How to read your Google Ads search term report: find wasted spend, add the searches that convert, build negative lists safely and set a review rhythm for your budget.
Your keywords are not the searches you pay for. With broad and phrase match, and to a lesser degree with exact match, Google shows your ads for searches it judges close enough in meaning, and its judgement is not always yours. The search term report lists what people actually typed before your ad appeared, which makes it the most direct record of where your money went.
Search term mining means reading that report regularly and acting on it: blocking irrelevant searches with negative keywords, adding the searches that convert as keywords of their own, and matching your ads to the way people describe their needs. It costs nothing but time. The money at stake is real: if you spend €6,000 a month and one euro in seven goes on searches that could never produce a customer (an illustrative share, not a benchmark), that is about €860 a month, or more than €10,000 a year.
This article covers what the report shows and what it hides, the three questions to ask in every review, how to spot patterns instead of arguing about single searches, how to build negative keyword lists without blocking good traffic, and how often to do all this for your level of spend. The methods apply most directly to Search and Shopping campaigns, where the report is most complete.
The report lists the searches that triggered your ads, with impressions, clicks, cost and conversions for each, alongside the keyword that matched them. Comparing the two tells you how widely Google is reading your keywords. A phrase match keyword such as "office cleaning" might match "office cleaning prices", which you want, and "office cleaning jobs", which you don't.
Two limits matter. For privacy reasons Google doesn't show every search: terms used by very few people are left out, so part of your spend always sits behind searches you can't see. And Performance Max reports search data differently from Search campaigns, so treat what it shows as a partial view and apply the same logic wherever you can.
Sort the report by cost and look at the period since your last review rather than all time. The searches that spent the most deserve your attention first, whatever their conversion rate, because that is where a single decision moves the most money.
A useful review asks three questions, in this order. Each leads to a specific action, so the review ends with changes in the account rather than notes in a document.
A threshold makes these decisions faster. If a search has spent one to two times your target cost per acquisition (CPA), what you are willing to pay for one customer or lead, without converting, decide: exclude it, or keep it deliberately and note why. If it has converted at or below your target, promote it to a keyword.
Converting searches also tell you how customers describe what they want. If "same day" or "near me" keeps appearing in searches that convert, those words belong in your headlines and on your landing page, not only in your keyword list. Your search term report is your customers telling you, in their own words, what they wanted.
Single searches are easy to argue about; patterns are easier to act on. Export the last 90 days of search terms into a spreadsheet, split every search into its individual words, and add up cost, clicks and conversions for each word. This is often called n-gram analysis, and it shows which words cost money across hundreds of searches, even when no single search spent much. The patterns usually fall into a few familiar groups.
A word like "free" shows why patterns need judgement. For a software company, "free template" is wasted money; for a furniture shop, "sofa free delivery" may be one of its best searches. Block the phrase that causes the problem, not every search that happens to contain the word.
Your search term report is your customers telling you, in their own words, what they wanted.
Negative keywords can be added to a single campaign or to shared lists that apply to several campaigns at once. Keep a shared list for searches you never want anywhere (job seekers, free versions, products you don't sell), and use campaign-level negatives to steer searches to the right place, for example to stop a generic campaign from taking your brand searches. In Shopping campaigns, which have no keywords, negatives are the main way to keep your products away from searches you don't want.
Negative match types behave differently from normal keywords, and this is where most mistakes happen. Check these four rules before adding anything in bulk:
Build lists from your own search term data rather than a generic list downloaded from the internet. A generic list contains words that are irrelevant to one business and valuable to another, and it can quietly block searches that would have converted. Before adding any negative, check whether the term has produced conversions in the last 90 days; if it has, it needs a closer look, not a block.
Review the lists themselves twice a year. Products, services and markets change, and a negative that made sense before you offered home cleaning becomes an expensive mistake the month you start.
The right rhythm depends on how much you spend and how much your campaigns rely on broad match. As a rule of thumb from day-to-day account work, not an official standard: below about €5,000 a month, review every two weeks; between €5,000 and €20,000, weekly; above €20,000, twice a week.
New campaigns need more attention. For the first few weeks of a new campaign, or after moving a campaign to broad match, check every two or three days until the share of irrelevant searches settles. Broad match combined with Smart Bidding, Google's automated bidding, can find valuable searches you would never have thought of, but the wider the net, the more often you need to check what is in it.
You'll know the rhythm is right when each review takes 20 to 40 minutes and produces a handful of changes. If every review produces dozens of new negatives, look more often or tighten your match types; if a month of reviews produces nothing, you can safely look less often.