65% of Ad Operations Are Already AI-Driven. Beyond Meta Advantage+'s 'Full Automation': What Marketers Actually Do

65% of Ad Operators Are Already Using Advantage+

As of April 2026, one fact is settled. 65% of Meta advertisers have already adopted Advantage+ (DigitalApplied).

Two out of three are already entrusting their ad operations to AI. Some of you reading this number might say, “I haven’t touched it yet.” That caution is understandable. I was the same way at first.

Still, the numbers don’t lie.

Advertisers who adopted Advantage+ cut their CPA (cost per acquisition) by up to 32% (AdStellar). This is the result of campaign consolidation. If your monthly ad spend is 1 million yen, that translates to a 320,000 yen efficiency gain.

Your competitors are reaping this benefit. Continuing manual operations in that environment isn’t “staying the course”—it’s effectively “accepting a 32% disadvantage,” wouldn’t you say?

Why has it spread so widely? The reason is simple: “easy setup” combined with “real results.” Recall traditional ad operations. Setting target age, gender, and interests. Preparing multiple creative variations. Choosing placements and structuring bidding strategies. Advantage+ is a system that hands all of these steps over to AI, reducing effort while improving outcomes. The reasons not to use it are running out.

If you’re thinking, “So we don’t need ad operators anymore?”—rest assured. The conclusion, stated upfront: what becomes unnecessary is “execution,” not “judgment.” This distinction is the core of today’s article.

In Part 1 of the AI Chat Ads series, I explained the structure where conversational data started being used for ad targeting. In Part 2, I introduced 7 patterns where the “quality of input” you feed Advantage+ determines results.

For Part 3, I’ve raised the perspective one level. In a world where 65% are already using it, what work remains for marketers? After accurately understanding what “full automation” actually contains, I’ll walk you through the 3 judgment skills you should sharpen.

Ads Run on “URL + Budget” Alone. What Full Automation Actually Looks Like

Meta is moving forward with plans to complete “full automation” of advertising within 2026 (Marketing Dive).

What specifically does full automation refer to?

You input your business URL and budget target. That’s it. The AI handles creative generation, audience selection, placement optimization, and bid management—all end-to-end (VXTX).

img: Meta Advantage+ full automation flow diagram. Input box “URL + Budget” on the far left, “Delivery Optimization” on the far right. In between: creative generation, audience selection, placement | type: diagram | style: clean flowchart with arrows, minimalist design

Picture it concretely. Say you run an e-commerce site. You paste the URL of a new product page into Advantage+. You input “monthly ad budget 500,000 yen, objective: site sales.”

The AI then reads product images, text, and price information from the URL automatically. It generates multiple ad creative variations. It delivers them at sizes optimized for Facebook Feed and Instagram Reels. It continuously optimizes bidding toward users with high purchase probability.

Steps that previously consumed a full day of an ad operator’s time complete in minutes. The AI keeps optimization running 24/7 without rest. The density is utterly different from operations where you check numbers each morning and make adjustments.

At present, the complete “URL → instant delivery” form is being tested with a subset of advertisers. That said, targeting and budget automation are already running as standard features. Once fully automated creative generation is added, “full automation” will be complete—so the pieces are essentially all in place.

Worth noting here is Meta’s strategic intent. Full automation has particularly large impact for SMBs and sole proprietors. The segment that couldn’t get started “because I don’t have ad operations skills” will be able to launch ads just by pasting a URL. As more advertisers join, Meta’s ad revenue grows too. It’s a structure that benefits both sides.

On the other hand, as more entrants join, competition intensifies. If everyone uses the same AI, the difference comes from “what you feed the AI.”

Think about it this way. As a cooking analogy, it’s like everyone gaining access to the same high-performance oven. In that case, the difference comes not from oven performance, but from judgment about what to bake and which ingredients to choose. The exact same structural shift is happening in advertising.

If everything gets replaced by AI, what remains for marketers?

20 Images Become Video. The Capabilities of Video Generation 2.0

The element with the greatest impact within full automation is the automation of creative generation.

Meta’s Image-to-Video feature deserves attention. It automatically generates video ads from up to 20 product images (1ClickReport). Already used by over 4 million advertisers, it functions as a standard tool rather than an experimental one.

“I want to make video ads, but I can’t afford filming costs.” This feature addresses head-on a worry that SMBs and freelancers have long carried. We’ve entered an era where video ads can run as long as you have still images.

Video Generation 2.0, rolling out in 2026, has 3 distinguishing features.

Multi-scene generation. A feature that strings together multiple cuts with cinematic transitions into a single video. Upload 5 product images, and each becomes its own cut in the composition. A 15-second video ad gets completed without a production crew.

Dynamic text overlay. A feature that automatically extracts text from your ad copy and layers it onto the video with animation. For instance, text like “Limited time 20% OFF” fades in over a product image. The design work for static banners becomes entirely unnecessary.

Smart cropping. A feature that automatically detects the product based on placement and centers it. Reels uses a 9:16 vertical format; Feed uses 1:1 square. The effort of resizing per placement drops to zero.

img: Comparison table of Video Generation 2.0’s three features. Left column shows “Multi-scene generation,” “Dynamic text,” “Smart cropping”; right column shows traditional manual work (production crew | type: comparison-table | style: clean two-column table, professional design

“Aren’t AI-made videos cheap-looking?” This doubt is natural. I thought so at first too. When you actually check the output, there are cases where quality exceeds what a human designer rushes to produce. Vertical videos for Reels are an especially strong area for AI.

Of course, it’s not suited for brand advertising where you want strict control over the worldview. Even so, for direct response ads aiming for direct conversions, the quality is more than sufficient. The very ability to discern “which uses can be entrusted to AI” is the skill required of marketers going forward.

The “3 Judgments” Automation Can’t Replace

Here’s where it gets to the main point.

What Advantage+ automates is the “execution” layer. This includes targeting, creative generation, bidding, and placement optimization. AI is faster, more accurate, and never tires in these areas. Better to admit it’s not a domain where humans can win.

On the other hand, the “judgment” that happens before what gets handed to AI is still human work. In a world that’s crossed the 65% watershed, differences in this judgment skill translate directly into differences in results.

Judgment 1: Choosing What to Sell

When you hand “URL + Budget” to Advantage+, humans decide which URL to hand over.

Which product gets pushed as this month’s flagship? Where do seasonal inventory priorities sit? Considering competitor movements, which service page should be the landing page? Only humans who understand the business context can make these decisions.

Say you’re running an apparel e-commerce site. In April, do you push new spring arrivals, or move inventory clearance items first? Do you prioritize profit margin, or cash flow? This judgment requires “business intent” that AI doesn’t possess.

AI is good at optimizing ads for a given URL. That said, it can’t judge whether that URL is the right choice. This is the first point in the remaining 35%.

In fact, most of the failures I’ve seen with clients I support are cases where “we put the best budget on the best product page, but it didn’t sell.” The reason is always the same: “that product wasn’t what current customers needed.” AI didn’t get it wrong—the choice of materials handed over was wrong.

Judgment 2: Resolution on Who You Want to Reach

“Doesn’t AI handle targeting?” Many people think so. Indeed, Advantage+‘s audience selection is highly accurate.

What needs caution is that AI optimizes for delivery to “people likely to convert.” This doesn’t necessarily match “the people you truly want to reach” as a brand.

Is this a phase where you prioritize acquiring new customers? Or a phase where you want to raise LTV (lifetime value—total profit one customer generates across their entire lifetime) of existing customers? When you delegate this strategy-level judgment to AI, it tends to skew toward short-term CPA optimization. You could call this “the automation trap”—the better the numbers get, the harder it is to notice.

In my experience, when I gave Advantage+ full authority, delivery tended to skew toward retargeting existing users. Short-term CPA drops, but the pie of new acquisitions doesn’t expand. That’s why the strategic judgment of “this month, we allocate 60% to new acquisition” needs to be set by a human first.

The “quality of input” I explained in Part 2 of the series ties directly to this judgment. If the design of signals fed to AI is sloppy, the direction of optimization gets misaligned. Conversely, when you carefully design first-party data and CAPI (Conversions API—data sent to Meta via server). When you do that, AI moves as expected.

Judgment 3: An Eye for Brand Consistency

Video Generation 2.0 is excellent. Even so, whether generated creative aligns with your brand guidelines is hard for AI to judge.

Color tone, choice of words, consistency of worldview. Because they’re hard to quantify, they’re a domain that doesn’t easily become an AI optimization target. The judgment to approve an AI-made ad with “this is okay to run” can only be made by a human who understands the brand.

Let me give a concrete example. I once had AI generate ads for a brand whose selling point is luxury. The CTR (click-through rate) was solid. However, the output creative used pop fonts and casual color schemes. The finish was far removed from the brand image. By the numbers alone it was a “success,” but considering the brand-damage risk, it was a “failure you can’t overlook.”

Meta’s policy is also to retain creative review in the post-full-automation workflow (AdTaxi). Meta itself doesn’t recommend “leaving it entirely to AI.” This fact carries weight.

Building approval workflows becomes the foundation of brand management in the automation era. Don’t begrudge the time spent on review. The accumulation of these judgments is what nurtures the vast volume of AI-generated creative into something with “your company’s character.”

img: Marketer’s three judgment skills expressed in a triangle diagram. Top vertex: “What to sell (product selection),” bottom left: “Who to reach (strategic targeting),” bottom right: “Brand consistency (creative | type: diagram | style: clean triangle infographic, professional design

What I Noticed About the “Remaining 35%” Through Advantage+

“I get the theory, but how is it actually?” Some of you may be thinking this.

There are things I can share from my own experience running campaigns with Advantage+.

Honestly, I was scared at first. Handing over targeting to AI felt like driving a car with your hands off the wheel. You can’t help thinking “I’d feel safer setting it myself.” The longer you’ve been in marketing, the stronger this “fear of letting go” tends to be.

When I actually tried it, AI’s accuracy exceeded my expectations. Click-through rates improved compared to manual setting, and CPA dropped. I still remember the moment I thought “Oh, this is something I can let go of.”

On the other hand, there was plenty I noticed too.

AI mass-produces “ads with good numbers.” That said, it doesn’t make “the ads I want to put out as a brand.” One day, looking at generated creative, I felt “the numbers are good, but this isn’t our worldview.”

That sense of discomfort is exactly why human judgment is needed, I believe. AI handles CTR and CPA optimization. But it doesn’t answer the question, “Can I be proud of myself for running this ad?”

After I became conscious of the “7 patterns of input quality” introduced in Part 2, the quality of materials I fed AI changed. As a result, the direction of generated creative improved. Rather than “managing” AI, it might be closer to the feeling of “handing over good materials so it can do good work.”

I feel this mindset isn’t limited to advertising. With Claude Code, which I use regularly, the quality of output changes based on how you give instructions. Even when AI’s performance is the same, the precision of information you hand over makes a big difference in the results.

If I had to name one thing you can start right now, open your Advantage+ settings screen. Check campaign consolidation, and if there are ad sets scattered separately, consolidate them into one. That alone changes CPA. Keep screenshots of the numbers before and after consolidation, and you can use them for internal reporting too.

One more thing—check Advantage+‘s “Asset Customization” feature. Input your brand colors and font settings, and the direction of generated creative stabilizes. This is the practical implementation of the “quality of input” I explained in Part 2. It’s a 5-minute setup, so try it today.

Summary—Which Side of the 65% Watershed Will You Be On?

Meta Advantage+‘s adoption rate has reached 65%. Within 2026, full automation with “just URL + budget” is moving toward reality. Technology that auto-generates video from images is already being utilized by over 4 million advertisers.

This flow won’t stop. The adoption rate, which was around 50% in 2025, hit 65% in just a few months. It wouldn’t be surprising if it exceeds 80% by year-end.

But the more automation advances, the higher the value of human judgment rises. This isn’t a contradiction—it’s a structural inevitability. Precisely because AI takes on “execution,” the scarcity value of judgment that decides “what to execute” grows.

The 65% watershed—keep this phrase in mind. Now that more than half of ad operators have shifted to AI-driven operations, the difference doesn’t come from “what to entrust to AI.” It comes from the precision of “what to decide before entrusting to AI.”

In Part 1, you understood the structure of “conversational data becoming ad fuel.” In Part 2, you learned the mechanics of how “input quality determines results.” In this Part 3, I’ve organized “the judgment skill that lies beyond automation.” Whether or not you’ve grasped these three creates a difference in results even when using Advantage+.

I’ll close with 3 questions you can check starting today.

Product selection: Is the URL you’re feeding this month truly the page of the product you should be selling this month?
Strategic targeting: Does the AI’s optimization direction match the segment you want to reach?
Brand approval: Could you put your company’s name on the generated creative without embarrassment?

If you can instantly answer “yes” to these three, you’re already on the other side of the 65%. If you can’t answer instantly, start there first. One at a time is fine. You don’t need to change everything at once.

I’m still in the middle of trial and error too. There’s no “perfect answer” for how to engage with AI. But standing on the 65% side, you keep sharpening the 35% only you can do. I’ll continue sharing that process here.

The AI Chat Ads series wraps up with this installment. Next time I plan to deliver “GEO Practical Guide 2026—How to Change Citation Rates in the First 40-60 Words.” I’ll introduce how to write articles that get cited by AI, distilled into practical steps.