The next phase of AI advertising won't feel like a new ad format. It'll feel like handing your media buyer a company badge, a budget, and permission to make decisions while you sleep.
That shift is already visible. Google says campaigns with automatically created assets or campaign-level broad match will begin moving into AI Max for Search on September 1, 2026. Gartner now predicts that more than 70% of global ad spend will flow through AI-influenced self-serve advertising platforms by 2028, while warning CMOs to strengthen independent measurement.
The important word is not AI. It's influenced. The platforms aren't just helping marketers build campaigns anymore. They're shaping who gets targeted, which creative gets made, where a click lands, how much a conversion is worth, and which budget deserves another dollar.
The Platform Wants the Whole Loop
For years, paid media was divided into visible jobs. A strategist chose the audience. A copywriter wrote the ads. A designer made the assets. A buyer adjusted bids. An analyst decided whether the money worked.
AI advertising platforms are compressing those jobs into one optimization loop. Give the system a goal, a feed, a conversion signal, and a spending limit. The platform finds opportunities, produces variations, expands query matching, routes people to pages, and reallocates money toward the outcomes it thinks you value.
Google's AI Max migration makes the change unusually clear. The September upgrade doesn't simply rename a campaign type. For some campaigns, automated asset creation will arrive with search term matching and text customization. That means the system can influence both sides of the transaction: what a person searches for and what the person sees in response.
The operational details of Google's announcement matter because they expose the direction of travel. Query expansion, machine-written copy, and final URL expansion are not separate conveniences forever. They are parts of a connected machine.
That connection is powerful. It's also where marketers lose the ability to see which decision created the result.
The New Control Problem
The old question was, "Did the campaign work?"
The new question is, "What did the platform actually do?"
Those aren't the same question. A campaign can hit its reported conversion goal while quietly changing the audience, broadening the query set, substituting creative, rerouting traffic, and spending more in placements your team didn't intend to prioritize.
The dashboard may show an acceptable cost per acquisition. It may not show that the cheapest conversions came from a low-value audience, that branded demand was counted as new growth, or that the system learned to prefer an offer your sales team can't fulfill profitably.
This is why the recent wave of AI advertising isn't mainly a creative story. It's a control story. The platform is making more decisions, but the advertiser still owns the legal exposure, brand promise, customer experience, and financial result.
The gap gets wider when the optimization goal is vague. "Maximize conversions" sounds precise until the system has to decide whether a form fill from an unqualified lead is better than a smaller number of high-margin purchases. The algorithm isn't confused. It is following the signal it was given.
Your measurement design becomes your media strategy.
That is the uncomfortable idea behind Gartner's forecast. Gartner's August 2026 prediction pairs the growth of AI-influenced platforms with a call for independent measurement. The two belong together. The more decisions the platform makes, the less reasonable it is to let the platform grade its own work.
The Creative Gets Flatter
There is another cost that won't show up in a media efficiency report: sameness.
If every advertiser feeds similar product data into systems trained to maximize predicted response, the output will converge. Headlines get shorter. Benefits get safer. Images get cleaner. The system learns what already performs, then gives everyone a polished variation of the same answer.
Google's machine-written text, Meta's automated creative plans, and AI-generated video treatments all promise speed. They can deliver it. But speed makes a weak idea easier to distribute, not a strong idea easier to find.
The creative director's job used to be partly about making a brand recognizable inside a crowded feed. In an AI-heavy buying system, that job becomes more strategic. Someone has to decide what the model must not flatten. Someone has to protect the strange phrase, the sharp opinion, the visual cue, or the proof point that makes the brand feel like a brand.
This is connected to the argument in AI advertising is becoming a media-buying system. The platform doesn't need to replace the creative team to reduce differentiation. It only needs to reward the safest version of what has worked before.
The brands that hold onto an edge will use automation for variation and distribution, not for taste. They'll let the system produce ten versions, then refuse to let performance data become the only definition of quality.
What Marketers Still Own
The answer isn't to turn off every automated feature. That would be a nostalgic response to a structural change.
The better answer is to move human control upstream, into the parts of the system that determine what automation is allowed to do.
The business objective. Revenue, leads, purchases, and visits are not interchangeable. Tie the optimization event to the economic result the business actually wants, even if that makes the platform's early performance look worse.
The evidence layer. Product feeds, landing pages, conversion events, exclusions, customer lists, and creative inputs are now instructions to the machine. Treat them like a source-of-truth system. Bad inputs don't merely produce bad reporting. They make bad media decisions at scale.
The boundaries. Write down where the platform can expand and where it can't. Define protected claims, regulated language, audience exclusions, landing page rules, budget ceilings, and placements that require human review. A guardrail that lives only in someone's memory is not a guardrail.
The independent check. Compare platform reporting with server-side events, finance data, customer quality, retention, margin, and incrementality tests. The platform can tell you what it optimized. It cannot be the only authority on whether that optimization mattered.
The brand constraint. Create a small set of non-negotiables that the system cannot reinterpret. The more creative control moves into the platform, the more important it becomes to define the words, images, offers, and promises that must remain unmistakably yours.
This is also why paid search is becoming a useful training ground for AI search strategy. As I wrote in Paid Search Is Becoming AI Search's Advantage, query data and conversion signals can teach a brand what people actually want. But learning from that data is different from surrendering every decision to the channel that collected it.
The Budget Becomes a Belief
A media budget used to express a plan. Put $100,000 into search, $50,000 into social, and the allocation said something about how the team believed demand would develop.
In an AI-influenced platform, the budget becomes a training signal. It tells the system which outcomes deserve more attention. It also tells the system what the company is willing to tolerate while it searches for those outcomes.
That makes budget governance more important, not less. A platform that can move money quickly can also amplify a mistaken assumption quickly. If the conversion event is weak, the model can become very efficient at finding more of the wrong people. If the product feed is incomplete, it can route demand toward the inventory that is easiest to understand, not the inventory that is most profitable.
The finance team should be closer to paid media than it has been. So should sales, customer support, merchandising, and legal. That isn't because every department needs to manage campaigns. It's because the campaign is increasingly connected to decisions those departments own.
The media buyer doesn't disappear. The role changes. The strongest operator becomes the person who can translate business intent into clean signals, useful constraints, and honest tests. Less time is spent adjusting bids by hand. More time is spent asking whether the system is learning the right lesson.
The Advantage Is Judgment
The 2028 forecast may be right about where the money goes. It doesn't answer who will get the value from that shift.
The easy prediction is that the platforms will win because they own the data, the models, and the distribution. That is probably true if advertisers keep treating automation as a black box that produces convenient reports.
A different outcome is possible. Brands can use the platforms for reach and machine-scale execution while building their own layer of evidence, constraints, creative conviction, and financial measurement around them.
That layer will feel slower at first. It will require uncomfortable conversations about what counts as a conversion, which audiences deserve investment, and whether an efficient campaign is actually creating demand. It may also produce fewer impressive dashboard screenshots.
Good.
The future media buyer is not the person who knows every platform setting. It's the person who knows which decisions should never be delegated, which signals deserve trust, and when a cheap conversion is just an expensive mistake wearing a green label.