I have audited a lot of Google Ads accounts over the past twelve years, both for clients and as a test task when I am hiring. If you are interviewing for a paid search role, there is a good chance someone hands you a messy account and asks what you would change, so a clean, repeatable audit is one of the most useful skills to build early.
The goal of an audit is not to list every tiny thing that is wrong. It is to find the few changes that will actually move results, then explain them clearly. Below I walk you through the exact order I check things, why that order matters, and how to turn what you find into a prioritized fix list.
Start with conversion tracking, always
Before you judge anything else, you have to trust the data. I have seen accounts that looked like they were failing, when really the conversions were double counted or not firing at all. If tracking is broken, every other number you read is fiction, so this is step one and it is not optional.Open Tools, then Conversions. Check that the actions make sense for the business, that they are recent (look at the last conversion date), and that nothing is set to a count that inflates results. For lead gen, form submits and calls should usually count once per click. For ecommerce, purchases should pass real revenue values, not a flat number.Watch for the classic mistakes: a primary conversion that is really a soft action like a pageview, conversions imported from Analytics that overlap with a Google Ads tag, or attribution windows nobody set on purpose. If you can, confirm the tag is live with Google Tag Assistant, then move on.Review the account structure
Once the data is trustworthy, look at how the account is organized. Good structure makes everything easier to manage, and bad structure quietly wastes money. A common red flag is one giant campaign trying to do everything, or the opposite, dozens of tiny campaigns that spread data so thin the bidding never learns. Brand and non-brand traffic should almost always be separated, because brand searches convert cheaply and will make non-brand look better than it really is if they share a campaign.Inside campaigns, check that ad groups are tight and themed. A group stuffed with unrelated keywords cannot have copy that matches every search, and that hurts relevance. You do not need one keyword per ad group, but each group should cover one clear idea so the ads can speak directly to it.Hunt for wasted spend in the search terms report
This is where I usually find the fastest wins. The search terms report shows the actual queries people typed before your ad showed, which is different from the keywords you bid on, and broad match especially can pull in searches you never intended to pay for.Sort by cost and read the terms with high spend and zero or few conversions. You will often spot irrelevant queries, the wrong intent (someone looking for free, jobs, or a competitor by name), or searches that belong to a different product. Each is a candidate for a negative keyword, and a good negative list is one of the cleanest ways to recover budget.While you are there, look at match types too. If most spend sits on broad match without a tight negative list and a conversion-based bidding strategy to guide it, that is a warning sign. Broad match can work well, but only when the account has enough conversion data and strong negatives keeping it on track.Check Quality Score, bidding, and budgets
Add the Quality Score columns (expected clickthrough rate, ad relevance, and landing page experience) to your keyword view. Quality Score is a diagnostic, not a target you chase directly, but the three components show where the friction is. Low ad relevance points to copy that does not match the keyword. A poor landing page experience points to a page that is slow, off topic, or weak on the offer.Next, check whether the bidding strategy fits the goal and the data. Target CPA and Target ROAS need a steady flow of conversions, so putting them on a brand new campaign with three conversions a month rarely goes well. If the account has volume, smart bidding usually beats manual. If it is thin, Maximize Conversions or a spell of manual control can be safer while data builds.Then review budgets across campaigns. Are the campaigns that drive efficient conversions capped by budget while weak ones spend freely? Check the impression share lost to budget, because moving money from a poor performer to one limited by budget is often a bigger lever than any single keyword tweak.Review ads, extensions, and build the fix list
Look at the ads themselves. Each ad group should have at least one strong responsive search ad with headlines that use the main keyword and a clear, specific offer. Check the asset strength rating, but read the copy too, since a Good rating with vague headlines is worse than it looks. Make sure assets (formerly extensions) are in place: sitelinks, callouts, structured snippets, and call assets. Missing assets is free real estate left on the table.Now pull everything into a prioritized fix list, the part interviewers actually care about. I sort findings by impact and effort, and high impact with low effort goes first. Here is the checklist I run through and turn into that list:- Conversion tracking: confirm it is accurate and counts the right actions before trusting any metric.
- Structure: brand separated from non-brand, tight themed ad groups, sensible campaign splits.
- Search terms: add negatives for wasted queries, fix match type and intent mismatches.
- Quality and relevance: improve low ad relevance and weak landing pages.
- Bidding: make sure the strategy fits conversion volume and goals.
- Budgets: shift spend toward campaigns limited by budget that convert well.
- Ads and assets: strong responsive search ads plus a full set of extensions.
Key takeaways
- Check conversion tracking first, because every other audit decision depends on the data being accurate.
- The search terms report is usually where you find the fastest wins through negative keywords and match type fixes.
- Finish with a prioritized fix list sorted by impact and effort, with the expected result spelled out for each item.