The Google Ads audit checklist we run on every account we take over: 15 checks across measurement, structure, settings, bidding, creative and landing pages, ranked by what each…
Every Google Ads account gets audited eventually. Sometimes it happens on your terms: a new quarter, a new budget, a healthy instinct to check the machine before feeding it more money. More often it happens under pressure, because performance has plateaued, a new marketing lead has inherited the account, or someone senior finally asked what the agency actually does every month.
Most audits produced in those moments are screenshot decks. Forty slides of interface captures, a traffic light on each one, and no answer to the only question that matters: where is the money going, and what single change would recover the most of it? This is the checklist we run inside our Ads engine when we open an account for the first time, in the order we run it, with the reasoning behind each check.
The core position
An audit is not a list of everything that is wrong. It is a ranked list of what each problem costs, and the ranking starts with the signal the account learns from, not the settings screen.
Set the frame first, because the frame decides what counts as a finding. Pull the last 90 days as your working window and a year for context. Write down what a customer is worth, what the business considers a qualified lead, and what the account is expected to deliver. An account spending £20,000 a month against a £15 cost per lead target gets audited differently from one protecting a brand at £2,000 a month.
Then resist the interface. Google’s Recommendations tab is not an audit; it is a sales channel with an optimisation score attached. Some recommendations are sensible, many exist to increase spend, and auto apply should be off before you do anything else. That is check zero.
Measurement comes first because every other number in the audit is downstream of it. If conversions are miscounted, the bidding data, the search term analysis and the creative tests are all built on sand.
Open the conversion summary and read every action set to primary. You are looking for inflation: page views counted as conversions, button clicks alongside the form submissions they lead to, imported goals duplicating native tags, or a legacy action from three agencies ago still feeding the algorithm. Smart Bidding optimises towards the sum of primary actions. If that sum mixes real enquiries with noise, the account is being steered by noise.
Lead generation actions should almost always count one conversion per click, not every. A visitor who submits the same form three times is one lead. Check the attribution model and conversion windows while you are there, and confirm the same conversion is not arriving twice through two routes, once from the Google tag and once from an import.
Trigger a test conversion, or at minimum inspect tag status and recent conversion timestamps. Accounts routinely run for months on a broken or half broken tag after a site change. This matters double after a rebuild or migration, which is why conversion checks sit near the top of our lead quality diagnostic as well: a tag that fires on spam teaches the algorithm to buy more spam.
If the business runs lead generation, enhanced conversions for leads should be configured and receiving hashed email data. In markets where consent mode applies, check it is implemented and that conversion modelling is active rather than silently dropping a share of conversions. The gap between accounts with and without this plumbing widens every year as third party signals decay.
The strongest accounts import offline outcomes: qualified lead, opportunity, closed won, with values by stage. The weakest optimise towards raw form fills forever. Ask for the connection between the CRM and the account. If the answer is a shrug, you have found the single highest value fix in the audit, because every bidding decision improves once the machine learns from revenue instead of activity.
Structure should mirror margin and intent, not the site map. Look for the classic faults: everything consolidated into one campaign so budget flows to the cheapest clicks rather than the best customers, brand and generic terms sharing a campaign so brand conversions subsidise generic waste, and ad groups holding forty keywords that could never share one relevant ad.
Read a full month of real queries, not the summary view. Classify what you see: right intent, wrong intent, ambiguous. With broad match and Smart Bidding everywhere, the search terms report is the clearest window into what the algorithm believes your business does. If more than a modest share of spend lands on queries that could never become a customer, quantify it. That number usually funds the rest of the audit’s recommendations by itself.
Healthy accounts have structured shared negative lists that grow weekly: job seekers, students and tutorials, freebie hunters, adjacent services the business does not offer. Unhealthy accounts have a few dozen scattered negatives added the last time someone complained. Check the date the newest negative was added. It tells you when someone last actually looked.
Segment performance by network. Search Partners and Display Expansion often show flattering cost per conversion numbers that sales can never validate, for reasons we cover in depth in the lead quality piece. If partner traffic converts on paper and evaporates on the phone, exclude it and watch quality recover.
Confirm location targeting is set to presence for any business that serves a defined area, then open the user location report and look for spend leaking into places the business cannot serve. Check excluded locations and language settings against reality rather than assumption. These are two minute checks that regularly explain a stubborn tail of useless leads.
Look at performance by hour, day and device against how the business actually operates. A lead form that nobody follows up over the weekend, a call campaign running while the office is shut, mobile traffic pouring into a desktop era landing page: none of these appear on any single settings screen, which is exactly why they survive for years.
Target CPA and target ROAS need conversion volume to learn from. An account pushing aggressive targets on thin data will oscillate between overspending and going quiet. Check conversion volume per campaign per month against the strategy it runs, look for target changes stacked too quickly in the change history, and confirm portfolio strategies are not blending campaigns with incompatible goals.
Identify campaigns limited by budget and ask whether the constraint is deliberate. Budget limited winners and unconstrained losers are the most common allocation fault in accounts we take over. The change history tells the rest of the story: an account with no meaningful changes for eight weeks is not being managed, whatever the monthly report says. What management should cost, and what it should include, is a separate conversation we cover in our breakdown of Google Ads agency pricing.
Review responsive search ads for actual persuasion: a specific offer, proof, a reason to act, keyword relevance in the headlines that users see. Ad strength labels are a rough guide, not a goal. Check that assets, the components formerly called extensions, are present and specific: sitelinks that go somewhere useful, callouts that say something true, structured snippets that match the offer. Then look for evidence of testing. If every ad was written on the same date a year ago, creative is not being managed.
Click every final URL in the account. You are checking message match between query, ad and page, one clear action above the fold, a form that works on a phone, and load speed that respects the visitor. Speed is a conversion input long before it is an SEO concern; the discipline is the same one behind bringing LCP under 1.5 seconds for a Shopify Plus store, and the revenue effect on paid traffic is immediate because you paid for every visitor who gives up.
A finding without a cost attached is an opinion. For each fault, estimate the monthly spend it touches and the severity, then sequence the fixes: measurement first, because everything else depends on it, then query and negative hygiene, then structure, then bidding targets once the signal is clean, then creative and landing pages as the compounding layer. A typical account needs a 30 day sequence, not a 40 slide deck.
Run this yourself and you will find most of what a professional audit finds. The difference is usually in the costing, the sequencing and the follow through, which is the job description of a Google Ads agency worth its fee: not reporting on the account, but compounding it.
Book 15 minutes and we’ll run this audit on your account, live.
Daniel runs paid media at Gyrodile, from account architecture to creative testing. He cares about qualified revenue, clean tracking and killing wasted spend.
More insights from the Gyrodile team