversion looked high.
2. Add generic copay assistance keywords to capture more volume.
3. Fix the “Limited by policy” status on the campaigns.

The account’s own data told a different story.

The expensive keywords were doorways, not dead ends. In the month reviewed, 96 to 97 percent of conversions came from search queries that did not match any keyword in the account word for word. The flagged keywords each matched well over a hundred distinct queries, and the queries that converted (cost and infusion questions about specific therapies) entered the account through them. Patients in this category tend to search several times before they act, and last-click reporting credits the final search, which makes the early category and brand terms look unprofitable on paper. Removing them would not have returned the money either: a campaign spends toward its budget, so the same dollars would have flowed into fewer, weaker queries.

The copay idea had already been tested, and it failed where the platform could not see it. An earlier run of generic copay keywords produced leads that looked fine in Google Ads. On the phone, many callers turned out to be looking for a drug manufacturer’s assistance program, not a pharmacy to fill a prescription, and they rarely became patients. That only showed up at intake. Copay terms tied to a specific therapy, from people who already held a prescription, were among the account’s strongest performers.

The policy status was normal. Google’s healthcare policy states that when a campaign targets allowed locations and the domain is properly certified, ads will be labeled “Eligible (limited)” and can run in those locations (Google Advertising Policies Help, Healthcare and medicines). For a certified pharmacy advertiser, that is the expected state, not a defect.

One more detail mattered: Google omits low-volume search terms from the search terms report for privacy reasons (Google Ads Help, search terms report). In that account, the visible terms covered only part of the spend, so any analysis built purely on the export was working from an incomplete picture.

None of this means the reviewing agency was careless. It means an audit has to check three things before recommending a cut: what happened to those leads in the CRM or at intake, what the account has already tested, and which rules of the category override the general checklist. If an audit of your account cannot show that it looked at all three, treat its biggest recommendations as hypotheses to test, not instructions.

Prioritize findings by confidence and business consequence

A findings list that mixes “tag fires twice” with “consider testing new headlines” is hard to act on. Log each finding with the same fields so the team can sort by consequence and certainty.

Findings worksheet (blank template with one hypothetical row)

Rank in this order:

1. Measurement faults with high confidence. Duplicates, untracked calls, wrong primary actions. Everything else depends on them.
2. Spend clearly outside the business. Wrong geography, job seekers, services you do not offer.
3. Promise-to-page mismatches on high-spend ads.
4. Structural and automation questions that need a test to answer.
5. Creative and copy refinements.

Decide what to repair, test or leave alone

Not every finding is an experiment. Treat them differently.

Two cautions about reading results. A before-and-after comparison after several changes cannot tell you which change did the work. And fixing measurement often makes performance look better on paper because conversions that were always happening start to count. Report both, and keep a change log with dates so a later reviewer can see what moved and when.

When you may not need an outside audit

An outside audit is not always the right purchase. You may not need one if:

• The account is small enough that one experienced in-house buyer can run this checklist in a day or two.
• Reports already reconcile with the CRM, and the question is simply whether to scale.
• The problem is clearly downstream: leads are qualified, but intake does not call them back. That is an operations fix.
• You already know the answer and need execution capacity, not diagnosis.

In those cases, use the checklist above internally, or hire a specialist for the specific gap (tracking implementation, for example).

Next step

If the account, page and CRM tell different stories, use the Marketing Audit to establish the diagnosis before changing the budget. As published on October 1, 2026, it is a fixed-fee, three-week review of landing pages, ad creative and media accounts, ending in a written diagnosis and a 90-day plan you can run with any team.