Performance Max 2026: 7 tips to tame the beast

In 2023–2024, Performance Max was rather like handing Google your credit card with your eyes closed. You supplied a budget and some assets, then the algorithm did whatever it wanted, often badly. In 2026, the black box finally has controls. Not all of them, but enough to make it useful if you know what you're doing.

I've read around twenty articles and studies published in recent weeks (Search Engine Land, Optmyzr, Yellow Jack Media, the official Google Ads blog…) and compared them with what I see in client accounts. Here are seven tips that make a real difference in April 2026.

1. Segment campaigns, but do it carefully

When you first use PMax, it's tempting to put everything into one campaign to "give the algorithm as much data as possible". Google even recommended it. But mixing lead generation, product sales and subscriptions sends conflicting signals, so the algorithm optimises for the lowest common denominator.

The best practice in 2026: one PMax campaign = one business objective. But there's a catch. Over-segment, with fewer than 30 conversions a month per campaign, and you spread the data too thin, making things worse. A simple rule: only segment if each campaign will receive 30+ monthly conversions. Otherwise, combine them.

2. Conversion signals: less is more

This is THE mistake I see most often when taking over an account. Someone tracks product page views, add-to-carts, contact forms, phone calls and completed purchases, all as primary conversions in one basket. The result: the algorithm spends your budget generating page views. Well done.

The golden rule in 2026: only use a primary conversion for something with real business value (a purchase, qualified lead or signed quote). Everything else becomes secondary : tracked and observed, but not used to guide bidding. You optimise for the signals you send, so send the right ones.

3. Negative keywords finally get taken seriously

Big news from early 2025 that many people missed: Google raised the PMax negative keyword limit from 100 to 10,000 per campaign, managed directly in the interface. No more begging an account manager to add three exclusions.

What I routinely add to client accounts: competitor brand names (unless there's an intentional conquest strategy), informational queries ("how", "definition", "what is"), their own brand if a dedicated Search campaign already covers it (otherwise PMax eats into branded traffic and inflates its numbers), and locations outside their market.

Remember, PMax negative keywords filter only Search and Shopping inventory. They don't cover YouTube, Display or Discover.

4. Exclude existing customers (new in April 2026)

The most useful update so far this year: you can now exclude entire customer lists from PMax campaigns. Recent buyers, newsletter subscribers, free-trial users who didn't convert… remove them from targeting in three clicks.

That means your acquisition budget stops retargeting people who already know you. For an online retailer with 50,000 customers, the savings are immediate and visible within fifteen days. I enabled it last week on a client account; CPA fell 18% without changing anything else. Magic.

5. Themed asset groups: the algorithm needs clarity

A catch-all asset group with 20 generic headlines and ten random images is a sure route to a Poor or Average rating. That rating directly and measurably affects bidding: a Poor group is at a disadvantage against an Excellent one, even at the same CPC.

The right approach: one asset group = one clear theme. If you sell boilers and air conditioners, create two asset groups. Product-specific headlines, product images, benefit-focused descriptions and targeted audience signals. Google wants Excellent; according to its own data, that can mean up to 6% more conversions than Good. It may not sound huge, but on an account spending €10,000 a month, it adds up over a year.

6. Video really isn't optional anymore

I still hear advertisers say "I don't have the budget for video". In 2026, that's almost self-defeating. Google's internal data shows 25–40% better performance for campaigns with a complete video library compared with image-only campaigns. If you don't provide a video, Google generates one from your assets, and frankly, the result is painful to watch.

The good news is that in 2026, AI tools such as Runway, Google Vids and Synthesia can produce decent videos in a few hours without a studio. Keep them short (15–30 seconds), with one message per video and your logo visible in the first three seconds. No need to be Scorsese.

7. PMax and Standard Shopping: a combination that works

The final tip is particularly useful for e-commerce: don't leave PMax on its own. The best e-commerce advertisers I follow in 2026 run a Standard Shopping campaign alongside it, bidding more aggressively on flagship products. PMax handles volume; Standard Shopping protects long-tail products and top sellers.

The logic is simple: PMax optimises for the average, not your bestsellers. A Standard Shopping campaign alongside it keeps you in control of what really matters to your profit and loss.

My view after 15 years in digital

Performance Max took three years to become a tool humans could use. In 2023 it was a gamble. In 2026 it's powerful, provided you understand that you don't manage it like an old Search campaign. You don't fight automation; you guide it with better signals.

If you remember one thing: the quality of your inputs — conversions, assets, audience signals and negative keywords — matters ten times more than any bidding setting. Create good assets, clean up your tracking, exclude existing customers and watch CPA fall.

If you'd like someone to look at your current campaigns, I audit Google Ads accounts for free as part of my Google Ads service. If you're completely new to Performance Max, I wrote an introductory article on PMax updates a few months ago that still provides a useful foundation.

Translated from the original French article. Publication dates, examples and figures refer to that original version.