The Ad Industry Hasn't Changed in a Decade — Until AI Sat Down at the Table

A while back, I was using ChatGPT to plan a trip.

I asked it: a weekend getaway, with kids, a budget of two thousand yuan — where should we go? It didn't throw ten blue links at me. Instead, it gave me three complete plans, each with flights, hotels, and the best-reviewed family restaurants in the area.

That stopped me for a moment.

I realized: I hadn't opened any search engine. I hadn't scrolled any social feed. I hadn't even visited a travel platform. My searching, comparing, and deciding — all of it happened inside a single chat window.

You may have had a similar experience.

And that raises a question every marketer should seriously think about: if users start getting used to doing everything inside an AI conversation, where does advertising go?

What Hasn't Changed in a Decade Is Starting to Crack

For the past ten years, the underlying logic of advertising hasn't really shifted.

Brands fight for rankings in search results. They fight for attention in social feeds. They fight for placement on e-commerce platforms. The channels evolve — from text to images to short video — but the essence has always been the same thing: users are "browsing," platforms are "the middleman taking a cut," and ads are "monetizing that journey."

This logic ran for a decade. Everyone took it as gospel.

But AI is rewriting the formula.

Large language models are no longer just a tool to help you write emails or polish copy. They're becoming the default entry point for how people search for knowledge, discover products, and make purchasing decisions. One 2025 study showed that generative AI usage for shopping-related tasks grew 35% year over year.

What does that mean?

It means the front door of the internet is moving.

When hundreds of millions of people start getting used to "asking AI" instead of "searching the web," commercial monetization will inevitably follow. It's the exact same path that search engines and social media walked — first gather attention, then sell ads.

The large model is evolving from a neutral copilot into a system with commercial interests.

53% of Companies Are Already Paying Up

Marketers are reacting faster than I expected.

Conversational advertising has already emerged as its own budget line in many companies' media mixes. It's not counted as search advertising, and it's not programmatic — it's a separate allocation entirely.

53% of organizations have already invested budget here, and nearly three-quarters of them plan to significantly increase spending in the next two years.

That number surprised me. A new ad format doesn't usually penetrate this fast. But think about it — how many years did it take for search engines to be taken seriously by the ad industry? And social media? The user growth rate of AI assistants is far outpacing both.

So the question is no longer "will advertising appear inside AI?" It's "in what form will it appear, how will it be governed, and how will brands compete within it?"

Three Interfaces, Three Layers of the Cake

Consumers are spending less time on traditional web pages. They're increasingly completing the entire journey — from discovery to comparison to purchase — directly inside AI interfaces.

This has given rise to something new. You could call it the "AI attention stack" — the place where time and intent pool together, and where ad inventory will naturally grow.

Specifically, three types of interfaces are taking shape:

The first: search-embedded AI. Tools like Google AI Overviews, Perplexity, and Microsoft Copilot compress information from multiple sources into a single synthesized answer. You ask a question, it gives you the answer directly — not a list of links.

The second: assistant-native AI. ChatGPT, Gemini, Claude, Meta AI. These embed themselves into your daily work and life flow. They help you plan, research, and make decisions. You're chatting with it, and somewhere in the conversation, it recommends something.

The third: retail and e-commerce AI. Amazon Rufus, Walmart Sparky, Instacart Ask — shopping assistants trained on retailers' proprietary data. You're no longer scrolling through product grids. You tell the AI "I want a dishwasher that fits a small apartment," and it picks for you.

Across these three interface types, advertising is emerging in three forms:

In-answer advertising — you ask a question with commercial intent, and the AI embeds a sponsored item right inside its synthesized answer.

In-conversation advertising — while you're exploring a topic with an AI, a relevant recommendation pops up alongside, even if you had no intention of buying anything.

Agentic advertising — when the AI is planning, comparing prices, or placing orders on your behalf, sponsored options appear naturally in its recommendations.

Each layer has its own rules and its own competitive dynamics. To win in any of them, knowing "where the attention is" isn't enough. You also need to understand how the model interprets your product and how it judges commercial relevance.

Search Is Being Torn Down and Rebuilt

The foundational logic of search advertising is undergoing its biggest rewrite since paid search was born.

In the past, getting seen was about keyword coverage. Buy enough keywords, bid high enough, optimize enough landing pages.

Now? Google's AI Overviews is already tentatively inserting Shopping ads and search ads directly into AI-generated answers, ahead of organic results. Microsoft Copilot turns search into a continuous conversation, with Bing's advertising system running underneath. Perplexity is testing a more direct approach — generating follow-up questions marked as "sponsored," telling you plainly this is an ad.

The direction is remarkably consistent: discovery is becoming "synthesized answers" — more conversational, and more commercialized.

For brands, this means what you need to worry about has changed. You used to worry about keywords. Now you need to worry about how the model "understands" your product data, your creative assets, and your bidding strategy. The logic of being seen has shifted from "keyword matching" to "model comprehension."

Assistants Are Becoming Billboards — But There's a Big Problem

General-purpose AI assistants are embedding themselves rapidly into people's daily routines, from planning trips to researching solutions, from executing tasks to buying products.

OpenAI has already announced plans to test ads in ChatGPT, starting with US users. Meta is even smarter — it uses your interaction data with Meta AI to improve ad precision across all its platforms. The line between ad and assistant is blurring.

But what do ads inside an assistant actually look like?

They will absolutely not look like search ads.

They need to be part of the conversation. Maybe a "next-step suggestion." Maybe a "recommended tool." Maybe a seamless transition to a transaction.

Think about it — when someone treats an AI assistant as an advisor, even as a friend, any hint of a "sales pitch" gets magnified. One study found that 69% of consumers feel manipulated when brands use AI to advertise without disclosing it.

Sixty-nine percent. That's a high number.

So the limiting factor in this game isn't technology. It isn't budget. It's trust.

Users treat the assistant as an advisor. Slip an ad into the advisor's mouth without telling them, and if they catch on, the cost is far greater than running a banner in a social feed. Whoever can achieve clear disclosure, genuine relevance, and disciplined restraint will win.

The Next Hurdle for Retail Media

Retail media has reached an inflection point too.

In the past, when you invested in retail advertising, you were buying attention on the retailer's own turf. Amazon's homepage recommendation slots, Walmart's search results pages — that was the logic. But AI-mediated search is changing this. When a user lets an AI intermediary make the judgment call about "what to buy," they might never visit the retailer's website at all.

Retailers' response? Build their own shopping AI.

Amazon added advertising to Rufus in 2024. Walmart just announced ads in its shopping assistant Sparky. Sponsored products can now appear naturally in conversational shopping flows.

Even more interesting is a standard Google is pushing called the Universal Commerce Protocol. It allows retailer data, product selection, and fulfillment capabilities to be accessed and monetized by external AI systems — no need for the user to go back to the retailer's own website.

This means the economics of retail media are being redefined.

For retailers, value shifts from "monetizing on-site traffic" to "participating in distributed, agent-driven commerce." For brands, the advantage shifts from "winning placement and traffic" to "being selected by the model." What matters now is data quality, availability, and relevance.

Put bluntly: you used to pay for position. Now you have to make the AI think you're worth recommending.

Some Categories Have Been Waiting for This

For certain product categories, AI-driven interfaces might be a genuine turning point.

Digital advertising has always excelled at one thing: driving immediate transactions. A pair of shoes, a cup of coffee, a phone case — nice image, right price, one click and it's bought.

But there's a whole swath of categories that feeds, banners, and keyword ads could never properly explain. Financial services, insurance, healthcare, enterprise software — the purchase decisions in these categories require research, explanation, and trust-building. You can't lay out the clause differences of an insurance product inside a 300x250 ad unit.

Conversational AI changes that.

Users can explore complex questions within a single interface. Ask follow-ups, weigh trade-offs, compare options — the entire process takes place in a trusted environment. When advertising appears in these contexts, it's no longer an interruption. It's more like decision support.

For categories driven by "understanding" rather than "impulse," AI assistants offer an entirely new way to participate in the purchase journey.

New Rules, New Red Lines

Consumers have red lines when it comes to embedding ads in AI-generated answers.

Nearly 70% of people believe there's certain data AI should never touch: private messages, health information, precise location. This sensitivity creates a triple risk:

Content-level risk — when AI models synthesize third-party content, paid influence gets embedded inside, diluting the credibility of original content.

Data-level risk — every word you say in a conversation could become raw material for ad targeting.

Algorithmic neutrality risk — organic recommendations and sponsored recommendations get tangled together. Users can't tell which is the AI's genuine opinion and which was paid for.

The regulatory shoe hasn't dropped yet, but the direction is clear: disclosure requirements will tighten, and the use of conversation data will be restricted. Platforms and brands that achieve transparency first will have an advantage when the rules harden.

What Does the Future Hold? Four Scenarios

The future isn't a single possibility. Here are four scenarios:

Scenario one: Search 2.0. Ads get embedded directly in synthesized answers. The optimization target shifts from keywords to intent clusters.

Scenario two: Agentic commerce. AI agents handle the full journey from research to purchase. The core of retail media becomes "influencing the agent's default recommendations."

Scenario three: Ubiquitous commercial influence. Advertising is no longer a standalone "ad placement." It's distributed throughout the algorithm's contextual judgment — everywhere and nowhere.

Scenario four: Regulated neutrality. Regulators step in to limit precision targeting and optimization, mandate separation of organic and sponsored results, and require clear disclosure.

Reality will likely be a blend of all four. That means being prepared matters more than predicting correctly.

So, What Should You Do?

Change is coming fast. The migration from links and feeds to answers and agents is uneven in pace but firm in direction. The next 18 months are the critical window.

If you're in marketing, here's what you should start doing now:

First, get a seat at the table and start experimenting. Alpha and beta programs on emerging AI platforms — squeeze your way in. Test how ads perform in AI overviews. Test the effectiveness of sponsored answers. Test how product data renders in retail assistants. Once something works, scale fast. Don't wait for the ecosystem to mature before you enter — by then, the seats will all be taken.

Second, build an AI-native media operating system. Get your data hygiene in order. Integrate creative and media workflows. Make rapid experimentation the norm. AI interfaces break down departmental walls — they blend content, media, and creative together. If your organization is still siloed, you'll be a step behind the competition.

Third, set rules for conversation data and disclosure. What data can be used, what can't, how to label ads, how to get user consent — establish these rules now. In AI interfaces, trust isn't a soft metric. It's your entry ticket.

Fourth, start reimagining your marketing organization. When generative AI compresses planning, creative, and media buying into one function, the marketing org of the future will be more integrated, more in-house, and more dynamic in resource allocation. If you don't start thinking about this now, you'll be dragged along later.

---

For the past decade, the keyword for the advertising industry has been "channels." Search, social, short video, livestream — every time a new channel emerged, everyone rushed in.

But AI isn't bringing a new channel.

It's bringing an entirely new commercial environment.

The logic of channels is: "advertising follows wherever the people go." The logic of AI is: "AI makes decisions on behalf of people, and you have to make the AI think you're worth it."

These are two completely different things.

In the next era of advertising, the winners won't be the ones with the most ad placements. They'll be the ones who best understand how models make sense of the world.

This game has just begun. But the chips are already on the table.