The Ad-Buying Tools of All Three Platforms Are Quietly Turning Into an "AI Autopilot"

A while ago, a friend who runs ad campaigns vented to me.

He said that back when he trained new hires, he had to teach them everything by hand. How to define an audience, how to pick interests, how to adjust bids — he practically had to walk them through every single switch. But lately he'd noticed something: the new hire who didn't know the first thing was getting results no worse than his. Sometimes even a little better.

I asked him why.

He sighed. "The tools figure it all out for you now. A lot of what I taught them — it's not needed anymore."

My stomach dropped a little. That's a signal.

Three Platforms, One Move

Think about it — a company of any real size, which of these can it afford to ignore?

Meta's Instagram, Facebook, WhatsApp. Google, plus YouTube. And TikTok.

These three practically are the entire map of digital advertising.

And right now, all of them, in lockstep, are doing the same thing: handing the core decisions of ad buying over to AI.

On Meta's side it's called Advantage+ (Meta's AI-driven ad-buying suite), on Google's side it's Performance Max (PMax for short — Google's automated campaign type), and on TikTok's side it's Shopping Ads. Different names, same job. You give it a goal and a budget, and it decides for you where to place the ad, in what format, and how much to bid.

What do I mean by "where to place the ad"?

It means that for any given ad, the system itself judges: today should this show up in an Instagram Story, or pop into the feed; should it surface on YouTube, or get slotted into Google search results, or even Gmail. Inventory, budget, conversion goals — it works all of that out for you.

This sort of thing had a prototype a few years ago. But what really made the platforms throw their full weight behind it, go all in, was around 2024.

A $38.3 Billion Bet

Why go all in? Let me show you a number.

$38.3 billion.

That was Meta's advertising revenue in the second quarter of 2024. Up 22% year over year.

Mark Zuckerberg, at that earnings call, put it very bluntly. He said that in the past, advertisers came to them carrying "the audience I want." Age, region, interests — filtering through all of it. And now?

"All advertisers really have left to do is make the creative."

Then he added something even sharper: "In the coming years, AI will make the creative itself, and distribute it to each person in a highly personalized way."

This is the vision Zuck is painting for AI advertising. After the metaverse face-planted, this vision is, for Meta, practically its lifeblood. He even gave an end-state: an advertiser needs only to hand over a business goal and a budget, and the platform takes care of everything else.

"We're moving toward that step by step. This is going to be a big thing."

But Is AI Really That Magical?

This is where it gets interesting.

The platforms hype their tools relentlessly. But once you actually use them, you notice something counterintuitive: they keep saying "AI is all-powerful," while at the same time constantly adding "switches a human can control."

Take this example.

Meta added a new feature to Advantage+: advertisers can specify "which kind of customer I'm willing to bid more for," without having to set up a separate campaign. Sounds like a small update. But think about it — if the AI really understood everything on its own, why would they hand you a feature like this?

It amounts to the platform admitting it itself: AI isn't that smart yet. Some judgments still need a human watching.

Google is similar. PMax started out mostly running Shopping ads; now it's also opened up buying "by leads." Every time a new scenario gets unlocked, in essence it's patching a gap in the AI.

Data, Data, and More Data

So where exactly is AI stuck?

Stuck on not having data.

The biggest headache for ad buyers these last couple of years is that the numbers don't reconcile. You see one set of conversion figures in Meta's Ads Manager, then go look in Google Analytics and they can be off by a mile. Which one is real? Each side is convinced its own numbers are right.

Why?

Apple. After iOS 14.5, tracking got cut off at a stroke. If the user doesn't give consent, you can't follow them. Add the whole GDPR framework in Europe on top of that, and the data advertisers can get their hands on keeps getting thinner.

AI grows up eating data. Take the data away, and its judgments start to drift.

So when Meta updated Advantage+ this time, it took the chance to settle a long-standing grievance: letting advertisers bring in Google Analytics data, with Adobe Analytics to be opened up next.

Think about that. Before, Meta and Google each kept their own books. Now Meta is proactively saying, "let me see your ledger too."

Fred Leach, a Meta VP of product, put it candidly to Ad Age: "The system has already formed some of its own judgments about the differences between Meta and other channels. We'll keep updating to give advertisers more value."

In plain English: feed me the other guys' data too, and I can calculate more accurately.

What Does This Actually Mean?

After that conversation with my friend, I reached a fairly firm conclusion.

AI ad-buying tools are only going to matter more. How much more? So much that no one will be able to do without them.

Not because they're omnipotent. Exactly the opposite — right now they're still riddled with gaps and still need humans to tune them. But their direction is correct. Today, with data chopped to pieces by regulation, only automated optimization can help you react fast and keep adjusting. You can no longer rely on a human watching every detail.

So if your team is still agonizing over "should we adopt AI ad-buying tools," that question itself is already obsolete.

The real question to ask is: how do you plan to combine "human steering" with "AI driving," instead of treating it as either/or.

AI does the driving. But your judgment decides whether it's heading in the right direction.

After we finished talking that day, my friend sent me a message: "That kid I'm mentoring now — he's actually the kind who asks the right questions."

I messaged back: "So what you taught wasn't wasted. It's just being used differently now."