Performance Max: should you trust Google's black box?

Your Google Ads rep has one favourite recommendation, and it’s always the same: switch to Performance Max. It’ll “unlock all of Google’s inventory.” It’ll “let AI do the heavy lifting.” What they leave out is that PMax is the one campaign type where you hand Google the steering wheel and the map, then find out where it took you after the money’s gone.

That doesn’t make it bad. It makes it a tool — and tools have a right and a wrong job. Here’s when Performance Max earns its keep, when it quietly bleeds you, and the exact settings that decide which one you get.

What PMax actually is

One campaign that serves across every Google surface at once. Per Google’s own documentation, a single Performance Max campaign runs on YouTube, Display, Search, Discover, Gmail, and Maps, with Google’s AI deciding the bids, the placements, and which combination of your assets to show where.

You don’t pick keywords. You don’t pick placements. You feed it creative assets, a conversion goal, and some audience signals, and it optimises toward that goal across all of Google. That’s the pitch. It’s genuinely powerful — and it’s also exactly why it goes wrong.

Where the black box wins

PMax is legitimately good in a few specific situations:

• E-commerce with a clean product feed. This is its home turf. If you’ve got a Merchant Center feed and real conversion data, PMax turns Shopping, YouTube, and Display into one machine that’s hard to beat manually.
• Accounts with lots of conversion volume. The AI needs data to learn. Thirty-plus conversions a month gives it something to optimise on. Three does not.
• Broad reach goals where you’d struggle to build channels by hand. If the honest goal is “sell more of this product to anyone likely to buy,” PMax’s job is to find those people everywhere at once.

Notice the pattern: clear commercial intent, a real feed, and enough data to learn from. When those three line up, the black box is a genuine advantage.

Where it quietly bleeds you

Now the part your rep skips. PMax fails in predictable ways:

• Thin conversion data. A local plumber getting five leads a month doesn’t feed the algorithm enough to optimise. It’ll spend confidently and learn nothing.
• It cannibalises your brand traffic. Left unchecked, PMax loves serving to people already searching your name — the cheapest, easiest “conversions” there are — and then reports them as wins. You’re paying to buy customers who were already yours.
• You can’t see where the money went. Placement and search-term reporting is deliberately thin. When results dip, you often can’t tell whether it was a bad YouTube placement or a junk search query, because Google doesn’t fully show you.
• It’ll chase cheap, worthless conversions. Point it at “form fills” without qualifying them and it’ll happily buy you a pile of tyre-kickers, hit the goal, and call it a success.

This is the through-line of everything we say about paid traffic: your ads aren’t expensive, your funnel is. PMax makes that worse, because it optimises toward whatever you told it to count — so if you’re counting the wrong thing, it’ll get very good at buying you more of the wrong thing.

The settings that decide your outcome

If you do run PMax, these are non-negotiable. Most of the horror stories come from skipping them:

• Exclude your own brand. Add account-level negative keywords for your brand name. Google supports a global negative keyword list that applies across eligible inventory — use it so PMax stops taking credit for people who already know you.
• Feed it a real conversion, not a proxy. A qualified lead or a sale — not “clicked the phone number.” Garbage goal in, garbage spend out.
• Give it good creative. PMax builds ads from your assets. Weak images and lazy headlines produce weak ads on premium placements. The landing page and creative still decide whether the click turns into anything.
• Watch it for the first month. The AI has a learning period. Don’t judge week one, but don’t leave it unsupervised for a quarter either.

So — should you use it?

Honest answer: if you’re e-commerce with a feed and conversion volume, probably yes, and it’ll likely beat what you’d build by hand. If you’re a local service business with a handful of leads a month, a tight Google Ads search campaign with real keywords and negatives will almost always outperform PMax — and you’ll actually be able to see what’s working.

The black box isn’t the enemy. Handing it the keys before you’ve earned the data to trust it is.

FAQ

Is Performance Max replacing regular Search campaigns? No. Google positions PMax as complementary — it respects your existing keyword-based Search campaigns rather than replacing them. For many small and local accounts, a well-run Search campaign is still the better core.

Why does PMax look so good in the reports? Because it often counts easy conversions — people already searching your brand — as wins. Exclude your brand and qualify your conversion goal, and the numbers get honest fast.

How much conversion data do I need before PMax makes sense? Rough rule: enough that the algorithm has something to learn from — think dozens of conversions a month, not a handful. Below that, you’re paying Google to guess.

The bottom line

Performance Max is a fast car with tinted windows. In the right hands, on the right road, it’s excellent. Pointed at the wrong goal with no guardrails, it’ll drive your budget somewhere you can’t see and hand you the bill.

Not sure which one you’re running? Tell us what you’re spending and what you’re getting — if a plain Search campaign would serve you better, we’ll tell you that instead of selling you the black box.