Ad Buying Is Being Rewritten — by Three Platforms at Once

A few days ago, a friend who runs an e-commerce business complained to me.

He said he'd been running Facebook ads for almost ten years. From audience targeting to bid optimization, he handled it all himself. From the year before last, he started noticing his workload shrinking. A thing called Advantage+ showed up in his account. At first he just played with it; then he kept moving more budget into it. By the second half of last year, he had basically one job left — uploading product images.

He asked me, is this normal?

I told him — totally normal. And it's not just him. Almost the entire industry is going through it.

Meta Has Boiled the Advertiser Down to Two Things: a URL and a Budget

Let's start with Meta.

Meta's vision for advertising in 2026 is so simple it's almost unsettling: you hand over a business URL, you set a budget, and the AI takes care of all of it.

What does "all of it" actually mean?

Creative — AI takes your product images and generates copy, variant images, and even turns up to 20 images into multi-scene video ads. Audience — Advantage+ strips out manual audience targeting; it uses conversion behavior, engagement signals, and cross-platform data to find high-value users on its own. Bidding, placement, budget allocation — all dynamically adjusted by AI.

This isn't some future roadmap. Meta has publicly said it intends to ship "fully automated ads" before the end of 2026, and these building blocks are already running today.

The results? Advertisers who consolidated scattered accounts into Advantage+ have seen CPA drop by as much as 32%. That number isn't a small fix — it's a structural gap.

Even more striking is the revenue figure. In a Q4 2025 earnings report, Meta's AI video-generation tools hit a $10 billion annualized run-rate, growing at almost three times the rate of the overall ad business. What does that tell you? Advertisers aren't dabbling — they're moving real money in.

By this year, Meta disclosed that 65% of its advertisers had already expanded their budgets into Advantage+.

If you're still manually running a full account structure, you're competing in the same auction against AI buying systems that are dramatically more efficient than you.

Google Killed the Thing You've Been Doing for Ten Years: Picking Keywords

Now let's look at Google.

Google's cut goes even deeper — it took the keyword research you've been doing for a decade and removed it.

Late last year, Google pushed AI Max into ad accounts, and by January of this year every Ads MCC (Manager Accounts, formerly My Client Center) in North America could switch it on. The product's logic is so simple it's almost funny: you provide a landing page, a daily budget, and a CPA or ROAS target. Everything else goes to Gemini.

It reads your landing page, understands what you sell, automatically matches it to user search intent, writes headlines and descriptions, and sets bids. The keyword list — gone.

For a whole class of advertisers who made a living off keyword research, this is nothing short of a career earthquake.

Even more interesting is AI Mode. In February this year, Google launched something called "shopping ads with Direct Offers inside AI Mode." In plain terms: a user asks a question in a conversational search, the AI returns an answer, and sponsored product recommendations are interleaved inside it. This is a brand-new paid channel — ads aren't being pasted next to AI text; the ad actually grows inside the conversation.

Why does Google dare to do this? Because it has the leverage. In Q4 2025, Google Search revenue hit $63 billion, partly because AI Overviews had already reached 1.5 billion users, and AI Mode had long been inserting ads. Early data says that on high-purchase-intent queries, the click-through rate of AI Mode ads is even higher than that of traditional search ads.

The signal underneath this is worth thinking about more than the surface: Google moved ads from "pasted next to search results" to "growing inside the conversation." This is the beginning of a new channel form, not a redesign.

TikTok Chose a Third Path: It Dismantled Creative Production, the Slowest Link in the Chain

Meta is retuning audiences. Google is killing keywords. TikTok didn't fight for either of those battlefields. It poured its AI bet somewhere else — creative production.

Why?

Because on TikTok, content gets consumed far faster than on the other two. Ad fatigue sets in quickly, and you have to keep feeding new creative. Creative production has always been the biggest bottleneck on this platform.

TikTok's answer is called Symphony, and it's plugged straight into Ads Manager. It does a few things:

• Image to Video: upload product images, reference images, or brand assets, and in seconds it generates a 5-second vertical video ad — feed it up to 20 images at a time.
• Text to Video: don't even have an image? A text description is enough to generate a video that fits TikTok's tone.
• AI avatar: upload a product image, pick a virtual persona, and an AI avatar presents your product on camera. No need to book influencers, no need to rent a studio.
• AI voiceover + localization: one asset is automatically translated and voiced into multiple languages, while preserving the original speaker's voice. One asset, global reach.

How direct is the impact? Brands report that on average content-production time has been cut by 70%.

TikTok also did something fairly clever. It connected Symphony with Adobe Express and integrated it with WPP Open. In other words: creative teams don't have to jump back into TikTok's backend every time — they can call these tools from inside their existing workflows.

This is a platform-neutral posture, but hidden behind it is a bigger judgement: the role of creative teams is shifting from "producing assets" to "directing AI systems." More on that below.

The Common Destination of All Three Platforms: You Set the Goal, AI Delivers the Result

By this point, have you noticed something?

Meta, Google, and TikTok look like three different products on the surface, but underneath they're running the same playbook.

They're all converging on one endpoint — advertisers only do two things: define business goals and supply creative assets. Everything in the middle — audiences, bidding, placement, budget allocation — AI handles.

This isn't a coincidence. The three platforms draw on different data signals (Meta has the social graph, Google has search intent, TikTok has content engagement), but they arrived at the same conclusion: manual buying can no longer keep up on efficiency.

What does this mean for advertisers?

It means the era of "winning by being good at tweaking accounts" is ending.

The new competitive moat has moved. From "knowing how to tune ads" to three things: clean first-party data, diverse creative assets, and clear business-goal signals. Whoever does these three well, AI amplifies their advantage. Whoever is still playing the 2024 playbook will see the gap keep widening.

A Counterintuitive Strategy: Consolidate Accounts, Don't Split Them Fine

On the operational side, there's a pattern that keeps showing up, and it's pretty counterintuitive.

Advertisers' instinct used to be — split. Split by audience, by product, by region; the finer you split, the more "precise" it felt.

In the AI era, it's exactly the opposite. The more data you feed AI, the better it optimizes. By shattering your accounts into pieces, you're effectively starving its AI.

Meta's own data shows that consolidating accounts with similar audiences into one Advantage+ lowers CPA by as much as 32% versus fragmented structures. Google's AI Max performs better in broader targeting than in narrow keyword groups. TikTok is the same — it rewards advertisers who run fewer accounts with higher individual budgets.

Put simply, less is more. One account, with enough budget and data stuffed into it, gives AI something to actually learn from.

If you're going to put this into practice, here are a few principles to follow:

• Meta: merge accounts that target similar audiences into one Advantage+; ideally keep no more than 5 active ad accounts under one consolidated structure, and let its budget optimization allocate on its own.
• Google: roll keyword-based search accounts into AI Max; open a separate Performance Max for cross-channel. Don't let two accounts bid against each other in the same auction.
• TikTok: merge ad groups with overlapping audiences, and pull the daily budget per account to at least 50 times your target CPA, so it exits the learning phase earlier.
• A rule that cuts across all three: deploy server-side conversion tracking (Meta's CAPI, Google's Enhanced Conversions) to feed cleaner data back. First-party data quality directly determines the speed and accuracy of AI optimization.

The Easiest Trap to Fall Into: Don't Touch the Learning Phase

Speaking of the learning phase, here's another counterintuitive point.

On AI-driven accounts, during the first week or two, CPA is usually higher than normal, and it bounces around. A lot of people's first reaction is — turn it off, change the settings, roll back.

Hands off.

AI needs to see roughly 50 to 100 conversions before it stabilizes. During this period, it's building its model. The moment you touch it, you push it back to the start of the learning phase.

For example, say your target CPA is $50. The learning-phase budget for an account should be planned around $2,500 to $5,000. Judge whether it's working above that threshold — don't call it broken after two days of bouncing.

The market isn't in a hurry — your patience is.

The Role of Creative Teams Isn't Disappearing — It's Morphing

Finally, something that's a bit wistful to talk about.

Creative production used to be the slowest link in the chain — scheduling, booking people, shooting, editing, review. AI tools have dismantled that chain. A team that used to test 5 to 10 creative variants can now test 50 to 100.

So are creative teams about to disappear?

No. The role has changed.

The creative director, who used to review assets one by one, now defines the boundaries of the brand voice, sets the guardrails AI can't cross, and picks the input assets. He's gone from director to system architect.

The production designer, who used to produce assets, has shifted into quality inspection — AI produces, humans review what crosses the line and what isn't good enough.

The copywriter, who used to write every single ad line, now defines the messaging framework and lets AI fill in the details within it.

AI hasn't replaced creativity — the layer of creative work has shifted up by one. Roles that can only execute, not judge, will be eaten. People who can define "what the right thing is" get amplified instead.

Data from Q4 2025 to Q1 2026 shows that on direct-response metrics, AI-generated creative has already caught up with or surpassed human creative in most categories. But on brand recall and emotional resonance, human creative still wins.

So a standard answer has emerged in the industry: AI handles volume and variant testing; humans handle brand narrative and direction. Two separate jobs, each doing what it's best at.

So What Do You Do Now

After all this, it comes down to one question: as an advertiser, how do you actually take this on in 2026?

My own judgement is that this is no longer a "should we adopt AI" choice. It's the default. All three platforms have pushed AI into the default position. If you don't follow, you're competing for the same users against AI systems that are a notch more efficient than you.

The real question has become: how well do you feed it what it needs.

Three levers, in order of importance:

First, first-party data. This is the new moat. The customer lists, purchase behavior, and value segmentation in your hands — these are things AI can't learn on its own and your competitors can't get. Clean them up and feed them back to the platforms through CAPI and Enhanced Conversions.

Second, diversity of creative assets. Not one good creative, but a pile of different ones, letting AI test which one runs. Symphony, Advantage+ creative tools — they're all built for this.

Third, clarity of business goals. AI doesn't know what you want unless you tell it. CPA targets, ROAS targets, conversion-event definitions — the clearer these are, the more AI can run in the right direction.

Everything else — let AI handle it.

My friend later told me he'd figured something out. The craft he'd spent ten years building — its value isn't in his hands, it's in his head. The hands-on work will be taken over by AI; the judgement in his head is his real moat.

When he put it that way, I thought — that's really well said.

Maybe you think so too. Maybe you don't. But there's one thing I'm pretty sure about: this isn't something that will happen ten years from now — it's happening right now. By the time we look back this Q3, the gap between manual accounts and AI accounts will most likely already be unbridgeable by a little catch-up.

The sooner you feed AI what it needs, the sooner it starts working for you.