Performance Max Audience Signals Are Not Targeting

An audience signal is a starting hint, not a fence

An audience signal tells Performance Max who to try first. It does not limit who sees your ads. Google describes signals as optional input that helps its system learn faster. The campaign is free to show ads to anyone it predicts will convert, inside your signal or not.

In practice, you attach a signal to an asset group. It can include customer lists, website visitors, custom segments built from search terms or URLs, interests and demographics. Early on, PMax leans on that signal to decide where to spend. As conversion data builds up, it relies on the signal less and on its own predictions more. After a few weeks, your signal may barely matter.

This is by design. PMax is built to find conversions you would not have targeted yourself. That is the selling point. But it means the signal is not a control. Treat it like a briefing you give a new hire on their first day, not a rule they have to follow.

What follows from that:

• Removing a signal does not shut traffic off. It just takes away the starting hint.
• A narrow signal does not mean a narrow audience. A signal of 2,000 customer emails can still lead to ads shown to millions of people.
• A "wrong" signal does not lock you into the wrong audience. It slows the start, and then the system corrects course using your conversion data, for better or worse.

In Search, audiences can restrict. In PMax they cannot

The confusion comes from Search campaigns, where audiences have two settings. Targeting limits your ads to people in that audience. Observation shows ads to everyone your keywords match and just reports how the audience performs. Plenty of advertisers have used Targeting for years, for remarketing-only campaigns, for example, or for bidding on broad terms only for past visitors.

PMax has no Targeting setting for audiences. Every audience signal behaves, at best, like a softer version of Observation. It shapes where the system looks, and you can see a report on it, but no switch turns it into a limit.

There is a second difference people miss. In Search, the keyword is the main fence. An audience set to Targeting is a second fence on top of it. Your ad needs a keyword match and a person in the audience. In PMax, there is no keyword fence either. Search themes are hints in the same way signals are. We cover that in Search Themes vs Negative Keywords in Performance Max. So a PMax campaign can have two layers of suggestions and no hard limit on who it reaches or what they searched.

Here is the side-by-side:

If you move a working Search setup into PMax and keep the same mental model, you lose both fences without noticing.

Where the misunderstanding costs money

It costs money when you plan budget, bids or reporting as if the signal held. The ads go wider than you expect, and the wider traffic is often the least qualified. Three patterns come up again and again in the accounts we audit.

The B2B campaign that "only targets decision makers"

A software company runs PMax with a signal built from its customer list plus a custom segment of people who searched "accounting software for small business" and "best invoicing software". The account manager reports to the client that the campaign targets finance decision makers.

Then the search terms report shows spend on "free invoice template", "invoice template word", "what is accounts receivable" and "bookkeeping jobs near me". None of those people are in the customer list. The signal did nothing to stop them. PMax matched them because they look loosely related, and because they are cheap enough to try.

Say the campaign spends $3,000 a month and those four kinds of terms take $450 of it. That is 15% of the budget going to traffic the client was told the campaign did not reach. The problem is not only the $450. The client trusts a description of the campaign that isn't true. For the template searches specifically, Blocking "Free" and "Cheap" Searches Without Losing Buyers covers how to cut them without blocking buyers who type "free trial".

The local business with a "homeowners" signal

A plumbing company adds a signal for homeowners and people in the market for home services. It feels safe: tenants and job seekers are surely out.

They aren't. The search terms include "plumber salary", "plumbing apprenticeship" and "how to fix a leaking tap". Those people may well own homes. Some of them are even in the in-market segment. But they are not hiring a plumber today. An audience signal can't separate "owns a home" from "searching for a job". Only the search term shows that difference, and only a negative keyword acts on it.

The e-commerce store with a "past purchasers" signal

A store selling running shoes uses its buyer list as the only signal and assumes PMax is mostly reaching existing customers. Then it reads new customer revenue as proof that the campaign works for retention.

In reality, PMax reached plenty of people who never bought, including searches like "running shoes repair" and "how to clean running shoes". Some converted, which is fine. But the report the store built on top of the signal measured the wrong thing. If you are reading PMax results by audience, remember the signal tells you where the campaign started, not who it ended up paying for.

What actually restricts Performance Max

These are the settings that limit where PMax spends: location and language, conversion goals, listing groups and URL settings, brand exclusions, placement exclusions, and campaign-level negative keywords. Everything else is a hint.

Location and language. These are real limits. Check the location option too. "Presence or interest" lets in people who are only interested in your area, which for a local business often means people far away. "Presence" is stricter.

Conversion goals. PMax optimises for whatever you count as a conversion. If a page view or a "contact page visit" counts, the system will chase cheap traffic that triggers it. This is not a fence in the usual sense, but it is the strongest lever you have over what kind of traffic PMax looks for. A tight signal with a loose conversion goal ends up as a loose campaign.

Listing groups and feed. For retail campaigns, the products you include decide what can show. Excluding a product removes it completely. Including everything and hoping the signal steers spend does not work. How Listing Groups Hide Wasted Spend in Performance Max goes into the detail.

Final URL expansion. If it's on, PMax can send traffic to pages you didn't pick, like the careers page or a blog post. That pulls in searches that match those pages. Turning it off, or excluding URLs, is a real restriction.

Brand exclusions. These stop PMax from serving on searches for brands you list, including your own. They have gaps, which we cover in Performance Max Brand Exclusions and What They Miss.

Placement exclusions. These keep ads off specific sites, apps and channels. Like brand exclusions, they work but have limits; see Performance Max Placement Exclusions, Honestly.

Negative keywords. A PMax campaign accepts up to 10,000 campaign-level negative keywords. It does not accept shared negative keyword lists, so a list you maintain for your Search campaigns does not carry over automatically. You have to add the terms to each PMax campaign. Negatives are the only control that acts on what someone actually typed, which is exactly the gap the signal leaves open.

Google has been adding more exclusion controls to PMax over time. Check what your account currently shows instead of relying on an old guide. But none of those additions changes how signals work. A signal is still a hint.

Negative keywords close the gap signals leave

Signals describe people. Negative keywords describe searches. Wasted spend in PMax mostly comes from searches: the right sort of person typing the wrong thing at the wrong time. A homeowner researching plumber salaries. A finance manager downloading a free template. A runner looking up how to clean shoes.

That's why negatives matter more in PMax than many people think. A few practical points:

Match type matters. Broad, phrase and exact are the three negative match types, and they don't behave like the positive versions. Broad negatives in particular behave unpredictably across match types. For PMax, phrase negatives on clear intent words ("salary", "jobs", "template", "repair") tend to be the clearest choice, with exact negatives for specific queries you want gone. Negative Keyword Match Types Don't Work Like Positives explains the differences.

The 10,000 limit is real but rarely the issue. Most PMax campaigns we audit carry somewhere between a few dozen and a few hundred negatives, nowhere near the cap. The more common problem is having none, because the advertiser thought the signal handled it.

Not every search is visible. Search terms for Search, Shopping and Performance Max are reported in campaign_search_term_view. That covers the search side of PMax. The Display, YouTube and Discover parts of PMax report no search terms, because no one searched. Negative keywords do nothing there. That traffic is controlled by placement exclusions, your conversion goals and your assets, not by negatives. What the PMax Search Terms Report Actually Shows lays out what you can and can't see.

Don't over-block. The danger with a campaign that has been leaking is overreacting. "How to fix a leaking tap" is waste for an emergency plumber. "Leaking tap repair near me" is a customer. Block the first and keep the second. If you block too much, you'll see impressions drop and cost per conversion go up. Any negative should be easy to reverse, as described in How to Undo Negative Keywords in Google Ads.

How to use audience signals well anyway

Signals are still worth setting up. They help a new campaign spend its first weeks on likely buyers instead of guessing. Just give them the job they can do and don't count on them for the jobs they can't.

Put your best first-party data in. Customer lists and converters from your website are the strongest starting hints. Google has more to learn from 5,000 real buyers than from a broad interest category.

Build custom segments from search terms that converted. If "emergency plumber" and "burst pipe repair" brought you calls in Search, use them in a custom segment. This points PMax at people who show that intent. It's a hint, but a good one.

Match the signal to the asset group. A signal for "running shoe buyers" attached to an asset group with trail running creative and trail product images makes sense. A broad signal across a mixed asset group gives the system little to work with. When one asset group starts pulling unrelated traffic, the fix is rarely the signal; see When One PMax Asset Group Pulls Irrelevant Traffic.

Don't describe the signal to clients as targeting. For agencies and freelancers, this is the part that matters most. Write "we've told Google to start with people similar to your customers", not "the campaign targets your customers". The first sentence is true and will still be true after the client reads the search terms report.

Don't judge the campaign by the signal. The audience insights in PMax show which segments converted. That's useful for creative and messaging. It is not a list of everyone who saw your ads, and it does not show what you paid for searches that didn't convert.

How to check what PMax actually reached

Look at what you paid for, not at what you suggested. Thirty minutes once a month in the search terms and insights reports will tell you how far PMax has moved from your signal.

1. Open the PMax search term data for the last 30 days. Sort by cost. Read the top 50 terms and mark any that your signal was supposed to "exclude": job searches, how-to questions, free tools, wrong product types, wrong locations.
2. Add up the cost of those terms. That's how much the signal didn't stop. Keep the sum simple. If 12 terms cost $40 each, that's $480. For a more careful method, see How to Calculate Wasted Ad Spend Honestly.
3. Check conversions on those terms before acting. A term that looks off but converted is not waste. Leave it.
4. Look at the location report. If you see clicks from places outside your service area, check the presence setting.
5. Look at the landing pages. If PMax sent traffic to pages you didn't choose, look at URL expansion.
6. Add negatives for the clear cases. Phrase match for intent words, exact match for specific queries.

Running this check is often what convinces people that signals aren't a limit. It's hard to keep believing a campaign "only targets decision makers" once you've seen it pay for "bookkeeping jobs near me". For more on why those terms show up at all, Why Google Ads Shows Irrelevant Search Terms explains how matching works.

Where PPCleaner fits, and where it doesn't

PPCleaner handles the search side of the gap. Every night, it reads the search terms your Search, Shopping and Performance Max campaigns paid for. It judges which ones were never going to convert and adds them as negative keywords. In PMax, those are campaign-level negatives, since PMax doesn't take shared lists. It never blocks a term that converted in the last 90 days. Every block can be undone in one click for 90 days.

What it does not do:

• It doesn't touch your audience signals. It won't rewrite them, tighten them or remove them. It doesn't need to, because signals were never the control.
• It can't do anything about non-search placements. Display, YouTube and Discover inventory inside PMax reports no search terms, so there's nothing to judge and no negative to add. That part is up to your placement exclusions, conversion goals and assets.
• It doesn't change budgets, bids, ads or keywords. If your conversion goal counts page views, PPCleaner won't fix it. It will just keep blocking the search terms that goal draws in.

The honest summary: a signal tells PMax where to start. Negatives, exclusions and settings tell it where to stop. Most of the wasted spend we find in PMax comes from accounts that set up the first and skipped the second.