For years the pitch on automated campaigns asked you to trust the algorithm and look away. Hand over the budget, feed it a goal, and resist the urge to touch anything. In 2026 that pitch quietly changed. Google shipped a wave of Performance Max controls and reporting, from account-level negative keywords to first-party audience exclusions and network-segmented placement data. Microsoft built its entire Activate 2026 message around a single word: control. The platforms are no longer selling automation as a black box. They are selling it as something you can steer.
That shift matters more than any one feature, because it settles an argument a lot of advertisers have had backwards. Control never actually left PPC. It moved.
It used to live in the manual levers, the bids and exact-match keywords and placements you set by hand. Those levers are mostly gone or fading, and the marketers who feel out of control are the ones still reaching for them. Control now lives upstream, in the inputs you feed the machine before it makes a single decision: your data, your goals, your guardrails, your account structure, and the way you measure what comes back. The advertiser who masters those inputs is not fighting automation and not blindly trusting it. They are steering it.
Here is where that control actually sits now, across five layers, and what changed in 2026 that gives you more of it than you had a year ago.
1. Signal quality: the real steering wheel
If you only fix one thing this quarter, fix your data before you touch a single setting.
Automated bidding is only as good as the signal you feed it. Thin or noisy conversion data means you have handed the algorithm your budget and asked it to guess. Strong signal means you have pointed it at exactly the outcomes that make you money. The gap between those two accounts is enormous, and it has nothing to do with how clever your campaign settings are.
Three things separate a well-fed account from a starving one. First, you are importing conversion values, not just conversion counts, so the system optimizes toward profit rather than raw volume. A lead worth ten thousand dirhams and a lead worth two hundred should not look identical to the machine. Second, you are using enhanced conversions and offline conversion import to close the loop between the click and the sale that actually happens days or weeks later, often on the phone or in person. Third, you are leaning on first-party data as your durable advantage, because third-party signal keeps degrading and the accounts that own their customer data will keep pulling ahead of the ones that rent it.
This is the least glamorous layer and the highest leverage one. Most posts on automation skip straight to settings. Start here instead.
2. The goals you set
Automation drives. You still choose the destination.
Your bid strategy target is the most direct piece of control you have left, and it deserves to be treated as an active decision rather than a field you filled in once. A target ROAS or target CPA tells the system what a good outcome looks like, and moving it a few points changes how aggressively the machine chases volume versus efficiency. Value rules go a step further, letting you tell the platform that certain conversions, certain locations, or certain devices are worth more to your business than the defaults assume.
Then there is the acquisition question, which quietly eats budgets. Left alone, automated campaigns will happily spend to re-buy customers who were going to come back anyway. Both major platforms now let you steer against this. Google offers a high-value new customer mode, and Microsoft has a new customer acquisition goal, both of which push the system toward genuine growth instead of expensive loyalty. If you are paying to acquire, make sure you are actually acquiring.
3. Guardrails: where 2026 changed the most
This is the layer that gained real ground this year, and it is worth getting the details exactly right.
Performance Max, long criticized as a place where budget disappeared with little explanation, now accepts a proper set of fences. Advertisers can apply negative keywords at the campaign and account level, exclude their own brand terms, and exclude existing customers and first-party audiences so acquisition budgets are not wasted on people who already converted. Microsoft has matched much of this with self-serve negatives, brand inclusions and exclusions, and term exclusions. Taken together, these are the tools that keep automation inside your brand safety and efficiency boundaries instead of wandering wherever the model finds cheap clicks.
One caution, because precision is the whole point here. Not every control is fully released. The ability to opt out of Search Partners and the Display Network from inside Performance Max surfaced as a Partners alpha around July 2026 and was reported as limited rather than generally available. Treat anything still rolling out with a dated note, and check availability in your own accounts before you promise a client a lever that may not be switched on yet. As of late 2026, some of these are live for everyone and some are not.
That habit of separating what is released from what is in testing is not just good practice. It is exactly the kind of accuracy that earns a piece of content citations and trust, which matters more every year as buyers and AI answer engines alike reward sources that get the specifics right.
4. Account structure is a control lever
The lever people forget is a lever.
When you pour everything into one undifferentiated campaign, you force the algorithm to optimize toward a blended average that flatters no part of your business. It will quietly rob your high-margin lines to feed easy, low-value conversions, because on a blended view those conversions look like wins. Structure is how you stop that.
Splitting campaigns and asset groups by margin, by purchase intent, by geography, and by language forces the system to optimize toward your business logic rather than its own convenience. A high-margin product line gets its own target and its own budget protection. A prospecting effort is kept separate from a retention effort so their very different economics do not get averaged into mush. None of this fights automation. It gives automation a cleaner problem to solve.
5. Measurement and verification
Do not only grade the homework the platform grades itself on.
Every automated system tends to report the numbers that make it look good. That is not a conspiracy, it is just incentives, and it is why the final layer of control is refusing to accept the platform’s story at face value. Use the network-segmented placement reporting and channel-level insights the platforms now expose to see where your money actually went, not just what the headline performance says. Use budget and pacing forecasts to catch overspend before the month closes rather than after. And where the stakes justify it, verify lift with independent or incrementality-based measurement, so you know the automation is creating results rather than taking credit for demand that already existed.
Control without verification is just trust with extra steps.
The Gulf lens
There is a regional dimension to all of this that global playbooks miss.
Automated systems lean on patterns learned from enormous pools of data, and those patterns are thinnest exactly where your market is most distinct. In the Gulf, that means the quality of your own first-party signal matters even more, because the model has fewer reliable priors to fall back on. It means Arabic and English deserve their own asset groups rather than being blended, so creative, query matching, and reporting stay clean in both languages. And it means regional placement control is not a nice-to-have, because brand fit and cultural context are harder for a globally trained system to judge on your behalf.
The advertisers who treat the Gulf as a rounding error in a global account tend to get generic automation. The ones who feed the system strong local signal and hold it to local standards get automation that actually works here.
The new rules, in one line
Control did not disappear when the platforms automated. It moved upstream, into the data you feed the machine, the goals you set, the guardrails you place, the structure you build, and the way you verify what comes back. That is where the work is now. Fight the automation and you lose. Blindly trust it and you also lose. Steer it deliberately, from the layers that still answer to you, and you get the scale of the machine with the judgment of an operator who knows the business.
That is the whole discipline, and it is a good time to get fluent in it. The platforms have just handed you more of the wheel than they have in years.
