Google Ads Audit Checklist: A 50-Point Template for 2026

Most Google Ads audit checklists fail for the same two reasons. They tell you to "review your conversion tracking" without saying what a broken setup actually looks like, so the checklist only works for people who already know the answer. And they were written against a Google Ads interface that no longer exists — half the published templates still send you to menus Google retired, or list attribution models Google no longer supports.

This is a working template, not a reading list. Fifty numbered checks, grouped by audit area. Every check states where to look in the current interface and what "bad" looks like — an explicit fail condition, so two different people auditing the same account reach the same verdict. Every check also carries a verification flag: whether an AI agent can settle it from the Google Ads API, whether it needs API data plus human judgment, or whether a human simply has to go look.

That last column is the part we have not seen anywhere else, and it comes from building the thing. Soku ships Google Ads auditing as agent skills, so we had to decide, check by check, what a machine can actually prove. The answer is 32 of these 50 checks are fully machine-verifiable, 11 are hybrid, and 7 are human-only — which tells you exactly how much of your audit can be automated and where your attention is still the scarce resource.

If you want the method behind the thresholds — how spend-weighted Quality Score is computed, how search terms get classified, how creative fatigue is confirmed — this template deliberately does not repeat it. Those live in the deep dives, and each group below links to the right one. Start with the step-by-step guide to auditing a Google Ads account with an AI agent if you want the full narrative version first.

How to use this template

Set the window before you start. Anchor on the last complete day (yesterday in the account's reporting time zone), and use a 30-day diagnostic window with the previous 30 days as comparison. Partial days and lagging conversions produce false failures more often than real problems do.

Score each check PASS, WARN, FAIL, or UNVERIFIED. UNVERIFIED is not FAIL. If an API call did not return a field, or you could not get access to the analytics property, mark it unverified and say what a human needs to look at. Audit scores inflated by missing data are how audit reports lose credibility.

Fix in dependency order, not in checklist order. The numbering is for navigation. The fix order is: measurement first (group B), then stop active bleeding (group F), then structural fixes (groups D, E, G), then reallocation (group I). Changing bids on top of double-counted conversions just moves the error around.

Run each group on its own cadence. A quarterly audit is a labor constraint, not a best practice — and once an agent does the data pulls, the constraint disappears.

In the tables below, the Verify column reads Auto (an agent can settle it from API data alone), Hybrid (API data plus a human decision), or Manual (no API surface — someone has to look).

Group A — Access, billing, and audit trail (checks 1–4)

The group everybody skips, and the group that produces the most alarming findings. You are auditing who can change the account and what has already been changed to it.

Group B — Conversion tracking and measurement integrity (checks 5–12)

If measurement is wrong, every downstream finding is noise, which is why this group is scored before anything else. It is also the most commonly broken area in audited accounts — one review of audited accounts found conversion tracking errors in roughly 70% of them.

Group C — Campaign settings and targeting (checks 13–18)

Silent leaks. Each of these is a single toggle that nobody chose deliberately and everybody pays for.

Group D — Account and ad group structure (checks 19–22)

Structure sets the ceiling for everything else. A messy account can be optimised, but it cannot be optimised past its structure.

Group E — Keywords, match types, and Quality Score (checks 23–27)

Four of these five are single numbers with published thresholds, which is exactly what you want in a checklist. The method behind them — why the average is spend-weighted, why the two QS cutoffs differ — is in the Google Ads Quality Score audit; here you just need the fail condition.

Group F — Search terms and negative keywords (checks 28–32)

Usually the largest recoverable dollars in the account, and the only group that deserves a daily cadence. The classification logic — the semantic veto, the isolate-instead-of-negate rule, the Lin-Rodnitzky ratio — is covered in the Google Ads search terms audit.

Group G — Ads, assets, and extensions (checks 33–38)

Every check in this group is a count or a status — no taste involved. That is why it is entirely machine-verifiable. Fatigue detection, which is genuinely harder, is covered in the Google Ads creative audit.

Group H — Landing pages and post-click (checks 39–41)

The smallest group and the most human. Two of three checks have no API surface at all, which is precisely why they get skipped in automated audits and why they are on this list.

Group I — Bidding and budget (checks 42–45)

The group most affected by a change landing this year — see check 44.

Group J — The 2026 automation layer (checks 46–50)

This is the group that dates a checklist fastest, and the group most 2025-vintage templates do not have at all.

What a machine can actually prove: 32 auto, 11 hybrid, 7 manual

Here is the part we had to work out by building it rather than by reading about it. When you implement an audit as an agent skill, every check forces a decision: is there an API field that settles this, or is the agent about to invent an answer?

Three things fall out of that distribution, and they change how you staff an audit.

Counting checks automate completely; judgment checks never do. Group G is 6 for 6 auto because every check is a count or a status field — headline counts, disapproval status, sitelink counts. Group H is 0 for 3 because "does the page deliver on the ad's promise" has no field. The line is not difficulty, it is whether Google exposes a fact or you have to form an opinion.

Hybrid checks are where audits get faked. A hybrid check is one where the API gives you the data and a human supplies the business context: is this micro-conversion deliberate, is this competitor campaign strategy, is this location exclusion correct for our service area. An agent that answers these alone is guessing, and a human answering them without the data is guessing differently. The discipline that matters is that the agent reports the evidence and marks the decision as human-owned — Soku's audit skills only assign a root cause when Google's own field explicitly says so, and mark everything else UNVERIFIED rather than counting it as a failure.

The 7 manual checks are disproportionately expensive when they fail. Stale Admin access, a missing consent banner, a landing page that 404s on mobile — none are API-visible, and all of them are worse than any Quality Score problem in the account. The practical consequence: automate the 32, present the 11 with evidence attached, and calendar the 7. Do not let a "100% automated audit" pitch talk you out of the seven checks a machine cannot see.

Run this checklist as an agent prompt

The whole point of a machine-readable checklist is that you can hand it to a machine. If you have an agent connected to the Google Ads API — through the Google Ads integration, or your own setup via the Google Ads MCP guide — this prompt runs the auto and hybrid portions of the template above:

Audit Google Ads account {ACCOUNT_ID} against the 50-point checklist.

Window: 30 complete days ending yesterday in the account time zone,
with the prior 30 days as comparison. Never use partial days.

Baseline: resolve a CPA baseline per campaign in this order —
(1) the campaign's target CPA if set, (2) the campaign's actual CPA
over the window, (3) N/A. If N/A, suspend performance checks for that
campaign rather than scoring against an invented number.

For each of the 50 checks, return: check number, PASS / WARN / FAIL /
UNVERIFIED, the evidence (the exact API field and value you used),
severity, and the affected spend. Rules:
- Assign a root cause only when an API field explicitly states it.
- Mark a check UNVERIFIED, never FAIL, when a field is unavailable.
- Take every cost figure from the API cost fields; estimate nothing.
- Skip checks for campaign types the account does not run, and say so.

Then output: (a) the scorecard rolled up by group, (b) a fix list
ordered measurement -> waste -> structure -> reallocation, and
(c) the 7 manual checks as a task list for a human.

This audit is read-only. Do not pause, edit, bid, or write anything.

Two things make this produce an audit rather than an essay: the evidence requirement (every verdict cites the field it came from) and the read-only constraint. Auditing and acting are separate jobs, and mixing them is how an agent ends up grading its own homework. When an audit does produce executable changes — a negative keyword list, most often — the write belongs behind an explicit human approval gate, presented as a full proposal table first.

Why 2025-vintage checklists are already wrong

Five changes in the last year invalidate specific lines in most published templates. If you are adapting an older checklist rather than starting from this one, patch these first.

The general lesson is worth more than the specific patches: a Google Ads checklist has a shelf life of roughly a year, and the parts that expire first are the automation surfaces. The structural checks — is the ad group tight, is the conversion action the right one, does the landing page deliver — barely move.

Where to go deeper

This template is the run sheet. Four companion guides carry the method behind it:

• How to audit a Google Ads account with an AI agent — the full eight-step narrative version, including the fix-order logic and how an agent runs each step versus a human.
• Google Ads Quality Score audit — how spend-weighted QS is computed, why checks 25, 26, and 27 use three different cutoffs, and the per-keyword root-cause diagnosis behind them.
• Google Ads search terms audit — the semantic veto, the isolate-rather-than-negate rule, the Lin-Rodnitzky ratio, and where each negative keyword should live.
• Google Ads creative audit — the fatigue detection chain behind group G, and how to decide between kill, refresh, and watch.

If you want the checks run rather than read, connect an account through the Google Ads integration and ask for an account audit — the skills apply the same thresholds listed here, read-only, with the evidence attached to every verdict. For bulk fixes after the audit, Google Ads Editor is still the fastest way to apply structural changes offline.

FAQ

How long does this 50-point audit take?

By hand, six to ten hours for an account of any size — most of it exporting and cross-referencing. With an agent doing the 32 automated checks and pre-filling the 11 hybrid ones with evidence, the data work compresses to minutes and the human time collapses to reviewing findings plus the 7 manual checks. That is what makes a monthly cadence realistic instead of an annual one.

Do I have to run all 50 checks every time?

No, and the cadence table above is the point. Group F deserves a daily sweep, group I weekly, groups B, C, E, G, and J monthly, and groups A, D, and H quarterly. Running everything quarterly is the pattern that lets three months of search term waste accumulate between passes.

Which checks matter most if I only have an hour?

Checks 5 through 9 (is the measurement trustworthy), check 28 (how much spend is going to queries that will never convert), check 33 (is any ad group dark because of a disapproval), and check 43 (is a profitable campaign starved of budget). Those four cover the highest-severity failures we see in practice.

Does this checklist work for Performance Max-only accounts?

Groups A, B, C, H, I, and J apply in full. Groups E and G partially — Performance Max exposes no keyword Quality Score, and its assets are audited as asset groups rather than RSAs. Group F applies through PMax campaign-level negatives and the PMax search terms report rather than the classic search terms view.

Can I give this to an agency as an audit brief?

Yes, and that is a good use for it. The fail conditions are the reason it works as a brief: "review the account structure" is unfalsifiable, while "no enabled ad group may have zero enabled RSAs" is something a deliverable either satisfies or does not. Ask for the scorecard format — verdict, evidence, severity, affected spend, per check.

What is the difference between this and a free Google Ads audit tool?

Free audit tools score a subset of the automatable checks and are usually optimised to surface enough red to justify a sales call. The distinction worth holding onto is the one in the verification table: any tool can count your sitelinks, and no tool can tell you whether your competitor campaign is deliberate. Judge an audit by whether it shows its evidence per finding and admits what it could not verify.