Advertising Hasn't Changed in a Decade. Until AI Dismantled It.

A while back, a friend asked me a question:

"So, where should I put my ad budget next year?"

I was about to brush it off with "the usual trio — search, social, e-commerce." But the words died on my lips.

Because I suddenly realized something.

For the past decade, the advertising playbook hasn't really changed. Brands fight for position in search results, jostle for visibility in social feeds, and compete on price across e-commerce pages. Same playbook, different packaging. Lots of surface-level variety, but the underlying logic comes down to one sentence: people move around the internet, platforms sit in the middle playing matchmaker, and advertising collects the toll along the journey.

But AI is prying that underlying logic wide open.

How? Let's get into it.

The Neutral Assistant Starts Wanting to Sell

Over the last couple of years, large language models have quietly become many people's first stop for searching, learning, and shopping.

BCG conducted a study. In 2025, shopping-related generative AI usage grew by 35%.

What does that mean? It means the tool people originally used to casually ask "is this fridge energy-efficient?" is becoming their front door to buying.

Think about it — when traffic flows somewhere, can money be far behind?

AI platforms are burning through cash. Subscriptions and enterprise fees alone won't sustain them. The subscription revenue is a rounding error compared to infrastructure costs.

Advertising is the virtually inevitable next step.

And this isn't me guessing. There are already moves afoot. OpenAI has publicly announced plans to test ads in ChatGPT in the United States. Meta is more direct — it's taking user interaction data from Meta AI and using it to optimize ad recommendations across its own platforms.

In plain terms, that "I'll help you look things up, I'll help you find answers" neutral assistant is growing commercial teeth.

This has already started. It's not future tense.

Three AI Interfaces, Three New Ad Properties

So here's the question. There are many kinds of AI — where will ads actually take root?

I've broken the current landscape into three categories. Once you hear them, it'll click.

Category one: AI embedded in search.

Google's AI Overviews, Perplexity, Microsoft Copilot. What they do is synthesize a pile of web pages into a single answer and hand it to you directly. You don't need to click any links — the information comes to you.

So where do ads go? Inside the answer. Google is already inserting Shopping ads and Search ads into AI Overviews, often ahead of the organic results. You used to be able to buy keywords and get visibility. Not anymore. You have to make your product data "legible" to the model — only then will it fold you into the answer.

Perplexity's approach is more straightforward: it generates a clearly labeled "sponsored follow-up" — who the advertiser is and why it's showing up, stated plainly. The transparency is quite interesting.

Category two: Assistant AI.

ChatGPT, Gemini, Claude, Meta AI. These run 24/7, helping you plan, research, and decide.

Right now, these AI assistants are basically ad-free. But the infrastructure is being built fast. OpenAI's ad testing is the signal.

However, ads inside an assistant can't just copy the search playbook.

Why?

Because when you're talking to an assistant, you treat it like a person. A consultant. If mid-conversation it suddenly shoves a hard sell at you, you feel betrayed.

One statistic left a deep impression on me: 69% of consumers feel that brands using AI in advertising without proactively disclosing it constitutes manipulation.

69%. That's nearly seven in ten.

So ads inside assistants have to take the form of "suggested next steps," "recommended tools," or "summary of best options." They need to feel like they're helping you decide, not interrupting your decision.

That's a razor-thin line. But whoever finds their footing first, wins.

Category three: Retail and commerce AI.

Amazon's Rufus, Walmart's Sparky, Instacart's Ask. These are AI assistants built by retailers, fed on their own first-party data.

The playbook is different from the first two categories. It's not about helping you "discover" — it's about helping you "buy."

Amazon was already placing ads in Rufus by 2024. Walmart followed suit, letting sponsored products appear naturally in Sparky's conversations.

Something even more worth watching is a project Google is pushing called Universal Commerce Protocol, or UCP.

What does it do?

It lets retailers' data, products, and even fulfillment capabilities be called upon and monetized outside the retailer's own website. In other words, retail media is shifting from "my own turf" to a distributed business that follows the goods — whoever calls the data, pays.

For brands, this means the old "grab the top spot" strategy is losing its edge. The new core question is: is your product data clean, can the model understand it, and will the model pick you when it makes a choice on the user's behalf?

Categories That Got "Rediscovered"

Speaking of which, I want to single out one thing.

In digital advertising, certain categories have always been notoriously hard to advertise. Finance, insurance, healthcare, enterprise software — these aren't things you buy after glancing at an image. Users need to research, compare, and ask questions.

The old banners, feed ads, and keyword campaigns couldn't tell these complex stories.

But AI conversational interfaces are naturally suited for exactly this.

Think about it — someone asks an AI "should I get term life or whole life insurance," and they can follow up, ask the AI to build a comparison, and request a recommendation tailored to their situation. The whole process is continuous, with full context intact.

This means high-consideration categories, for the first time, have an ad slot that can actually tell their story.

The more I think about it, the more fascinating it becomes.

New Rules, New Costs

But don't rush in.

The rules of AI advertising are completely different from what came before. They touch on something extraordinarily sensitive: conversation data.

What you say inside an AI chat carries far more information than what you type into a search box. Your preferences, your circumstances, what you just discussed — it's all in there.

Nearly 70% of consumers draw a red line around three things: private message content, health information, and precise location. These three are off-limits for AI.

Regulation is tightening too. Early proposals are already pushing for clearer disclosure, stricter labeling, and tighter boundaries on how conversation data can be used.

So trust isn't a soft metric in AI advertising.

It's the price of admission — your ticket to the table.

Four Possible Futures

BCG laid out four scenarios. I think they're worth pondering.

One: Search 2.0. Ads grow directly inside AI-synthesized answers, optimized by intent clustering, no longer dependent on keywords.

Two: Agentic commerce. AI handles everything from research to checkout on your behalf, and the core of retail media becomes "influencing the AI's default recommendation."

Three: Ambient advertising. There's no clear ad slot anymore — commercial influence is fully dissolved into algorithms and context.

Four: Heavily regulated neutrality. Rules box in targeting and optimization capabilities, forcing platforms to be far clearer about disclosure and separation.

Reality will likely be a blend of all four. This isn't an either-or proposition.

So What Should You Do Now?

I've been talking to several marketing leaders lately, and the shared sentiment is: this is moving fast, and unevenly.

Some sectors are already in motion; others are still watching from the sidelines. But the direction is set.

If you're the head of marketing, I think there are a few things you should be doing right now.

One: Get your ticket and start testing.

All the major AI platforms now have alpha and beta programs — push your way in. Ad slots in AI Overviews, sponsored answers, placements in retail assistants — get a foot in the door. Figure out how the model reads your product data, how it understands your category signals, and how it evaluates your creative assets.

At this stage, whoever deciphers the rules first, gets to set the terms.

Two: Align your internal operating system.

AI interfaces have a ruthless characteristic. They blur content, creative, and media into one thing. If your organization still has brand, performance, and retail running as three separate lines, you'll be terrifyingly slow.

Data needs to be clean. Creative and media processes need to be connected. Experimentation needs to be fast.

Three: Set your own rules before regulators set them for you.

Which conversation data can be used, what the labeling standards are, how you obtain user consent — get these aligned internally now. If you wait for regulation to land, it'll be too late.

Four: Start imagining what the "marketing org of the future" looks like.

GenAI is compressing planning, creative, and activation into one. The marketing organization of the future will be more integrated, more in-housed, and have more dynamic budget allocation.

This change won't happen overnight. But the seeds need to be planted now.

Back to That Question

At the beginning of this article, that friend asked me where to put the budget.

My answer: the usual trio still needs investment for now, but you need to carve out a piece of this year's budget for this new AI territory.

BCG's data shows that 53% of organizations are already setting aside dedicated budgets for conversational advertising, and nearly three-quarters of those plan to significantly increase spending over the next two years.

This isn't a question of "should we do this."

This is a question of "if you wait a year, the gap becomes years."

What advertising in the next decade will look like is being defined right now.

The people who define it won't be bystanders.