Marketing isn't dying because of AI

Posted on: 25 February 2026

There is a scene that repeats at every industry conference in 2026. A speaker takes the stage, opens a slide with a declining graph, and delivers the line that has become liturgy: "AI is changing marketing." The audience nods. Twenty minutes of solutions, tools, frameworks, and strategic pivots follow. Nobody asks the obvious question: what if AI isn't the problem?

The numbers are brutal and no longer contested. Organic traffic has collapsed by 15 to 64 per cent across categories where AI systems provide direct answers, according to data collected by Seer Interactive from over one hundred thousand keywords analysed in late 2025. DMG Media, the publishing group behind the Daily Mail, reported an 89 per cent drop in click-through rates on pages where Google's AI Overviews appear. The Big Six advertising holding companies saw their share of US ad spending fall from 44.6 per cent in 2019 to 29.6 per cent in Q1 2024, according to Advertiser Perceptions. People in the industry know these numbers. What they are reluctant to admit is what the numbers actually mean.

They mean that the business model of digital marketing was never built on the value it created for the customer. It was built on a market inefficiency. For thirty years, buying anything online required enormous manual effort from the consumer: searching, comparing, reading reviews, opening twenty browser tabs, going back, starting again. Every single step of that journey was monetisable. Every click was advertising inventory. Every hesitation was a retargeting opportunity. The funnel was never a model of how buyers actually behave; it was a model of how the industry that watched them extracted value.

AI did not kill this model. It simply made the model visible by eliminating the friction it depended on.

When a buyer asks ChatGPT "which cybersecurity platform suits a five-hundred-person financial services company needing SOC 2 compliance and Microsoft integration", they receive in thirty seconds what previously took weeks of research. The awareness, consideration, and decision stages that generations of marketers learned to occupy and monetise collapse into a single conversation. The data confirms it: traffic originating from AI converts at three to eight times the rate of traditional traffic, because it arrives with intent already formed. The "messy middle" that Google itself described in 2020, that limbo between exploration and evaluation where brands could intercept and influence, has not been optimised. It has been abolished.

Anyone with a long enough memory recognises the mechanism. It is identical to what happened to financial brokers between the late 1990s and early 2000s. For decades, brokers justified commissions that now seem surreal because "market access" was a rare competence. You needed intermediaries, proprietary platforms, relationships. Then online trading arrived and access became a commodity. Brokers who sold access disappeared within a single economic cycle. Those who sold intelligence, the ability to interpret and decide, survived and in many cases thrived. The parallel with marketing is almost embarrassing in its precision: agencies that sold "access to the customer" by managing funnel complexity are experiencing exactly the same collapse. Those that sell genuine strategic capability, the ability to understand why a customer buys rather than merely how to intercept them, face an opportunity that has never been larger.

But here is the paradox that nobody in the industry has any interest in naming. Marketing is trying to sell the cure for a disease it helped create. Look at what agencies are offering in 2026: "AI optimisation" services, "generative engine optimisation", strategies for getting cited by AI systems. They are effectively telling clients: "Pay us to solve the problem our previous model let you ignore, namely that your product needs to be genuinely good and your communication genuinely useful, because you can no longer buy visibility through keyword stuffing and aggressive retargeting." There is something circular in all of this that ought to give pause to anyone signing a consultancy contract.

The most revealing data point comes from Jasper's State of AI in Marketing 2026 report. Last year, 49 per cent of marketers said they could demonstrate the return on investment of their AI initiatives. This year, that figure has dropped to 41 per cent. Not because AI is delivering less value, but because productivity gains alone are no longer sufficient: management now wants measurable business outcomes. In other words, AI has stripped away a problem that always existed. For years, marketing could hide behind intermediate metrics, impressions, clicks, engagement rates, reach, that measured activity rather than results. When a tool arrives and automates all that activity, what remains exposed is the question that mattered from the beginning: are you selling more or not?

Research published by Funnel in January 2026 captures this crisis with surgical precision: 72 per cent of in-house marketers and 55 per cent of agency consultants say they cannot turn data into actionable decisions. Eight out of ten lack a clear signal on what is actually working across channels. Marketing in 2026 generates more reports than at any point in its history, yet most professionals cannot answer the most basic question: what should we do next?

This is where provocation becomes diagnosis. Marketing's problem is not technological. It is structural, and if one has the courage to look, it always has been. The industry built an enormous edifice on a premise that AI has just falsified: that intermediation between seller and buyer is inherently a valuable service. It is not. It never was. The value lies in a deep understanding of why people choose what they choose and in the ability to build something that deserves to be chosen. Everything else was friction dressed up as service.

And here the story takes a turn that few in the industry wish to contemplate. Agentic AI, the kind that does not merely answer but acts on behalf of the consumer, represents a power inversion without precedent in the history of commerce. For the first time, the buyer has an agent working exclusively for them, not for the seller. Shopify has already enabled checkout directly within ChatGPT. Google is rolling out agentic capabilities that allow users to book, purchase, and compare without ever landing on a website. OpenAI opened its doors to advertising in February 2026, which means that even the last "neutral" space of AI recommendation is about to be colonised by the same incentives that corrupted traditional search.

The information asymmetry on which marketing has always relied, the fact that the seller knew more than the buyer about prices, alternatives, and relative quality, no longer exists. And when information asymmetry disappears, the entire apparatus built to exploit it becomes dead weight.

There is also the matter of content quality, and here the provocation touches a nerve that is particularly raw. The word "slop" was named a word of the year in 2025 to describe the mediocre AI-generated content flooding the web. But anyone who has worked in the industry long enough knows an uncomfortable truth: eighty per cent of marketing content was already slop before AI existed. It was slop written by humans, billed by the hour, optimised for keyword density instead of intelligence, produced at industrial scale by agencies that invoiced by volume. AI did not create the mediocrity problem in marketing. It made mediocrity free, and when something becomes free it ceases to be a business.

For those working in the sector, the implications are not the ones that conference stages suggest. There is no need for a "pivot to AI". No need for a new framework, a new acronym, a new technology stack. What is needed is the courage to examine your own business model and ask: am I creating real value or am I taxing a friction that is disappearing?

The marketing businesses that survive this transition will not be those that learn to "optimise for AI". They will be those that do something AI cannot do and will not do for a considerable time: understand the deep mechanisms that drive human choice, build narratives that carry genuine meaning, design experiences that people actively seek rather than passively endure. In other words, the competencies that marketing has always claimed to possess but in practice replaced with algorithmic targeting and A/B tests on copy variants that no human being can tell apart.

Twenty-four per cent of British agencies expect to cut staff this year because of AI, according to the IPA's annual census. Omnicom has shed thousands of positions over the past year and is planning hundreds of millions more in "synergies". The sector is contracting. But the contraction is not caused by AI. AI is merely the catalyst that accelerated a process the market's own structure made inevitable. When you build an industry on inefficiency, the arrival of efficiency is not a crisis. It is a correction.

For those with eyes to see, the correction is also an unprecedented opportunity. AI systems reward precisely what good marketing should always have been: dense, genuinely informative, verifiable, useful content. Citations from authoritative sources. Analysis that adds real understanding. The old SEO rewarded those who played an algorithm's rules most cleverly. The new paradigm rewards those who actually know something and communicate it with clarity. For professionals who have always worked this way, who have always found the industry's obsession with vanity metrics and technical tricks frustrating, this moment is not a threat. It is the end of a system that penalised substance and rewarded noise.

The real question, the one that separates those who will adapt from those who will keep telling themselves that "adding AI to the process" is enough, is simple and unsparing: if tomorrow morning your client could ask an AI agent to do everything your agency does, what would be left? If the answer is "nothing", the problem is not AI. The problem is that there was never enough real value behind the invoice.