If you only ever open one report in Google Ads, open this one. Not the campaign dashboard, not the keyword planner, not the recommendations page with its cheerful optimisation score. The search terms report, because it is the only place the platform tells you what people actually typed before they cost you money.
The distinction matters more than it sounds. Keywords are what you bid on. Search terms are the real queries Google decided were close enough. On a broad match campaign that gap is wide, and on Performance Max you only get a filtered view of it. Almost every small account we look at is spending somewhere between a quarter and a third of its budget on queries the owner would have blocked on sight.
Keywords are the bid. Search terms are the reality.
You add the keyword “court resurfacing”. Google decides that “how to resurface a tennis court yourself”, “court resurfacing jobs”, and “cheapest court paint” are all near enough. All three cost you a click at the same bid. None of them is a customer.
This is not Google behaving badly, it is broad and phrase matching doing what they are designed to do. The system is built on the assumption that you will read the terms report and tell it what you do not want. If you never do, you have effectively handed over the targeting decisions and kept the invoice.
How to pull the report
- In Google Ads, go to Campaigns, then Insights and reports, then Search terms.
- Set the date range to the last 14 or 30 days. Anything shorter is noise.
- Add the columns you will actually use: cost, clicks, conversions, cost per conversion.
- Sort by cost, descending. Not by clicks, and definitely not by impressions.
- Download it if you prefer a spreadsheet. For anything over about 200 rows, you do.
For Performance Max and Demand Gen, the equivalent lives under Insights as search themes and search terms insights. It is aggregated and less complete than the standard search report, which is a real limitation rather than something you can configure away. Losing that visibility is part of the broader trade the platforms are asking us to accept, which is the theme of how AI is changing digital marketing. You still get enough to add negatives, and account-level negative keyword lists do apply to Performance Max.
Every term falls into one of five buckets
Do not read the report line by line looking for inspiration. Bucket it. The whole exercise takes twenty minutes once you stop deliberating over individual rows.
- Matches what you sell. Leave alone. If it converts and is not already a keyword, promote it.
- Right words, wrong intent. Research, comparison and definition searches. Sometimes worth keeping at a lower bid, usually not.
- DIY, free and how-to. Someone who wants to do it themselves. Block, unless you sell to that person.
- Competitor and brand confusion. Decide deliberately. Bidding on competitor terms is a strategy; doing it by accident is a leak.
- Jobs, courses and noise. Block without thinking about it. The word jobs alone usually pays for the exercise.
Negative keywords: match types, and where to put them
Negatives have their own match types, and they do not behave like positive ones. A negative phrase match will not block a misspelling, and a negative broad match requires every word to be present, not any of them. This is the part people get wrong, so it is worth the table.
| Negative match type | Blocks | Use it when |
|---|---|---|
| Broad | Queries containing all of the words, in any order | You want a concept gone, like free or diy |
| Phrase | Queries containing the exact sequence | You want a specific phrase gone, like how to make |
| Exact | Only that precise query | You are removing one high-cost term surgically |
Where you add them matters as much as how. Single terms go at campaign or ad group level so you keep the flexibility. Recurring themes go into a shared negative keyword list applied across the account, so that adding one word protects every campaign at once. Build one list per theme rather than a single sprawling one: jobs and careers, DIY and free, competitors, unrelated products. When you inherit an account, this is the first thing worth checking, because a good list set is the difference between fifteen minutes a week and two hours.
One warning about exclusions. Negatives do not apply retroactively and they do not rescue a bad landing page. If the term is right and the click still does not convert, the problem is downstream, and the fixes are in the landing page checklist for paid ads rather than in the terms report.
The weekly routine
Two steps in that list get skipped and both cost you money. Promoting winners is the upside half of the job: a search term that converted twice and is not yet a keyword is a term you are bidding on by accident and could be bidding on deliberately, with its own ad copy and its own landing page. And logging what you changed is what lets you tell in four weeks whether the work helped, rather than guessing.
A worked example, in five rows
Here is what one afternoon of this looks like on a services account with a modest monthly budget. Five rows from the top of the cost-sorted report, and the decision each one gets.
| Search term | What it is | Decision |
|---|---|---|
| court resurfacing cost | Buyer, researching price | Keep. Promote to its own keyword. |
| how to resurface a tennis court | DIY intent | Block: phrase negative on how to resurface. |
| tennis court resurfacing jobs | Job seeker | Block: broad negative on jobs, add to shared list. |
| [competitor] court resurfacing | Competitor brand | Decide. Usually block unless the comparison page exists. |
| pickleball court painting near me | Adjacent service you do offer | Keep. New ad group, matching landing page. |
Note that two of the five decisions are not blocks. That is the ratio to aim for. A terms review that only ever subtracts is half a review, and the additions are where the growth is: an adjacent service you already deliver but never advertised, showing up in your own report, is the cheapest new campaign you will ever launch.
Four ways to do this badly
- Blocking everything that did not convert. Twenty clicks and no sale is not evidence at a small budget. Give a term enough volume to mean something before you kill it.
- Adding negatives at account level in a hurry. One careless broad negative can silence a campaign that was working. Add narrow, add high in the hierarchy only for real themes.
- Negating your own long tail. The odd-looking four-word query with one conversion at half your usual cost per acquisition is the exact thing you were hoping to find.
- Doing it once. The report refills. New queries appear every week, and a quarterly cleanup means eleven weeks of leakage between visits.
What the exercise is actually worth
If a third of spend is going to the four blockable buckets and you recover half of that, you have cut your effective cost per acquisition by close to twenty percent without touching bids, budgets, creative or landing pages. That is a larger gain than most of the tactical changes people spend their week on, and it comes from reading one report properly.
It also changes the budget question. Once you know what share of search spend is real, you can compare channels honestly instead of comparing a clean Meta account against a leaky search one, which is the comparison we walked through in how a small business should split its budget between Meta and Google.
And if the weekly thirty minutes is the part you will not sustain, automate the pull rather than the judgement. A scheduled export into a sheet, with a rule that flags any new term above a cost threshold, turns the job into reviewing a short list. That pattern, human decisions on a machine-prepared list, is the one that survives contact with a busy week, and there are more examples of it in AI automation ideas that save small businesses hours.
Common questions
How often should I check the search terms report?
Weekly while you are actively spending, monthly on a stable account with a small budget. The trigger to look sooner is any sudden change in clicks or cost per conversion, because a new bad term entering the mix is one of the most common causes.
Why do I see fewer search terms than I used to?
Google stopped showing queries that do not meet a privacy volume threshold, so a slice of low-volume searches is permanently invisible. You still see the terms that cost real money, which is where the work is. Do not go looking for a setting to change; there is not one.
Can I do the same thing for Performance Max?
Partly. You get search themes and a filtered terms view rather than the full list, and you manage exclusions through account-level negative keyword lists. It is less precise than standard search, so lean harder on your negative lists and on watching cost per conversion by asset group.
Should I bid on competitor names?
It can work when your offer genuinely compares well and your landing page addresses the comparison, and it is expensive when it does not. The important thing is that it is a decision. Accidental competitor traffic, caught in the terms report and never reviewed, is just a subsidy. For local services in particular, a share of that demand is cheaper to win organically, which is what Google Business Profile optimisation is for.
Do negative keywords lower my quality score?
No. Blocking irrelevant queries usually improves the metrics that feed quality score, because your click-through rate and relevance improve once the bad impressions stop. There is no penalty for having a large, well-organised negative list.
