Click fraud on Performance Max: what Google Ads teams miss

Why PMax makes junk harder to see

Performance Max optimizes across Search, Shopping, Display, YouTube, Discover, and Gmail with limited channel breakouts. When invalid clicks land, the damage shows up as softer ROAS and odd asset-group learning — not a neat bot label in the UI.

Teams often respond by swapping creatives or raising budgets. That can help when demand is real. It fails when the mix underneath includes farms, scanners, and competitor drain teaching the model the wrong lesson.

Treat PMax like any other paid program: score clicks as they arrive, then queue those IPs for Google Ads exclusion. The format’s opacity is a reason to add protection earlier, not later.

Finance partners feel the lag first. Invoices still arrive on schedule while the model quietly learns from sessions that never had purchase intent. A weekly junk-share view next to ROAS keeps that conversation grounded.

Asset groups that look “learning” for weeks sometimes are learning from the wrong visitors. Protection is how you give the model cleaner raw material without pausing the whole campaign.

If your agency reports only campaign-level ROAS, ask for junk share next to it. Opaque formats reward teams that bring their own session layer.

Signals that matter on asset groups

Watch high-risk sessions with near-zero engagement on shopping and lead landing pages. Bursts from the same device cluster or ASN are another tell, especially overnight on brand and category queries.

Compare risk share before and after you expand to new countries or feed items. Fresh inventory attracts scrapers as quickly as it attracts shoppers — and PMax will happily spend into both until you intervene.

Asset-group level reporting can look healthy while one SKU cluster absorbs most of the junk. Saved views that slice by product type or landing path expose that imbalance faster than campaign totals.

Brand terms clicked with impossible travel patterns — same fingerprint hopping geos in minutes — deserve an immediate review even when the creative looks strong.

Screenshot or export the first clear example you find. Pattern libraries beat tribal knowledge when a new buyer joins mid-quarter.

Cross-check with search-term or product reports when available. PMax will not always hand you the channel, but landing paths still reveal intent.

• High risk + low engagement on product URLs
• Repeat fingerprints across asset groups
• Sudden ASN spikes after feed or geo expansion
• Brand terms clicked with impossible travel patterns
• Overnight bursts on newly listed SKUs

Block without starving real demand

Start with conservative defaults for a week. Promote clear junk to high-risk queues, keep gray cases on a watch view, then tighten once false positives look acceptable for your catalog.

Document every rule change next to CPA movement. PMax attribution is noisy; a short paper trail keeps finance and creative teams aligned when results shift.

Avoid blanket country exclusions that also remove legitimate shoppers in the same market. Prefer risk and engagement filters first, then geo only when the pattern is overwhelming and sustained.

If a creative refresh coincides with a junk spike, pause the blame game. Check traffic quality before you kill assets that may have been fine under cleaner traffic.

Run a soft pilot on one asset group before account-wide aggression. Prove the false-positive rate on a contained slice of spend.

When in doubt, watch for seventy-two hours. PMax noise is high; patience prevents swinging the ax at real demand.

A weekly PMax protection rhythm

Every Monday: open high-risk paid sessions, export the top junk sources, and confirm your Google Ads IP exclusions still match the ClicksArmor export.

Midweek: skim the watchlist. Promote anything that repeated with zero engagement; release anything that converted or looked like a real shopper on a slow connection.

Every month: compare junk share and ROAS on a like-for-like window. If junk fell and ROAS rose, protection is working even when top-line volume looks flat.

Quarterly: revisit default thresholds after major catalog or geo expansions. Yesterday’s “aggressive” rule can be too loose once scrapers discover a new price drop.

Keep the Monday export under two pages. If the ritual takes an hour, people will skip it — and junk will quietly return.

Coordinate with feed and creative owners

PMax mixes shopping and demand surfaces. When junk spikes after a creative refresh, check traffic quality before you kill the assets.

A short note — junk share up, exclusions tightened, creatives unchanged — prevents the wrong post-mortem in the next standup.

Tell merchandising when you exclude an ASN that was hammering newly listed SKUs. Feed hygiene and media protection reinforce each other when both teams see the same spike.

Keep a shared calendar of feed pushes, promo windows, and protection rule changes. Spikes become explainable instead of mysterious CPA swings.

Invite feed ops to one monthly review. Shared language about scrapers versus soft creative prevents the wrong tickets from being filed.

What success looks like after thirty days

Junk share among paid PMax sessions trends down. Remaining sessions show clearer engagement on PDPs and lead forms. Smart Bidding has fewer polluted conversion paths to learn from.

You have a dated log of exclusions, a weekly export habit, and at least one stakeholder who can explain why CPC or ROAS moved without guessing.

If those three are missing, extend the pilot rather than declaring PMax “fixed.” Format opacity rewards patience and evidence more than one dramatic blocklist dump.

Celebrate boring Mondays. When the high-risk band is small and the note writes itself, the operating system is working.

Give Smart Bidding cleaner raw material

Every week of junk learning costs more than a week of cautious exclusions. Prefer early watchlists over late creative funerals.

Pair monthly ROAS reviews with junk-share by asset group. Opaque formats reward teams that bring their own session layer.

If leadership only wants one PMax metric besides ROAS, make it junk share. Volume without quality is an expensive vanity number.

Keep the change log short enough that a buyer will actually read it on Monday morning.

Revisit the rule after one full billing cycle — not after the first scary day.

When in doubt, watch for seventy-two hours before promoting a gray case to Google Ads exclusion.

Treat every major feed or geo expansion as a risk event: open the watch view the same day, not in the monthly ROAS review.

Document creative freezes during junk spikes so asset owners are not blamed for traffic they never chose.

A quiet week of exclusions often recovers more ROAS than another creative refresh on polluted learning.