The Google Ads audit checklist for 2026

This is the checklist we run when we look at a Google Ads account for the first time. It is organised by where wasted spend is usually found fastest, not by the order the tabs appear in the interface — which is why conversion tracking comes first and account settings come last. Use a 30-day window for volume checks and a 90-day window for trend checks, and keep one running note: every issue you find, with a rough monthly cost next to it.

Audit in waste-finding order: verify the data first, then follow the money. An account audit that starts anywhere other than conversion tracking is grading homework with the wrong answer key.

Before you start

Three decisions before you open a single report — they stop the audit sprawling into an afternoon:

• Fix your windows. Pick the date ranges now (30 days for cost checks, 90 for trends, year-on-year for seasonal accounts) and keep them constant across every section, or you will compare incomparable numbers.
• Write down what a conversion is worth. A lead value or blended margin per sale, even approximate. Without it you cannot rank findings by cost, and ranking is the whole point.
• Time-box each section. Ten minutes per section on the first pass. The goal is a ranked list of leaks, not a dissertation — you can go deep after you know where the money is.

Conversion tracking — do this first, everything else lies if it is broken

Every other number in the account — and every automated bid — depends on conversions being counted correctly. If tracking is wrong, the audit stops here until it is fixed.

• Check what is marked primary. In the conversion settings, confirm only genuine business outcomes (purchases, qualified leads) are primary. Page views, button clicks, and newsletter signups set to primary quietly train smart bidding to chase junk.
• Hunt for double counting. The classic: the same purchase imported from both the Google Ads tag and GA4, both set to primary. Reported conversions double, calculated CPA halves, and budgets get scaled into the illusion.
• Cross-check volume against reality. Pull 30 days of reported conversions and compare them with the CRM or order system. A persistent gap in either direction means attribution, consent, or tag placement needs attention.
• Confirm values are flowing, not just counts. For ecommerce, revenue per transaction; for lead gen, at least differentiated values per conversion action. Value-blind accounts cannot use value-based bidding, which is where most of the modern upside sits.
• Verify enhanced conversions are enabled and diagnostics are green. With cookie loss and consent requirements, unmodelled accounts under-report — and smart bidding under-bids on what it cannot see.
• Look at conversion lag before judging anything recent. Check the days-to-conversion distribution. If the typical lag is two weeks, the last two weeks of ROAS are not bad — they are incomplete.

Search terms and negatives — the fastest waste to find

The search terms report is where intent mismatch turns into an itemised bill. This section usually produces the biggest single line items in the audit.

• Sort last month’s search terms by cost and read the top 50. Highlight anything with meaningful spend and zero conversions. This ten-minute read is the highest-yield check in the entire audit.
• Judge intent, not just conversions. Job seekers, students, DIY researchers, competitor brand names you do not convert on, “free” qualifiers on a paid product — waste is often obvious from the query text alone.
• Check negatives actually exist — and are applied. Shared negative lists that are maintained but attached to nothing are common. Confirm each search campaign carries the right lists.
• Use account-level negatives for the universal exclusions. Terms that should never match anywhere (jobs, careers, login, support) belong at account level, not copy-pasted per campaign.
• Look for non-brand campaigns eating brand traffic. If generic campaigns are matching your own brand terms, performance reports flatter the generics while you pay auction prices for clicks the brand campaign — or the free listing — would have caught.
• Check for cross-campaign cannibalisation. The same query matching in multiple campaigns means your structure is bidding against itself; tighten negatives to route each intent to one place.

Performance Max visibility — open the black box

PMax can perform well while hiding how. The audit question is never “is the ROAS number good” but “where did the money actually go, and would I have won that revenue anyway?”

• Split spend by channel. Use the channel-level reporting to see how much budget went to Search versus Shopping versus YouTube and Display. A “great” PMax campaign that is 80% remarketing-flavoured Display is a different conversation.
• Check brand exclusions. Without them, PMax happily serves on your brand terms and books the revenue as its own. Confirm brand lists are attached — and that a dedicated brand campaign exists to catch that traffic properly.
• Review asset group structure. One asset group for the whole catalogue means one message for every audience. Group by product line or intent theme, each with its own assets and signals.
• For ecommerce, audit the feed before the campaign. Disapproved items, missing GTINs, weak titles, and stale prices cap PMax performance from below the waterline. (Running heavy PMax spend? Our ecommerce workflow page covers the asset-level analytics side of this.)
• Scan placement reports for junk. Pull where Display and video actually served. Made-for -advertising sites and low-quality app inventory are excludable once you can see them.

Keywords and match types

• Broad match only where smart bidding has data. Broad match paired with manual bidding or thin conversion volume is how exploratory clicks become a standing tax. Check every broad-match campaign has a value-based or conversion-based strategy and enough volume to steer it.
• Find duplicate keywords competing with themselves. Same keyword, same match type, in multiple ad groups or campaigns — Google resolves the conflict, and you pay for the ambiguity in relevance.
• Review your money keywords’ quality signals. For the twenty terms that spend the most, check the ad relevance and landing page components. A weak score on a high-spend term is a per-click surcharge you can actually fix.
• Check what match types are really matching. Read search terms per match type. Exact and phrase have loosened over the years; if “close variants” are drifting off-intent, negatives are the steering wheel.
• Confirm keyword-to-ad relevance. Each ad group’s keywords should be answerable by its ads. If an ad group needs three different headlines to cover its keywords, it is three ad groups.

Bidding and budgets

• Match the strategy to the goal and the data. Target ROAS with twelve conversions a month will oscillate; Maximise Conversions with no target on a tight budget will spend it all regardless. Check each campaign’s strategy against its conversion volume and its actual business goal.
• Compare targets with delivered reality. A tCPA set at half the delivered CPA is not a target, it is a throttle — usually one someone set and forgot. Reset targets from the last 90 days of data.
• Find winners flagged “Limited by budget”. A campaign beating your target while capped is free growth being declined daily. Either fund it or lower the constraint deliberately.
• Check the budget split against the performance split. List campaigns by share of spend next to share of conversion value. Money pinned to legacy campaigns “because it has always run” shows up immediately.
• Look at pacing anomalies over the month. Spend spikes, sudden CPC jumps, and days at zero are easy to miss after the fact — this is exactly what continuous anomaly alerts exist for between audits.

Ads and assets

• Every active ad group has a complete responsive search ad. Full headline and description slots, written for the keyword theme — not three headlines and a shrug.
• Treat Ad Strength as a coverage hint, not a KPI. Fix “Poor” where it flags genuinely thin assets; do not chase “Excellent” by unpinning copy that exists for compliance or message control.
• Audit pinning deliberately. Pinning everything recreates expanded text ads and starves the combinatorics; pinning nothing can bury the legal line. Know why each pin exists.
• Check asset coverage: sitelinks, callouts, structured snippets, images. These are free real estate in the auction. Missing image assets on Search is the most common gap in otherwise healthy accounts.
• Sweep for disapprovals and “eligible (limited)”. Policy limbo does not send you an email. Filter for it explicitly — entire ad groups sometimes idle for months this way.

Landing pages

• Message match, ad to page. The promise in the headline should be restated above the fold on the page. Users who have to re-orient bounce, and the auction charges you for their confusion.
• Test the top pages on a mid-range phone. Most Google Ads traffic is mobile. If the page takes more than a breath to become usable, you are paying search prices for abandonment.
• Click every final URL that carries spend. Redirect chains, out-of-stock pages, and outright 404s survive in mature accounts far longer than anyone expects.
• Verify tracking templates and UTMs are consistent. Broken templates silently sever the click-to-conversion chain — which loops you back to section one.

Account structure and settings

• Search partners and Display expansion: on purpose, or on by default? Check the network settings on every search campaign and judge each network on its own segmented numbers.
• Location options: presence versus interest. “People interested in your locations” quietly ships local budgets abroad. Most advertisers want presence.
• Review device performance — and old device modifiers. Legacy bid adjustments layered under smart bidding create fights nobody is refereeing. Segment by device and clean up.
• Check ad schedules against the conversion clock. If leads only get worked weekdays 9–5, decide deliberately what weekend clicks are worth, rather than by omission.
• Turn auto-apply recommendations off — or curate the list line by line. Some auto-applied “optimisations” add broad keywords and raise targets. Nothing in the account should change without a human deciding it should. (This is, unsurprisingly, a principle we build products around.)

Turning the audit into an action plan

An audit that ends as a document is a cost. It becomes an investment the moment it is a ranked queue of fixes with owners and dates.

• Rank every finding by monthly cost. Use the conversion value you wrote down at the start. A mis-set primary conversion usually outranks fifty small keyword tweaks — the ranking protects you from busywork.
• Fix the top three, then re-measure. Give changes two weeks or a full conversion lag, whichever is longer, before judging them. Batch small changes; isolate big ones.
• Put monitoring where the audit found rot. Every section that produced a finding is a place the account decays — alerts and a quarterly re-run keep the same leaks from silently reopening.

Or let an agent carry the checklist. Adszy runs this kind of analysis against your live account — it reads the data, ranks what each issue is costing you, drafts the fix, and applies only the changes you approve. You can ask the questions in plain English from chat, or run the read-only checks from Claude or Codex through Adszy MCP. Read-only by default either way: an audit should never change an account by itself.