Google Ads Audit Checklist 2026: 12 Blocks, Real Thresholds, Real Example

The full Google Ads audit checklist PPC specialists use for large accounts: 12 blocks from conversion tracking to search terms, with pass/fail thresholds and a real (anonymised) audit example. Free to run.

Updated 2026-08-07 · 16 min read · ppcfy.ai team

Most Google Ads accounts do not fail because of one broken setting. They fail because five small leaks compound: a conversion action that never fires, a bidding strategy chasing clicks instead of leads, generic keywords sitting in model-specific ad groups, and nobody opening the search terms report for three months.

This is the checklist we use when auditing large accounts manually — the same structure ppcfy.ai's free Google Ads audit automates. It is organised into 12 blocks across three levels (account, campaign, ad group), each with what to check, why it matters, and the threshold that separates "OK" from "fix this now".

Throughout, we use one real audit as a running example: a machinery dealer in Germany, ~7 weeks of data, €3,644 spend, 3,168 clicks, 0 tracked conversions. Anonymised, but the numbers are untouched. It is a textbook case of how a technically "clean" account can still be completely blind.

How to score each item

Give every line one of five statuses. Resist the urge to invent more.

A finished audit is not 76 observations. It is one verdict, five prioritised fixes, and the data behind each one.

Account level

Block 1 — Conversion tracking integrity

Nothing downstream is trustworthy until this passes. If this block fails, stop the audit and fix it before touching bids, budgets or keywords.

Check:

• GTM / GA4 implementation. Is the tag actually firing? Load the site and watch GA4 Realtime. A property created two weeks ago with 13 total users and no Realtime hit is not "collecting slowly" — it is broken.
• Conversion actions. Are conversions recording at all? Zero conversions across thousands of clicks is a tracking failure until proven otherwise, never a "bad market".
• Primary actions. Exactly one primary action per real business outcome (lead, purchase, call). A GA4 event can be marked primary for its goal and still not be the account-default conversion — check both.
• Attribution model. Data-driven where volume allows. Last click on a secondary action is fine only if it does not steer bidding.
• Conversion window. Match the real sales cycle. 90-day click-through suits machinery or B2B; 1-day click on a lead form throws away most of the pipeline.
• Conversion value. Ecommerce: required. Lead gen: a flat default value of 1 on every lead is worse than no value — it tells Smart Bidding all leads are identical, and makes ROAS reporting fiction.
• Ghost conversions. No page views, scroll depth or anti-fraud pings marked primary.

Thresholds: last conversion recorded more than 3–4 days ago on an active account = broken tag. Two primary actions tracking the same purchase = doubled ROAS.

From the example: two lead actions enabled, both with a default value of 1, one on a 90-day window and one on a 1-day window, GA4 linked but the imported event excluded from the Conversions column — and 0 conversions on €3,644 of spend. Everything else in this audit was measured against data that did not exist.

Block 2 — Audiences and data collection

• Remarketing lists created and actively collecting? A list attached to nothing is not a list.
• GA4 audiences for: all visitors, product/service page visitors, form starters, converters.
• GA4 ↔ Google Ads link active, and imported events actually counting.

Threshold: no audience with a growing member count = no remarketing, no RLSA, no Customer Match. Warm traffic is being lost permanently.

Block 3 — Account settings and hygiene

• Auto-tagging on. Without it, GA4 reports paid traffic as direct.
• Linked accounts: GA4, Search Console, Merchant Center (if ecommerce).
• Account-level placement exclusions for mobile apps and games — cheap insurance before you ever launch Display, Video or PMax.
• Brand campaign exists if branded demand exists. Otherwise competitors buy your name at the bottom of the funnel and your brand CPA hides inside your generic CPA.
• Change history, last 90 days. Sudden performance shifts almost always match a change nobody mentioned.

Threshold: a change history that hits the 500-row cap in 30 days means the account is being rebuilt mid-flight — no performance comparison before and after that date is valid.

From the example: 390 creations, 107 removals and 3 updates in one month, capped at 500 rows. Any "the account got worse in July" conclusion would have been noise.

Campaign level

Block 4 — Targeting settings

• Geo targeting type: Presence, not "Presence or interest". The second option quietly buys traffic from people who merely searched about your country.
• Geo coverage: does the location list match where the business actually sells and delivers?
• Language targeting: only enable a language if ads, landing pages and lead handling support it. Three languages in one campaign with mixed-language RSA assets is a relevance problem, not a reach win.
• Network settings: Search Partners and Display expansion off on Search campaigns unless deliberately tested.
• Ad schedule: 24/7 is fine only until you have conversion data by hour. Then it is a decision, not a default.
• Device bid adjustments: based on actual data, never on a hunch.

Threshold: any device taking more than ~85% of spend deserves its own landing-page review before any bid modifier.

From the example: mobile took €3,289 — 90.2% of spend — with desktop at €284 and tablet at €71. With below-average landing-page experience on most keywords, the mobile page was the single highest-leverage fix in the account.

Block 5 — Auction competitiveness

Pull Search impression share, Lost IS (rank) and Lost IS (budget) together. They tell three different stories:

• Lost IS (budget) > 20% on a profitable campaign — you are leaving money on the table. Raise budget.
• Lost IS (rank) > 40% — budget will not help. This is Quality Score, ad relevance and landing page.
• Both high — fix rank first. Adding budget to a low-rank campaign just buys more expensive bad clicks.

From the example: 25.4% impression share, 47.4% lost to rank, 27.2% lost to budget. Recommendation: do not raise the €100/day budget; fix relevance first.

Block 6 — Budget and bidding strategy

• Does the strategy match the goal? Maximize Clicks buys traffic, not leads. It is a legitimate starting strategy while tracking is broken — and a permanent problem if nobody revisits it.
• Smart Bidding needs roughly 30–50 conversions per campaign per month to be stable; below ~15 it is guesswork.
• tCPA/tROAS values derived from business math, not from a blog post.
• Any strategy changed in the last 14 days is still learning. Do not judge it.

Rule: never switch to conversion-based bidding, and never scale spend, until Block 1 passes. In the example account, both were explicitly put on hold.

Block 7 — Campaign-level audiences

• Observation-mode segments attached for data collection.
• A separate remarketing campaign once lists have volume.
• Audience bid adjustments only once observation data shows a real difference — with unreliable tracking, no modifier is the correct answer.

Block 8 — Landing page and post-click

Auditing ads without the destination is half an audit.

• Mobile load time under ~3 seconds.
• Message match: ad headline ↔ page H1 ↔ CTA.
• Does the form or checkout actually fire the conversion tag? Test it yourself, end to end.
• Check the landing-page experience component in the keyword Quality Score breakdown — "Below average" across many keywords is a page problem, not a keyword problem.

Block 9 — Other formats (PMax, Shopping, Display, Video)

Only audit what is running — but note what is missing and whether it should be.

• Performance Max: asset groups themed by product or service (never one catch-all), audience signals set, search themes defined, all asset types supplied, ad strength Good+, final URL expansion controlled, brand exclusions applied so PMax stops absorbing branded Search conversions.
• Shopping: Merchant Center feed errors and disapprovals, title/image quality, campaign priorities, listing groups segmented by margin or category rather than "All products".
• Display: defined audiences instead of pure automatic targeting, placement exclusions reviewed regularly, frequency capping, responsive creative with full asset coverage.
• Video: subtype matching the goal (bumper for awareness, skippable for consideration), frequency cap.

If none of these are running, say so and state the condition for launching them — usually "after tracking is fixed and first-party audiences have volume".

Ad group level

Block 10 — Keywords and match types

• Structure: brand separated from generic. Ad groups themed tightly — model-by-model, service-by-service. Generic category keywords sitting inside a model-specific ad group will pull traffic for every other model you sell.
• Match types: phrase and exact by default. Broad match works only with a clean conversion signal plus a maintained negative list — which means it is disqualified whenever Block 1 fails.
• Quality Score: count enabled keywords at QS ≤ 4 and read the three components. If landing-page experience is "Below average" across all of them, you have one fix, not thirty.
• Zero-impression keywords over 90 days: remove. They add noise, not coverage.
• Irrelevant keywords: flag by intent mismatch, not by zero conversions — with broken tracking, "no conversions" means nothing.

From the example: 5 ad groups by machinery model, no broad match anywhere (good), but 28 enabled keywords at QS ≤ 4, all with below-average landing-page experience, plus generic category terms leaking across model groups.

Block 11 — Negative keywords and search terms

The highest-ROI 20 minutes in any audit. Sort the last 30–90 days of search terms by cost, filter conversions = 0, and read the top 50 rows.

Sort what you find into three buckets:

1. Waste — wrong intent, wrong geography, job seekers, "free", "diy", competitor complaints, other manufacturers' models you do not sell. These become phrase negatives.
2. Opportunity — converting or clearly-qualified queries not yet in your keyword list. Promote to exact match in the right ad group.
3. Ambiguous — monitor, do not act.

On the negative side, check:

• Negatives applied at account level, not re-created ad hoc per campaign.
• Duplicates across shared lists — a bloated, overlapping list is unmaintainable and hides mistakes.
• Negatives blocking your own converting terms. A broad -free will block "free shipping".
• Brand leakage: your brand appearing in generic campaigns inflates CPCs; other brands appearing where you cannot serve them is pure waste.

Threshold: in an unmanaged account, 15–35% of spend typically sits in queries the advertiser would never buy. That number is your headline finding.

From the example: 336 campaign negatives, 57 ad-group negatives, 4 shared lists with 114 criteria — extensive, but zero account-level negatives, duplicates across lists, and irrelevant queries still arriving the day after a full restructure. Coverage is not the same as control.

Our search term analysis tool does this classification against your live account data, and the negative keyword generator turns bucket one into an importable list.

Block 12 — Ads and assets

• RSA completeness: all 15 headlines and 4 descriptions, every ad group with at least two ads, ad strength Good or better.
• Pinning: pin only what legally must appear. Over-pinning removes Google's ability to test.
• At least one unpinned headline per RSA so combinations can actually rotate.
• No mixed languages inside one RSA — otherwise Google will happily serve a German headline with an English description.
• Assets: sitelinks (4+ per campaign), callouts, structured snippets, call asset, images, lead form. Check none point to 404s or redirect chains.
• Copy quality: unique per URL and USP, no expired promotions still live.

From the example: all five RSAs complete and approved with unpinned assets — but German and English mixed inside the same ads, no callout assets at all, and generic export/trade copy reused across model-specific ads.

Turning the checklist into a report

After the 12 blocks, compress everything into 5–7 recommendations ordered by money recovered per hour of work. From the example account, the whole audit reduced to:

1. Fix conversion tracking and record qualified lead submissions as the primary conversion.
2. Freeze budget and bidding until reliable conversion data exists — no conversion-based bidding, no scaling.
3. Restructure keywords: separate model-specific from generic terms, each pointing at the most relevant page.
4. Tighten search-term control and consolidate the overlapping negative lists.
5. Rebuild the mobile landing-page and enquiry experience — 90% of spend, below-average page experience.
6. Separate German and English traffic; fix copy inconsistencies.
7. Improve Ad Rank through relevance before touching the budget.

Seven lines. That is the deliverable a business owner reads and acts on.

How often to audit

• Weekly: search terms, budget pacing, disapproved ads, conversion volume sanity check.
• Monthly: blocks 1–3 and 10–12.
• Quarterly: full 12 blocks — structure, bidding, formats, landing pages, offer.

Run it automatically

Manually, this checklist takes an experienced specialist two to four hours per account. ppcfy.ai's free Google Ads audit connects with read-only access, pulls the same signals through the Google Ads API — conversion actions, impression share, Quality Score components, search terms, negatives, assets — and returns a scored report with prioritised fixes in minutes.

If you would rather drive it from a chat window, our Google Ads MCP server lets Claude or ChatGPT query your own account directly and work through this checklist block by block.

And if the report surfaces something you would rather hand off, you can get matched with a vetted PPC specialist straight from the results.

Related reading

• How to audit a Google Ads account for free (15 minutes)