• by Cory Treffiletti , Featured Contributor, September 30, 2026

There’s been a lot of ad platform changes over the last few weeks.  Google started moving legacy search campaigns into its AI Max product on Sept.  1, and the migration just happens, whether you were planning for it or not.  Amazon did something similar back in August, quietly enrolling Sponsored Products campaigns into creator reviews and buying guides, with the setting turned on by default. 

These changes were made automatically, with no way to opt out.  The platforms didn’t ask, “Are you sure?”  They just turned them on, assuming advertisers would be happy, and that performance would prove the wisdom of the change.

I’m not anti-automation.  If we think it will be good for the buyer, I support it. Automation saves time, and it can unlock many new paths that a human being might have overlooked.

What concerns me lines up with a previous article I’ve written. Product teams are rolling out so many new features and changes because the rate of development has accelerated so quickly.  In many cases, we are dismissing (or forgetting) to make the end user aware of these changes.  That opens inventory up to fully automated optimization, or to creator content, which may not always align with the advertiser’s strategy. 

If you check with them, that’s ok.  If you just do it without their knowledge, that feels more problematic.

Let’s look specifically at Google for a moment.  AI is very, very good at following a trend.  It sees what performed yesterday and leans into it, and it does so faster and more efficiently than any human team ever could.  But it’s not good at noticing when the trend is the wrong thing to follow. 

AI doesn’t know your product has a supply issue in the Southeast next month, or that the creator driving all those cheap clicks said something last week your CEO would not love seeing your brand support – unless you tell it.  It doesn’t know that the audience converting most efficiently today might not be the audience you’re trying to build for the next five years (once again, unless you feed it that information). 

Humans know these things, sometimes because of data, and sometimes simply because they were in the meeting, or they read the room, or they have a gut feeling they can’t fully explain yet.

Some of the best calls marketers make go against what the numbers say at the time.  Optimization rewards the average, but brands are often built on the exception.  Sometimes you pull back on a channel that’s performing because you can see where it’s heading, or you stay in one that’s underperforming because you believe in the long game, and those decisions look wrong on a dashboard right up until the moment they look brilliant.  (I made a version of this argument about the movie “Air” earlier this year, and I still don’t think an algorithm would have signed Michael Jordan.)

I think the fix is a lot simpler than the platforms are making it.  Give the marketer the switch option and execute good product marketing.  Explain what the feature does, show the tradeoff, and let them decide before the money moves, or give them the option to flip back.  I suspect most people would agree to the change, because most of these features work well enough and the efficiency gains are real.  The ones who say no usually have a reason, and that reason is probably the exact thing the algorithm can’t see.

The platforms are also facing a trust issue here. Marketers are being asked to hand more and more of their budgets to systems they can’t fully see inside, and you earn that kind of trust by respecting the person holding the budget, not by moving it for them.  Every time someone finds a change in their account they didn’t make, a little bit of that trust erodes, even if the change happens to perform.  Choice builds confidence.  Surprise builds suspicion.

I do think we’re heading toward a place where AI makes most of these calls, and I’m mostly OK with that.  I just want the human to decide when to let go of the wheel, rather than finding out six weeks later that somebody already took it. 

The machine can see the trend.  The marketer can see what the trend is missing.  Both of these matter, and right now only one of them is being asked for its opinion.