You didn't lose control of your Google Ads account.

You traded it.

Performance Max arrived with a simple promise: give Google your goals, budget, assets and data, and it will find conversions across Search, Shopping, YouTube, Display, Discover, Gmail and Maps. In exchange, you gave up most of the levers you used to pull.

This isn't an anti-automation rant. I spent years at Google helping advertisers get better results from AdWords, and I run PMax for clients today. It works. Often very well.

But "it works" and "it works for you" are not the same sentence.

PMax optimizes toward what you tell it to value.

Not what you meant.

What you told it.

Machine learning is a genie. Phrase the wish badly and you may get exactly what you asked for.

Tell it to maximize conversions and it will look for the conversions it can generate most efficiently. Sometimes that's exactly what you want. Sometimes it's an existing customer searching your brand name who was going to buy anyway.

Tell it you want leads and it will find people willing to fill out your form. That doesn't mean they're qualified. They may be students, competitors, bots or people who were never going to buy.

The algorithm isn't lying to you.

It's obeying you.

Google controls much of the targeting, placements and bidding. Fine. Here's what you still own, roughly in order of impact.

1. The conversion signal. This is everything. Count the actions that represent actual business value. If you sell to businesses, import offline conversions from your CRM so PMax learns which leads became qualified opportunities and customers, not merely who filled out a form. If you can pass values, pass them. Profit beats revenue. Revenue beats a flat conversion count. And a qualified sale beats a lead form every time.

2. New versus returning customers. Tell Google who your existing customers are, and use customer lists and new customer acquisition settings deliberately. Otherwise some of the performance you're celebrating may simply be repeat business with a media fee attached.

3. Brand exclusions. Decide intentionally whether PMax gets access to branded demand. Don't let it happen by accident. If a Search campaign already captures your brand efficiently, test excluding it from PMax and watch what happens to the numbers. The difference tells you a lot about what PMax was actually doing.

4. Inputs. As Google automates more of the mechanics, what you feed the system matters more, not less.

Your feed is targeting.

Your creative is targeting.

Your landing page is targeting.

Your audience signals are information, not instructions. A sharp product feed, strong creative and a landing page that converts can matter more than all the bid adjustments we used to spend hours tweaking.

5. Structure. Separate things that genuinely behave differently: different margins, customer types, geographies or business objectives. One campaign containing everything eventually gives the machine one blended version of everything. Sometimes that's useful. Sometimes it hides the part of the business that's actually making money.

There's a short list of things I don't want the machine deciding for me.

What counts as a conversion.

What a customer is worth.

Whether an existing customer counts as an acquisition win.

How much budget each business line deserves.

Those aren't media buying decisions. They're business decisions disguised as Google Ads settings. If you don't make them yourself, the system will optimize toward whatever best satisfies the objective you gave it, and that isn't always the same thing as growing the business.

Google has slowly opened the box, with better channel reporting, more search term insights, more exclusion controls and more visibility into what PMax is doing. Use all of it.

But the real test happens outside Google Ads.

The question isn't whether PMax reported more conversions.

It's whether the business actually grew.

Run an incrementality test. Use a matched geography, a Google experiment where practical or another credible holdout, then watch total sales, qualified pipeline or profit, not just reported conversions. If the business barely moves when PMax disappears, the campaign was likely capturing demand that already existed. If it drops meaningfully, you've found incremental value.

Now you have something worth scaling.

Most advertisers never run that test.

Because they're afraid of the answer.

Automation doesn't remove the need for judgement.

It moves it upstream.

You no longer win by spending all day choosing keywords and tweaking bids. You win by deciding what the machine should value, giving it clean signals, feeding it good inputs, structuring the problem correctly and checking its answer against reality.

The black box is fine.

Just make sure you're the one writing the instructions.