Add Negative Keywords From Search Terms Without Guessing

Why adding negatives one row at a time goes wrong

The search terms report shows one query per row, but a negative keyword applies to every query that contains it. That includes queries you have never seen. So "is this term bad?" is the wrong question. The right one is "what else will this negative block, and what is that worth?"

Most guides on this topic explain the clicks: tick the box, click Add as negative keyword, choose a campaign or a list. That part is easy, and Google's own help page covers it well. Then the guides say "add irrelevant terms" and stop, right before the step that causes damage: deciding how much of the term to block.

Say you sell standing desks and you see this row:

standing desk manual pdf — 40 clicks, $36, 0 conversions

You have at least four options:

• Block the exact query [standing desk manual pdf].
• Block the phrase "manual pdf".
• Block the single word manual.
• Block the single word pdf.

Each option blocks a different set of queries. An exact negative blocks almost nothing. That exact query may never come up again, so you paid for the lesson and bought no protection. A single word blocks a lot. Whether that's good or bad depends on what the word does in the rest of your account. The row you're looking at can't tell you.

The two mistakes don't cost the same

Blocking too much costs more than blocking too little, and it is much harder to spot. A query you didn't block comes back next week. It costs a few dollars, and you catch it on your next review. A negative that blocks buyers produces nothing at all: no rows, no clicks, no warning. The conversions just stop coming. The search terms report only lists queries that triggered your ads, so it can't show what's missing.

That's why the rule below is careful on purpose. It lets a little waste through to avoid the silent, expensive mistake.

The rule: price the word across every query it would block

Before you add a word or phrase as a negative, filter your search terms for every query that contains it. Use at least the last 90 days and read the totals row. If none of those queries converted, the word is safe to block on its own. If any did, it is too broad. Add a word and test again.

The totals give you three numbers:

1. Spend on the matching queries. This is roughly what the negative would have saved.
2. Conversions and conversion value on the matching queries. This is what the negative would have cost you.
3. How many different queries match. This tells you whether you're looking at a pattern or a one-off.

Here is the decision:

• Zero conversions, more than one matching query, and you can't picture a buyer typing the word → block the single word.
• Any conversions among the matching queries → don't block the word. Find the shortest phrase that matches the wasteful queries and none of the converting ones. Run the same test on that phrase.
• Only one matching query, with no pattern → add it as an exact negative if it cost enough to matter. If it cost $3, leave it and see whether it comes back.

Why 90 days, and why more is sometimes better

A short window misses buyers who take a while to convert, and seasonal queries that only show up a few months a year. Ninety days is the minimum we trust. In low-volume accounts, a word may only show up a handful of times in 90 days, so use six or twelve months. Leave out the last week or two if your conversion lag is long. Recent clicks may still be on their way to converting, and a "zero" there isn't a real zero yet.

How to run the check in Google Ads

Open Campaigns → Insights and reports → Search terms. Run it at the same scope where the negative will apply. If the negative goes into one campaign, look at that campaign. If it goes into a shared list used by six campaigns, select all six. Add the filter Search term → contains → manual, set the date range, and read the totals row.

A filter on "manual" also catches "manuals" and "manually". That's useful, because it shows you nearby words you need to decide on too. If you'd rather work in a spreadsheet, download the report and use a text filter. It's quicker when you're checking ten candidate words.

Performance Max search terms live in a different report and come with less detail. The check still works, but read what the PMax search terms report actually shows before you trust its totals.

A worked example: "manual" versus "pdf"

Back to the standing desk store. The row standing desk manual pdf looks like an obvious case for blocking manual. Here is what a "contains manual" filter returns over 90 days:

Blocking manual would save $169 over 90 days. It would also block six conversions. "Manual" is a product type here: a desk you raise with a crank instead of a motor. At an average order of $300, those six orders are worth $1,800. You'd give up $1,800 in sales to save $169. The waste comes from only two of the four rows, which add up to $58.

So manual fails the test. The next step is to find what actually makes the two bad rows bad. It isn't "manual". It's "pdf" and "download". Test each one:

Three different queries, $60, no conversions. Nobody looking for a PDF is about to buy a desk. pdf passes all three parts of the rule, so block it as a single word.

download matches desk manual download ($22) and standing desk app download ($9): $31, no conversions. Before you block it, ask whether a buyer could use the word. If you sell a smart desk with a companion app, someone who downloads the app is usually an existing owner. The answer is probably still no, so block it. If the app were a selling point people search for before they buy, you'd use the phrase "manual download" instead.

The two single-word negatives together block $91 of visible waste over 90 days. That's a small amount, and that's normal. The bigger value comes later: every future query containing "pdf" is blocked too, including ones you haven't seen yet. Blocking manual would have looked more decisive, and it would have quietly cost the account its crank-desk sales.

If you want to know whether savings like these matter for your account, calculating wasted spend honestly shows how to separate real waste from the normal cost of testing.

Choosing the match type after you've priced it

Pricing the word tells you how much to block. The match type is how you block exactly that much. In short: phrase match for single words and multi-word patterns, exact match for one-off queries, and broad match only when you have a specific reason.

Single words: phrase match

A one-word phrase negative such as "pdf" blocks any query containing that word, wherever it appears. That's what the test measured, so the negative does what you checked. We default to phrase for single words because it's predictable. Broad negatives behave less predictably across match types, and when a negative stays in an account for years, predictable is better.

Multi-word patterns: phrase match

When a single word fails the test, the fix is usually a two-word phrase. "manual pdf" blocks queries that contain those two words next to each other, in that order. Check the phrase the same way you checked the word: filter for "contains manual pdf" and confirm that no converting queries match.

One-off queries: exact match

Use exact match when the query is bad as a whole but each of its words is fine. Two common cases:

• A query that combines good words in a way that doesn't fit, such as standing desk for cats, where "standing" and "desk" appear in all your buyers' queries.
• Sending traffic to the right place. You block [electric standing desk] in a general campaign so it goes to the campaign built for it, without blocking anything else.

Variants are your job

Negative keywords don't extend to close variants the way positive keywords do. "instruction" doesn't block "instructions". If the filter shows both spellings, add both. Plurals and common misspellings are the usual gaps. Close variants and why exact isn't exact explains why positive and negative keywords behave so differently here.

Where to add the negative, and why it changes the test

The level where you add a negative decides which campaigns it covers, so it decides what you have to check. Price the word across every campaign it will touch, not just the one where you spotted the bad query.

Back to the desk store. Say it also sells a paid assembly service through a separate campaign. In the desk campaign, assembly only matches waste: standing desk assembly pdf, how long does desk assembly take. In the services campaign, desk assembly service near me converts. Add assembly at campaign level to the desk campaign and it's fine. Add it to a shared list that also covers the services campaign and you've switched off a product. Filtering only the campaign where you found the query would have hidden that.

What Google allows at each level:

• Search accepts negatives at ad group level, at campaign level and through shared negative keyword lists.
• Standard Shopping works the same way: negatives and shared lists.
• Performance Max accepts up to 10,000 campaign-level negative keywords but doesn't accept shared negative keyword lists. You add negatives to each PMax campaign directly, and you check each campaign's own terms.
• Display, Video and App don't report search terms, so there's nothing to check and nothing to add from this report.

For deciding between shared lists and campaign-level negatives in general, see account-level negative keyword lists vs campaign-level. The rule is the same either way: the scope of your check must match the scope of your negative.

What the check cannot see

The test only measures queries Google shows you, and only the past. Three gaps matter, and you should know about them before you treat a zero as proof.

Hidden search terms

Google doesn't report every query. Low-volume queries are grouped out of view for privacy, so the totals in your search terms report won't match the campaign's spend. A word with zero visible conversions might still appear in converting queries you can't see. The test can't fix that. What you can do is be more careful with words that sit close to your offer, and stricter with words that clearly sit outside it. "pdf" is outside. "manual" is close. Search terms hidden by privacy covers how big that gap can get and how to work around it.

Future queries

The report is history. A single-word negative also blocks queries nobody has typed yet. That's why the rule includes a judgement call: can you picture a buyer using the word? Take used. A store that only sells new desks might see used standing desk and used desk for sale and block used. But "best standing desk used by programmers" is a buyer, and so is "standing desk used for 8 hours a day". If those haven't come up yet, the totals look clean while the word is still risky. When the honest answer is "maybe", use the phrase ("used desk", "used standing desk") instead of the word.

Your own keywords

A negative can block a keyword you're bidding on, and Google doesn't always warn you. Before you add a single word, search your keyword list for it too, not just your search terms. Negative keyword conflicts Google Ads won't flag shows where these collisions tend to hide.

A routine that keeps working

The check takes a minute per word once you're used to it. The hard part is keeping it up. A simple routine:

1. Sort the search terms report by cost, highest first, for the period since your last review.
2. For each wasteful row, name the word that makes it wasteful. Not the whole query, and not the first word that looks odd. In standing desk manual pdf, it's "pdf".
3. Run the check on that word at the scope where you'll add it. Totals over at least 90 days.
4. Apply the rule. Single word, longer phrase or exact match.
5. Write down what you added and why, even in a one-line note. When conversions drop three months later, you'll want to know which negative to look at first.

How often to do this depends on spend and how much your queries change. How often to check search terms gives a schedule based on volume. Accounts that spend more need checks more often. One reason is that a bad negative costs more the longer it stays in place.

Then look back at your negatives now and then. Blocking buyers doesn't show up in the search terms report, so you have to look for it elsewhere: keywords whose impressions drop for no clear reason, or a product line whose conversions slowly fade. Six signs negative keywords are hurting performance lists what to look for. When you find a bad one, undoing negative keywords shows how to remove it cleanly.

Where PPCleaner fits, and where it doesn't

PPCleaner does this work every night. It reads the search terms your Search, Shopping and Performance Max campaigns paid for, judges which ones were never going to convert, and adds them as negatives. It follows the same cautious rule: it never blocks a term that converted in the last 90 days, and every block can be undone in one click for 90 days. It doesn't touch budgets, bids, ads or keywords.

It has the same blind spots you do. It can't see queries Google hides for privacy. It can't add search term negatives to Display, Video or App, because those campaigns don't report any. It can't put a shared list on Performance Max, because Google doesn't allow it. And it doesn't know your business the way you do. If "assembly" is a product you're launching next month, you know that and the data doesn't. That's why every block is listed and reversible, not hidden away.