The Craft of Ad Buying Is Being Rewritten by AI

A while back, a friend of mine who works in cross-border e-commerce asked me out for coffee. He runs ad accounts with an annual budget in the tens of millions (RMB).

The first thing he said when he sat down: these days, opening the Google Ads backend doesn't feel like running ads. It feels like herding a pack of interns.

I told him: that feeling of yours is dead-on.

Because in 2024, the assembly line of ad buying was rebuilt by AI from end to end. And not conceptually—this was done cut by cut, all the way down to the level of the buttons. The two engines behind the biggest changes: one is Google Ads, the other is Meta Ads.

Let's take it apart piece by piece.

First, Google: Letting Go with One Hand, Opening Windows with the Other

What does AI-powered ad buying actually mean?

Simply put: the machine has taken over the work you used to do by hand.

Start with the creative side. Making a single ad image used to mean booking a designer's schedule—cutting out an object, swapping a background. Since 2024, AI inside Google Ads can do all of this directly: remove the things cluttering the frame, replace the background, then automatically adapt one image into every layout Search, Display, and Demand Gen call for. You can even toss it a few reference images and have it steer toward your brand's look and feel.

One image edit: from half a day to one minute. That's the creative side.

But the really interesting part isn't that it does the work for you.

It's that it has started explaining itself.

You have to understand what Performance Max used to be: a black box. You threw in money and creative, it spat out results, and told you nothing about what happened in between. For people who run ads, this was the most frustrating thing of all: it's not that I'm lazy—it's that I can't see.

In 2024, Google cut two windows into that black box.

The first window: asset-level data. Which headlines are running well, which images are dragging the rest down—conversion details laid out right in front of you. It even makes specific suggestions: add a few more headlines, throw in a few more images. At last you can go after the weakest asset in the set instead of staring blankly at the campaign as a whole.

The second window: impression share. What is impression share? Among the relevant searches out there, the share in which your ad was eligible to show up.

What is this number good for? Immensely. It answers a question every media buyer has asked at some point: should the budget go up, or not?

Raising budgets used to be guesswork. Now you glance at impression share: still riding high? Then the ceiling isn't reached—more money buys more exposure. Already maxed out? Then more money only buys diminishing returns—the cash goes out, and your ad doesn't show its face one extra time.

A budget isn't there to be spent. It's there to buy. Money that can't buy exposure and conversions shouldn't be spent—not a cent of it.

There are two more moves that show Google reeling a little control back in.

One: campaign-level negative keywords, which back in 2024 were still in gradual rollout. Your ad showing up in front of more people isn't automatically better—only in front of the right people is it better. Blacklist the irrelevant search terms, and the money stops leaking through the cracks.

Two: in Demand Gen, you can pin assets. For high-visibility placements like YouTube and Gmail, the flagship creatives have to be your call. The face of your brand cannot be swapped out by the machine on a whim.

See the pattern? On one side, Google hands more and more of the operations to AI; on the other, it has to hand more and more of the explaining back to you.

The deeper the automation, the more transparency is worth.

Now Meta: Combinations Go to the Machine, the Problem-Setting Stays with You

Meta Ads takes a different route.

It's betting on one thing: a single ad shouldn't have just one look.

Ads used to be one line of copy paired with one image, distributed to everyone—everybody saw the same version. Now you can feed in several sets of copy and several versions of creative, and the system, drawing on platform and per-user data, automatically serves each person the combination that fits them best.

What does "a different ad for every person" look like? This. The version you scroll past and the version your colleague scrolls past may not be the same ad at all—but both come from your one ad.

Pair that with ever finer targeting: audience, behavior, interests, drilled down layer by layer. What Meta is doing is breaking "reaching the right people" into pieces fine-grained enough for a machine to handle.

AI has seeped into the daily operations, too. For ads already running, it proactively suggests optimizations to push ROI up; for copywriting, a few clicks produce several versions, all tuned for performance.

My friend's verdict on all this: it's like hiring a copywriting intern who works for free—and never sleeps.

Goodness, what a line. But it really is a vivid one.

So What Are the Ad Buyers For?

If the machines do all the work, what do the people do?

My take: the doers are being freed, and the problem-setters are being amplified.

Think about what AI actually took over: editing images, pairing assets, searching combinations, offering suggestions. All execution-layer work. But there are four things it cannot take away: which direction to test, what material to feed in, how to read the data, and when to hit the brakes.

In 2022 and 2023, AI mainly helped marketers write content. By 2024, it had formally stepped in to manage the money. And that line is only going to draw itself deeper over the next couple of years.

So the question "will AI replace me?" is asked backwards. The question you should really be asking is: when the machine runs at full speed, are the problems you set worthy of that speed?

The more automated the tools, the more your judgment is worth.

Back to That Cup of Coffee

With the coffee drained, my friend said: got it. I'm not an ad buyer anymore—I'm a coach. My players run faster than I do; my job is to make sure they run in the right direction.

Exactly. That's it.

Here's wishing you: every batch of creatives you feed out has someone there to catch it, and every cent of budget you spend actually buys something.