The attention economy that built Google, Meta, and every advertising business in the world for 30 years is being dismantled in real time. When AI agents do the shopping, the browsing, and the decision-making on behalf of humans — as they increasingly do — the entire infrastructure of digital advertising becomes irrelevant. You cannot serve an impression to a machine that doesn't have eyeballs. You cannot capture attention that was never given to you. The next economy of marketing is not about reaching people. It is about being chosen by the systems that reach people on their behalf.
Not investment advice. This letter draws on McKinsey AI in Advertising Survey (February 2026), Salesforce State of Marketing 2026, Gartner CMO Spend Survey 2026, HubSpot AI Trends 2026, IAB Digital Advertising Revenue Report 2025, and eMarketer 2026 forecasts. All figures current as of June 2026.
The digital advertising economy was built on a single foundational assumption: that human attention is the scarce resource, and that the company which can best capture and direct it wins. Google captured attention at the moment of intent — the search box. Meta captured attention at the moment of social connection — the feed. Amazon captured attention at the moment of purchase consideration — the product page. Together they built a $258.6 billion annual industry on the premise that humans browse, humans scroll, humans click — and that the economic value of those actions could be reliably measured, auctioned, and monetised at scale.
This assumption is structurally correct for a world where humans do their own browsing. It is structurally incorrect for a world where AI agents do the browsing on their behalf. When you ask Claude to find you the best noise-cancelling headphones under $300, Claude does not click ads. When you ask ChatGPT to book a restaurant for Friday, ChatGPT does not see the restaurant's Google Ads campaign. When Perplexity answers your question about which laptop to buy, it does not register an impression for any of the laptops it mentions. The trillion-dollar machinery of digital advertising is optimised for a human attention that is increasingly being mediated, filtered, and in many cases entirely replaced by AI systems that were not designed to be advertised to.
What is being replaced · What is replacing it · As of June 2026
For 25 years, the most important real estate on the internet was the first page of Google search results. The entire $80 billion SEO industry — agencies, tools, content factories, link-building services — existed to earn positions on that page. The mechanism was straightforward: Google crawls the web, ranks pages by authority and relevance, and humans browse the results. Your page ranking determined whether you existed commercially on the internet.
AI search is dismantling this mechanism at speed. When a human asks ChatGPT, Claude, Gemini, or Perplexity a question, they receive an answer — not a list of ten blue links. The AI has already done the browsing, synthesis, and recommendation. The human sees the conclusion, not the journey. Over 90% of pages cited in AI Overviews already contain AI-generated content, and Google AI Overviews now appear in nearly one in five US search results — up from one in thirteen just eighteen months ago. The implication is profound: the question is no longer how to rank in Google. It is how to be cited, recommended, and surfaced by the AI systems that are increasingly replacing the search engine as the first point of contact between consumers and information.
This new discipline — Generative Engine Optimisation, or GEO — is in its earliest stage. The practitioners who dominate it in the next three years will hold the same structural advantage that first-mover SEO agencies held in 2002. The difference is that GEO is not about accumulating backlinks and keyword density. It is about building the kind of authoritative, well-sourced, structured, trustworthy content that AI models preferentially cite. It is, paradoxically, a return to actually being good rather than merely appearing good. AI models are harder to manipulate than Google's algorithm ever was.
The most underappreciated structural shift in the entire advertising economy is what happens when AI agents begin doing the purchasing on behalf of humans at scale. This is not a distant future — it is beginning now. Klarna's AI assistant already handles 67% of customer service chats. Bank of America's Erica has handled over 1.5 billion client interactions. AI agents are booking travel, comparing insurance policies, selecting software vendors, and in some contexts making B2B procurement decisions with budget authority.
When the decision-maker is an AI agent rather than a human, the entire playbook of persuasion — beautiful creative, emotional storytelling, social proof, scarcity messaging, celebrity endorsement — becomes irrelevant. AI agents do not respond to fear of missing out. They do not feel aspirational about a lifestyle brand. They do not reward packaging aesthetics. They optimise for the parameters their human principals have set: price, quality score, delivery time, review credibility, return policy, carbon footprint. The brand that wins in the agentic commerce economy is the brand whose structured data, API integrations, review profiles, and price competitiveness are best optimised for machine evaluation — not the brand with the most emotionally resonant television commercial.
This requires a complete reorientation of what marketing departments actually do. The creative director whose job was to conceive a campaign that moved humans emotionally is not equipped for a world where the primary audience for marketing collateral is a large language model parsing structured product data. The media buyer whose expertise was targeting Facebook audiences is not equipped for a world where the ad unit is a citation in an AI-generated answer. The SEO specialist who built authority through backlink accumulation is not equipped for a world where AI models weight content quality over link graph position. The marketing profession is experiencing the most significant structural disruption since the invention of television advertising in the 1950s — and it is happening in under five years.
The surface-level story about AI and marketing creativity is that AI is replacing human creative workers. This is partially true and mostly misleading. It is true that 23% of agencies reduced junior copywriting headcount in 2025, with 31% planning further cuts in 2026. It is true that 94% of marketers plan to use AI for content creation in 2026 — up from 35% two years ago. It is true that the cost of producing a piece of content has collapsed by 75–85% at the volume end of the market.
But the full picture is more complex and more interesting. As AI floods every channel with algorithmically competent content, the scarcity — and therefore the value — of genuinely original human creativity is increasing, not decreasing. Consumers are already developing finely tuned sensitivity to AI-generated content, rejecting its uncanny smoothness in favour of the rough edges of authentic human expression. The brands running Super Bowl ads made with generative AI — Svedka's 2026 campaign was the first notable example — are discovering that audiences can tell, and that the reaction is not enthusiasm but discomfort.
The creative hierarchy is bifurcating sharply. At the bottom: high-volume, low-margin, algorithmic content production — entirely replaced by AI, and the agencies that built business models around it are already experiencing revenue compression. At the top: strategic creativity, brand narrative, cultural insight, and the kind of originality that creates genuine cultural moments — more valuable than ever, commanded by the small number of humans who can genuinely deliver it. The middle — competent but not exceptional, experienced but not visionary — is being hollowed out fastest. Senior strategist demand is climbing at exactly the same agencies reducing junior headcount.
The death of third-party cookies — accelerated by Apple's App Tracking Transparency framework, GDPR enforcement across Europe, and Google's eventual deprecation — has removed the tracking infrastructure that underpinned two decades of digital targeting. The advertiser who could previously follow a user across 500 websites, building a precise behavioural profile, and serve them a contextually relevant ad at exactly the right moment can no longer do so. The re-targeting that made digital advertising so measurably effective has been severely degraded.
What replaces it is not better third-party data — it is first-party data from direct customer relationships, contextual advertising that targets content rather than individuals, and the data-rich environments of walled gardens (Google, Meta, Amazon) where first-party data remains abundant because users are logged in. The structural consequence: advertising spend is concentrating in fewer, larger platforms at the expense of the open web. McKinsey's February 2026 survey confirms that roughly 40% of reallocated AI-driven spend is shifting away from traditional search and the open web toward platforms with stronger data, measurement, and transaction capabilities. The ad tech ecosystem that served the open web — DSPs, SSPs, DMPs, ad exchanges — is facing an existential reckoning.
Generative Engine Optimisation is not a technology problem. It is a credibility problem. AI models recommend sources, brands, and products based on the quality and trustworthiness of the information available about them in the AI's training data and retrieved context. A brand with strong Wikipedia presence, rigorous press coverage, consistent factual accuracy in its own published content, high-quality structured data on its product pages, and authentic reviews across multiple platforms will be recommended more frequently and more prominently by AI models than an equivalent brand with superior traditional SEO but weaker underlying authority.
The practical implications are specific. Every brand needs a structured data strategy that makes its product attributes, pricing, and logistics parameters machine-readable. Every brand needs a press and editorial strategy that generates accurate, substantive coverage in sources that AI models weight as authoritative. Every brand needs a review strategy that builds genuine customer advocacy rather than manufactured social proof. And every brand needs to accept that the question is no longer "how do we appear in search" but "how do we appear in the recommendation layer of an AI system that a consumer is using as a proxy for their own research." The answer to that question is not a technical shortcut. It is being genuinely good and making that goodness legible to machines.
The disruption of the attention economy does not destroy value — it redistributes it. The $258.6 billion annual digital advertising market is not going to zero. It is being restructured around new competitive advantages, new platforms, and new forms of consumer access. Understanding where value is moving is the core investment question.
Google, Meta, Amazon, and TikTok hold massive first-party data advantages that become more valuable as the open web loses its tracking infrastructure. Their ad ecosystems benefit from the migration of spend away from the open web. Meta's AI-driven feed (50%+ algorithmic) and TikTok's (95%+) already demonstrate the model: data-rich, AI-optimised, attention-maximising environments that deliver measurable performance advertising results that the open web cannot match.
As AI floods every channel with algorithmically competent content, human creators with genuine audiences and authentic relationships become scarce and therefore premium. The creator with 200,000 genuinely engaged followers is worth more than ever — not for reach, but for trust. In a world where AI-generated content is everywhere, the signal of human authenticity becomes a differentiating luxury that brands pay significant premiums to borrow.
The tools, platforms, and services enabling AI-powered marketing are the biggest structural winners. The AI marketing market reaches $107.5 billion by 2028 at 36.6% CAGR. Enterprise marketing teams spending $24,000–$48,000 per month on AI-specific tools — a line item that didn't exist two years ago. The picks-and-shovels play in the marketing AI gold rush.
Companies with deep, consent-based customer relationships — their own apps, loyalty programmes, subscription bases, and direct purchase history — hold the only targeting data that the post-cookie world validates. The brand that knows its customers is structurally advantaged over the brand that rented knowledge of its customers from third-party data brokers. This is why Amazon's advertising business has become one of the most valuable in the world — it has first-party purchase intent data that no other company can match.
DSPs, SSPs, DMPs, ad exchanges, and the hundreds of companies that built businesses serving advertising on the open web are facing structural compression. 40% of spend moving toward walled gardens represents tens of billions of dollars leaving the open web ecosystem. The intermediaries who extract margin from that spend go with it. The trade desk is already responding — but the structural headwind is significant and long-duration.
The agency model built on creative production, media buying commissions, and campaign management fees is under existential pressure from three directions simultaneously: AI collapses the cost of content production, eliminating the creative production margin; direct buying bypasses agencies entirely; and the skills required for GEO, agentic marketing, and AI-optimised advertising are not in existing agency talent pools. The agencies that survive will be fundamentally different organisations.
India's advertising market is growing at 12–15% annually — the fastest among any major economy — driven by the same smartphone penetration and cheap data (₹10/GB) that Letters 62 and 63 documented in the education context. The Indian advertising market crossed $12 billion in 2025 and is projected to reach $18 billion by 2028. But the structural disruptions happening in global advertising are arriving in India simultaneously, not sequentially — creating a market where old and new models compete directly without the legacy infrastructure buffer that Western markets have.
India's vernacular internet is the critical variable. AI models trained primarily on English-language content have weaker capabilities in Hindi, Tamil, Telugu, Kannada, Bengali, and the other languages in which the next 500 million Indian internet users will primarily operate. The brands and platforms that crack vernacular AI — that build the GEO infrastructure for regional language content, that develop AI marketing tools that work in Indian languages with Indian cultural context — will own the most important untapped advertising market of the next decade. The Jio-Google-Meta competition for India's next 500 million is a competition for vernacular AI reach. Advertising follows attention. Attention follows language.
The most honest thing to say about the future of advertising in the AI world is that nobody knows exactly what the equilibrium looks like — including the people building the AI systems. The shift from human-browsed to AI-mediated discovery is real and accelerating. But the rate, the specific mechanisms, and the ultimate competitive structure are genuinely uncertain in ways that the confident predictions of marketing consultancies consistently understate.
The second honest observation is that advertising has survived every prior prediction of its death with remarkable resilience. Television was going to kill newspapers. The internet was going to kill television. Social media was going to kill traditional digital advertising. DVRs were going to kill TV commercials. In each case, the old medium didn't die — it adapted, often discovering new forms and new audiences while ceding some territory to the challenger. It is possible that human attention — the raw material of the attention economy — remains more important than the AI transition narrative suggests, because humans remain the ultimate decision-makers for the choices that matter most.
The third honest observation is that the structural shift creates genuine investment opportunities regardless of the uncertainty. Whether the AI transition is 60% complete in five years or 30%, the direction is clear. The companies building AI marketing infrastructure, the platforms with first-party data moats, the creators with authentic trust-based audiences, and the brands investing in GEO and structured data now will outperform those that do not — not because AI replaces everything, but because AI gives the prepared a durable advantage over the unprepared that compounds over time.
Long-horizon thinking on capital, technology, and the forces shaping the next decade of wealth creation. Written from first principles. Not consensus. Not noise.