Negative Keywords and Placement Exclusions: A Guide to Cutting Wasted Ad Spend

Some of your budget pays for clicks from people who were never going to convert. Finding and blocking those clicks returns more per hour than most optimization work, but few teams put time for it on the calendar.

How much spend goes to waste

Seer Interactive analyzed millions of data points across 30 paid search accounts. About 15% of budget went to irrelevant keywords: searches that matched a keyword but had nothing to do with what the advertiser sold. Seer Interactive estimated that across the whole industry, this adds up to billions of dollars spent on terms nobody wanted to pay for.

Other studies put the number higher. A well-known WordStream analysis found the average small-to-medium business spends around $1,200 a month on pay-per-click and wastes roughly 25% of it through poor account management, with missing negative keywords as the main gap. That study is a decade old, and the problem it describes has not gone away. Lunio’s 2024 report put global spend lost to invalid traffic at about $75 billion, up 33% from 2022. Lunio The ANA found that only 43.9% of programmatic budgets reach consumers as quality impressions, and $26.8 billion is lost to fees, fraud and low-quality inventory. ANA benchmark

The average cost per click in 2024 was $4.66 and rising in 86% of industries. At that price, a hundred junk clicks a week cost about $466. The fix is a short review of a few reports, done on a fixed schedule.

A wasted-spend review pays back more than most account work. It is rarely urgent on any given day, so it keeps getting pushed to next week.

Start with the search terms report

The search terms report is the most useful screen in Google Ads and Microsoft Ads. It lists the exact queries people typed before your ad showed. Those queries often differ from the keywords you bid on. With broad and phrase match, that difference is where budget goes to searches you never meant to target.

Open it, set the date range to the last 7 to 14 days, and sort by cost, highest first. Look for terms with high spend and few or no conversions. If a term cost $180 and brought zero conversions, add it as a negative keyword. Read each query and work out what the person wanted. Common problems are words like “free,” “jobs” and “how to,” competitor names, products you do not carry and cities you do not serve. Each of these needs a different fix.

Negative keyword match types

The match type of a negative decides how much traffic it blocks. Set it too broad and you block searches that would have converted. Set it too narrow and the same junk query gets through with slightly different wording. Google’s three negative match types also work differently from positive match types, and this confuses experienced advertisers too.

Shared lists and account-level negatives

Adding negatives one campaign at a time stops working once you manage more than a couple of clients. Google Ads has two tools for this.

Shared negative keyword listslet you keep one list and apply it to many campaigns. Start with a few standing lists: general junk terms like “free / cheap / DIY,” competitor brand names, and job and career terms. Attach each list to the campaigns it fits. When you update a list, every campaign that uses it gets the change right away.

Account-level negatives are for terms that should never trigger an ad in any campaign, such as adult content, brand-safety issues or categories you will never sell. They apply to search and shopping inventory, now including Performance Max, and are capped at 1,000 keywords. Use them only for terms that are always wrong. If a term is irrelevant only in some campaigns, add it at the campaign or ad group level.

Placements and audiences outside search

Negative keywords only control search. In Display, Performance Max and Meta campaigns, you also have to control placements (where the ad shows) and audiences (who sees it).

Display and YouTube placements

Display campaigns show your ads on millions of sites and apps. A noticeable share of that inventory is mobile-game interstitials and content farms, which bring accidental clicks. Open the placements report, sort by spend, and exclude apps and domains that cost money and do not convert. Placement exclusion lists let you apply one block list to many campaigns, the same way shared negative lists work in search.

Performance Max

For a long time PMax gave advertisers few controls and little query data. That has changed. Google added campaign-level negative keywords for Performance Max in December 2024 and raised the limit to 10,000 per campaign in March 2025. It then released a PMax search terms report, so you can see which queries you pay for. Dataslayer’s guide explains the setup. Add brand exclusions to keep shopping ads off competitor brand terms, and use account-level negatives for terms that should never trigger an ad. Google’s 2025 update also added brand controls for specific ad formats.

Meta Advantage+

Meta recommends automatic placements because a wider choice of placements helps the algorithm find cheaper impressions. You can still limit where your ads run. Exclude placements that you have tested and that did not convert. Apply publisher and content block lists, and opt out of topics like gambling or mature content for brand safety. One caveat: Meta may still send about 5% of spend to “excluded” placements to test them. An exclusion cuts spend on a placement but may not stop it completely, so keep checking the placement breakdown.

Separate brand from non-brand first

Before you judge results, split brand from non-brand traffic. Branded searches, where people type your company name, convert at a much higher rate and lower cost, because these people are already looking for you. If brand and non-brand share a campaign, the cheap branded conversions pull your averages up and hide waste in the non-brand terms.

Add your brand terms as negatives in non-brand campaigns so the two do not compete. A brand-term negative list keeps prospecting campaigns on new demand. When you report brand and non-brand separately, you can see whether your non-brand campaigns bring in new customers or mostly reach people who already know you. This also depends on clean conversion tracking. If the conversion data used for bidding is wrong, the algorithm optimizes toward the wrong queries no matter how many negatives you add.

The weekly wasted-spend audit

A one-time cleanup does not last. New junk queries keep appearing as broad and phrase match pick up new searches. Set aside 20 minutes a week and run the same five steps each time.

1. Pull the search terms report for the last 7 to 14 days, for both search and PMax, and sort by cost, highest first.
2. Flag terms with zero conversions. Any high-spend term with no conversions is a candidate. Read the query and write down why it failed: wrong intent, research only, competitor name, product you do not sell, or location you do not serve.
3. Choose the level and match type. Terms that are junk in every campaign go on a shared list as negative broad. A single bad query gets a negative exact on its campaign. Add the obvious variants, because negatives do not match close variants.
4. Check placements.Open the Display and PMax placement reports and Meta’s placement breakdown. Exclude placements that spend money without converting.
5. Log what you cut. Write down what you excluded and why. Next week you start from that list and do not have to review the same terms again.

The hard part is doing this every week on both Google and Meta. The reports sit in different tools with different exports, and in a busy week the review is the first thing teams skip. Adriva is built to run this kind of recurring check across platforms. It shows wasted-spend queries from Google and Meta in one workflow, so the weekly audit happens even when nobody remembers to open the report.

After you cut the waste, move the saved budget to campaigns that convert. Automated bidding strategies and cross-channel budget pacing handle that part. It is also why running Google and Meta together works better than optimizing each platform on its own.