Search engine optimisation as a standalone discipline is collapsing into a sub-discipline of generative engine optimisation. Gartner forecasts a 25% reduction in traditional search volume by 2026, and AI Overviews has cut click-through rates on informational queries dramatically. Hiring an 'SEO Manager' in 2026 is staffing for a 2019 discovery layer. The work is now about being cited inside AI answers, not ranked on a results page.

Will AI replace marketers or augment them?

The 'augmentation' framing is corporate cope. WPP, Publicis, Omnicom and IPG have collectively removed tens of thousands of roles in 2024 to 2025 against flat-to-declining organic growth. Microsoft's own pilots show 29% faster task completion and roughly nine hours saved per user per month. The honest read is that AI is replacing routine marketing work and concentrating remaining roles around brand, agent platforms, and judgement. Companies that maintain entry-level pipelines through 2028 will hold a senior-talent advantage by 2032.

What is agentic commerce and why should a CMO care?

Agentic commerce is direct purchase delegated to an AI agent. ChatGPT's Instant Checkout went from zero to over one million Shopify merchants in approximately six months. 23% of Americans bought something via an AI agent in the past month, per Morgan Stanley AlphaWise. The strategic risk is that brands without product feeds compatible with the Model Context Protocol or Agentic Commerce Protocol will be excluded from the buying flow before they ever appear in the consideration set.

What is generative engine optimisation (GEO)?

GEO is the discipline of being cited and recommended inside AI answers from ChatGPT, Gemini, Perplexity, Microsoft Copilot and Google AI Overviews. It depends on structured data, schema markup, evidence-backed claims, FAQ formats, and earned third-party citations on Reddit, YouTube, Wikipedia and reputable trade publications. It is the natural successor to SEO and the prerequisite for visibility in the new discovery layer.

What are the biggest risks no one is pricing in?

Three sit in the high-probability, high-impact quadrant. A model-collapse event in which AI-generated content degrades the training corpus and triggers a quality regression. A regulatory shock from the EU AI Act, U.S. state-level rules, or antitrust action against Google. And the platform-tax dynamic in which two or three agent operators capture pricing power over advertisers in the way Google and Meta did at the search-and-social layer. Boards that do not commission scenario planning are allocating capital on the assumption that the next ten years will resemble the last ten. They will not.

What AI Content Scale Does to B2B Trust

By Juan Beltrán · 2026-05-06

The marketing function is being rewritten by a discontinuity, not a trend. Search is bifurcating, an agentic layer is forming, and creative production costs are collapsing while performance media concentrates in two or three platforms. Brands without a machine-readable, agent-addressable presence will become invisible by 2028. The honest framing for the marketing workforce is replacement, not augmentation. Most organisations are quietly betting their P&L on a world that no longer exists.

A discontinuity, not a trend

For two decades, the dominant mechanic of digital marketing was simple. A user typed a query. Ten blue links appeared. The user clicked. The brand captured the visit. Revenue followed. That mechanic, and the entire industry built on top of it, is breaking. Not gradually, not in some forecast horizon, but now, in measurable, audited, balance-sheet-visible numbers.

The distinction between trend and discontinuity is the one that matters for capital allocation. Trends are extrapolations of the existing system. Discontinuities are reconfigurations of it. When ChatGPT reaches 900 million weekly active users in eighteen months, having taken Google fifteen years to reach a comparable scale, and when global publisher traffic from search falls by a third inside a single calendar year, the system itself is being rewritten.

Three movements are happening at the same time. First, the discovery layer of the internet has fragmented across half a dozen AI surfaces in addition to Google. Second, an agentic layer is forming, in which autonomous software acts on behalf of users, researching, comparing and purchasing. Third, inside the marketing function, generative AI is collapsing the cost of creative production while concentrating performance media optimisation in the hands of two or three platforms.

Each movement on its own would justify a strategic response. Together, they amount to a reset. The thesis of this piece is that the marketing function as currently structured will not survive the next thirty-six months, and that most CMOs are quietly betting their P&L on a world that no longer exists.

My read from inside enterprise digital teams is that the danger is not ignorance. Most leaders can recite the AI-search narrative. The danger is planning inertia: budgets, roles, dashboards, and agencies still arranged around the click economy while the buyer is already learning through answer engines and agents.

Claim one. The click economy is dead. Most marketing leaders have not noticed.

Gartner forecasts traditional search engine volume falling 25% by 2026 as AI assistants become substitute answer engines. That figure was filed in February 2024 and now reads as conservative.

Independent studies from Pew Research, Seer Interactive, Digital Content Next, Chartbeat and SimilarWeb show informational click-through rates collapsing where AI Overviews appear, and publisher traffic falling by a third year on year. HuffPost, Business Insider and Forbes report declines greater than fifty per cent over twenty-four months. The numbers are uneven across categories. Retail and product queries are less affected than news and reference. The direction is consistent.

Google's VP of Search Liz Reid publicly disputed the publisher-decline framing in August 2025, arguing that "total organic click volume has been relatively stable." That position is contradicted by every independent measurement on informational verticals. The data is now the data. The contested question is whether to act on it.

The honest read for any CMO whose 2026 plan still assumes that organic search delivers the audience it delivered in 2022 is that the plan is wrong. The supply side has changed. The demand side has changed. The mechanism that translated demand into a session on a brand-owned property has been disintermediated by a conversational layer that increasingly answers the user without ever sending them to a website.

Anyone hiring an SEO manager in 2026 is staffing for 2019. The discipline that survives is generative engine optimisation: being cited and recommended inside the AI answer. The work is structural. Schema markup. Evidence-backed claims. FAQ formats. Earned third-party citations on Reddit, YouTube, Wikipedia and reputable trade publications. The brands that win the next cycle will be the ones whose name is reliably surfaced when a buyer asks "what is the best X."

Claim two. Your website is becoming an API, not a destination.

The most uncomfortable shift for marketing leaders is that the corporate website is no longer the strategic asset it was. By 2028 to 2030, the dominant interface for most consumer digital experiences will be a multimodal AI assistant. Voice. Text. Increasingly camera-based. Websites will persist, but they will narrow into two functions: source-of-truth APIs for agents, and high-fidelity flagship experiences for committed brand fans. The middle tier of mid-traffic informational sites will continue to be hollowed out.

The protocols that define this shift are already deployed. The Model Context Protocol from Anthropic and the Agentic Commerce Protocol from OpenAI and Stripe are the equivalent, in agent-economy terms, of HTTP and HTTPS. Brands that do not expose product, pricing, support and inventory through MCP-compatible servers will be unfindable to the agents that will increasingly do the choosing.

ChatGPT's Instant Checkout went from zero to over one million Shopify merchants in approximately six months. Etsy is integrated. The number of agent-callable surfaces is doubling roughly every two quarters. Conversational commerce reached $290 billion in U.S. consumer spend in 2025, up from approximately $41 billion in 2021. Shoppers who engage with an AI assistant during a session convert at roughly 12.3%, approximately four times the rate of non-AI sessions.

Twenty-three per cent of Americans bought something via an AI agent in the past month. Seventeen per cent of consumers begin holiday shopping on an AI platform. These are early-adoption numbers in late 2025 and early 2026. They are not steady-state.

The strategic test is brutally simple. Is the brand's product catalogue agent-callable? Is its support API documented for agent consumption? Is its checkout addressable through the Agentic Commerce Protocol? A "no" to any of these is an emerging form of digital invisibility. The new front door is the conversation, and the brand is increasingly invited or excluded by the agent rather than chosen by the human.

Claim three. Augmentation is a polite word for replacement.

The narrative most comfortable to executives is that AI augments marketers rather than replaces them. The headcount data from the holding companies says otherwise.

WPP, Publicis, Omnicom and Interpublic Group have collectively removed tens of thousands of roles since 2024 against flat-to-declining organic growth. Taligence's U.S. Marketing Job Postings Index for Q2 2025 shows entry-level marketing postings down sharply year on year. The agency P&L is restructuring around fewer humans producing more output, not around the same humans producing better output.

Microsoft's internal pilots show 29% faster task completion and roughly nine hours saved per user per month, but only with structured training rather than casual exposure. McKinsey's analysis sizes the generative AI productivity opportunity at $2.5 to $4.4 trillion annually, with marketing and sales the single largest functional value pool. That value does not arrive without consequence. It arrives by reducing the labour required to produce a unit of output.

The honest framing for the marketing workforce in the next thirty-six months is replacement at the routine layer, concentration of remaining roles around brand strategy, agent and data platforms, and judgement, and an accelerating "vanishing junior" problem. Companies that quietly stop hiring at entry level are storing up a senior-talent shortage that will hit them in 2030 to 2032. The right move is contrarian. Maintain the entry-level pipeline through the disruption, accept higher unit cost in the short term, and own the senior bench in the next cycle.

The B2B and B2C reset

In both B2B and B2C, the strategic problem can be stated in one sentence. The buyer no longer comes to you first. The buyer comes to an AI surface first, and arrives at the brand only if the AI surface has identified the brand as relevant.

In B2B, eighty-nine per cent of buyers have adopted generative AI in pre-purchase research. Seventy-four per cent self-serve digitally before talking to sales. Ninety-two per cent enter the purchase process with a vendor already in mind. Forty-one per cent enter with a single preferred vendor. The traditional motion of fill-the-funnel, qualify, hand-to-sales must be re-architected around the assumption that the buyer arrives already informed and already partly decided, with their AI assistant having pre-screened the field.

In B2C, AI Overviews and standalone chat products increasingly mediate the top of the funnel for considered purchases. Health. Finance. Travel. Durables. Loyalty fragments under these conditions. When the agent does the choosing, brand loyalty is harder to manufacture but more valuable when it exists. Brand and category leaders with strong "named-in-answer" presence will be disproportionate beneficiaries. Mid-market and unbranded competitors who relied on price comparison and last-click attribution face the steepest erosion.

The implication for marketing investment is direct. Spend that built the buyer's mental model before the AI surface intervened was always the most leveraged dollar. It is now the only one that matters at scale.

What the honest CMO does in the next ninety days

Five actions, all uncontroversial in 2027 and difficult in 2026.

One. Audit AI visibility. Manually test the top fifty buyer queries in ChatGPT, Gemini, Perplexity, Microsoft Copilot and Google AI Overviews on a monthly cadence. Track citation share against named competitors. If AI citation share is below fifteen per cent of category competitors, treat it as urgent.

Two. Stand up a generative engine optimisation function. Either expand the existing search team's mandate or appoint a dedicated AI-visibility lead. Implement structured data, schema markup, FAQ formats and evidence-backed claims. Earn third-party citations on Reddit, YouTube, Wikipedia and reputable trade publications.

Three. Inventory and govern the AI stack. Catalogue every AI use across creative, media, personalisation and customer experience. Implement watermarking and labelling before EU AI Act Article 50 enforcement on 2 August 2026. Maximum fines of €35 million or 7% of global turnover make this a board-level matter, not a marketing-operations one.

Four. Lock in first-party and zero-party data programmes. Audit consent infrastructure. Deploy a consent management platform and Consent Mode v2. Activate preference centres and loyalty programmes as data exchanges. If first-party data covers less than forty per cent of audience reach, treat it as urgent.

Five. Prepare for agentic commerce. If the brand sells physically, ensure product feeds are compatible with the Model Context Protocol or Agentic Commerce Protocol. Launch on the Etsy and Shopify-ChatGPT integrations. The trigger condition is straightforward. If competitors begin appearing in ChatGPT Buy or Google AI Mode purchasing flows and the brand does not, it has already lost a quarter of share that will be expensive to win back.

The risks nobody is pricing in

The base case in this analysis is broadly constructive. A productivity uplift. A re-architected channel mix. A healthy if smaller agency sector. A brand premium for those who navigate the transition well. There are, however, three risks of sufficient probability and impact to merit explicit attention.

First, model collapse. As 35 to 50% of new web content becomes AI-generated by 2027, the training corpus that future models depend on degrades. The risk is a noticeable quality regression in 2027 to 2028 that triggers a flight to verifiable human-authored content. The mitigation is to invest in proprietary data, original research, and provenance standards (C2PA, W3C verifiable credentials) before the price of those assets reprices upward.

Second, platform-tax concentration. The agent layer is on track to mirror the current Google-Meta dynamic in search and social, with two or three operators capturing pricing power over advertisers. The competitive question is which company controls the user agent. OpenAI, Google, Apple and Meta are the four with credible distribution claims. The U.S. v. Google antitrust remedy under Judge Mehta and EU Digital Markets Act enforcement will materially shape the answer. A duopoly or triopoly is the modal outcome. Plan for the cost.

Third, regulatory fragmentation. The EU AI Act, U.S. state-level regulation, and Asia-Pacific approaches are diverging. A global marketing operation cannot assume a single compliance regime. The cost of operating across regimes will rise, and the brands that under-invest in governance now will absorb the rise as enforcement actions in 2027 to 2028.

Boards that take this seriously will commission proper scenario planning rather than relying on point-estimate forecasts. Boards that do not will be allocating capital on the assumption that the next ten years will resemble the last ten. They will not.

Closing

The strategic message for senior leaders is unchanged by the caveats and the contested figures. The direction of travel is clear even if the speed and the destination are not.

Leaders who invest now in AI visibility, agent infrastructure, first-party data, and the human-led brand and trust assets that AI cannot replicate will compound advantage. Those who wait will find, in 2028 or 2029, that the discovery layer, the personalisation layer and the transaction layer have all been rebuilt, and that they are renting access to their own customers from someone else.

This is not a forecast. It is an inference from the evidence already on the table.

The full 33-page working paper, with figures, references and the staged programme of action across three horizons to 2030, is available below.

Research summary

Abstract

This working paper argues that the marketing function is undergoing a discontinuity rather than a trend. Three movements are happening at once: the discovery layer of the internet has fragmented across half a dozen AI surfaces; an agentic layer is forming in which autonomous software researches, compares and purchases on behalf of users; and inside the marketing function, generative AI is collapsing the cost of creative production while concentrating performance optimisation in two or three platforms. Together they amount to a reset. The paper sizes the financial impact, surfaces the structural risks, and proposes a staged programme of action across three horizons to 2030.

Research question

Is the change in search, marketing and digital experience between 2024 and 2026 a trend that can be managed within existing operating models, or a discontinuity that requires those models to be rebuilt before 2028?

Methodology

Synthesis of primary disclosures, analyst research and quasi-peer-reviewed industry data. Sources include Gartner, McKinsey, Forrester, Bain, eMarketer, WARC, Pew Research, Seer Interactive, SimilarWeb, Chartbeat, Stanford, Imperial College London, Internet Archive, holding-company interim results (WPP, Publicis, Omnicom, IPG), platform disclosures from OpenAI, Google, Microsoft and Anthropic, and U.S. Department of Justice antitrust filings. Where projections are speculative they are framed as such. Where sources conflict, both are reported.

Key findings

Gartner forecasts traditional search engine volume falling 25% by 2026 as AI assistants become substitute answer engines.
ChatGPT reached approximately 900 million weekly active users in 18 months. Google took 15 years to reach a comparable scale.
Publisher traffic from organic search fell by approximately one third in a single calendar year. HuffPost, Business Insider and Forbes show declines greater than 50%.
Global advertising spend crossed $1 trillion in 2025. McKinsey models $2.5 to $4.4 trillion in annual generative AI productivity, with marketing and sales the single largest functional value pool.
Conversational commerce reached $290 billion in U.S. consumer spend in 2025, up from approximately $41 billion in 2021. Shoppers who engage with an AI assistant convert at roughly four times the rate of non-AI sessions.
23% of Americans bought something via an AI agent in the past month (Morgan Stanley AlphaWise, 2025). 17% of consumers begin holiday shopping on an AI platform (Bain).
89% of B2B buyers have adopted generative AI in pre-purchase research. 74% self-serve digitally before talking to sales. 92% enter the purchase process with a vendor already in mind.
Holding-company workforces continue to contract. WPP, Publicis, Omnicom and IPG combined have removed tens of thousands of roles in 2024 to 2025 against flat-to-declining organic growth.
By 2027, 35 to 50% of new web content is projected to be AI-generated, raising the value of verifiable human authorship and proprietary research.
EU AI Act Article 50 enforcement begins 2 August 2026. Maximum fines of €35 million or 7% of global turnover make watermarking and labelling a board-level matter.

Executive implications

Treat 2026 as the last full planning year before the agent layer becomes the default discovery surface for considered purchases.
Audit AI visibility on the top 50 buyer queries across ChatGPT, Gemini, Perplexity, Microsoft Copilot and Google AI Overviews on a monthly cadence. Below 15% category citation share is an escalation trigger, not a benchmark.
Stop hiring SEO managers. Stand up a generative engine optimisation function with explicit ownership of structured data, evidence-backed claims, and third-party citations on Reddit, YouTube and Wikipedia.
Reframe the corporate website as a machine-readable API. Stand up an MCP server exposing product, support and pricing data. Brands without an agent-addressable surface will become digitally invisible by 2028.
Shift 5 to 15% of media budget back into brand, creator and earned media. Performance ROAS is being squeezed simultaneously by ad-platform AI tax and agent-mediated discovery.
Treat the 'AI-augmented marketer' framing with scepticism. Holding-company headcount data points to replacement at scale. Maintain entry-level pipelines through 2026 to 2028 to hold a senior-talent advantage by 2032.
Restructure marketing around three pillars by 2027: brand and strategy, agent and data platforms, performance and measurement. The change is meaningfully easier in 2027 than in 2030.

Limitations

Forecast uncertainty is high. Many headline figures (eMarketer's $25.9B AI-search ad forecast for 2029, Bain's $300 to $500B agentic commerce estimate by 2030, McKinsey's $4.4T productivity figure) are reasoned models, not measured outcomes, and should be read as scenario anchors. Numbers conflict across sources: 2025 global advertising spend ranges from $1.03T (eMarketer) to $1.17T (WARC); ChatGPT WAU figures range from 800 million to 900 million depending on source and month. The AI-driven traffic decline narrative is contested: Google's VP of Search Liz Reid in August 2025 publicly disputed publisher-traffic decline framings. Multiple independent studies contradict that position for informational verticals. Reality is likely category-dependent. The agent ecosystem is nascent, and real adoption may lag breathless market sizing. Regulatory trajectories diverge by region. Visionary projections to 2031 extend current trends and could be invalidated by structural breaks: a major model-collapse event, an antitrust break-up of Google, or the displacement of OpenAI by a Chinese frontier model in major markets.

References

Gartner. Forecast: 25% reduction in traditional search engine volume by 2026 as AI assistants substitute. Press release, February 2024.
OpenAI. ChatGPT weekly active user disclosures, 2024 to 2026.
McKinsey & Company. The economic potential of generative AI: the next productivity frontier. 2023, updated 2025.
Pew Research Center. Search engine results in the age of AI. 2025.
Seer Interactive. AI Overviews CTR collapse benchmarks. 2025.
Conductor / NYU AI Research. Generative engine optimisation studies. 2025.
SE Ranking. AI search visibility benchmarks, mid-2025.
Digital Content Next. Publisher traffic erosion analysis. 2025.
Chartbeat and SimilarWeb. Publisher traffic time-series, 2024 to 2025.
Bain & Company. Agentic commerce projections to 2030.
Morgan Stanley AlphaWise. Consumer adoption of AI agents survey. 2025.
Gladly and Rep AI. Conversational commerce sizing, 2021 to 2025.
Forrester. B2B sales forecasts and agent-led negotiation projections. 2025.
Salesforce. State of Sales reports, 6th and 7th editions.
The Pedowitz Group. Revenue Marketing Index. 2025.
eMarketer. Global advertising spend and U.S. AI-search ad spend forecasts, 2025 to 2029.
WARC. Global Advertising Spend Forecast 2025/2026.
WPP plc, Publicis Groupe, Omnicom Group, Interpublic Group. H1 2025 interim results.
Stanford University, Imperial College London, Internet Archive. The Composition of the New Open Web. 2025.
Statista / LXA. Global martech market sizing, 2025.
Taligence. U.S. Marketing Job Postings Index, Q2 2025.
Writesonic. Reddit's 450% growth in AI Overviews. 2025.
Anthropic. Model Context Protocol specification, 2024 to 2026.
OpenAI and Stripe. Agentic Commerce Protocol announcement and documentation, 2025.
United States v. Google LLC, U.S. District Court for the District of Columbia, Judge Mehta. 2024 to 2026.
European Commission. Artificial Intelligence Act, Article 50 enforcement timeline.

Key takeaways

Gartner forecasts a 25% collapse in traditional search by 2026. Publisher traffic from search has already fallen roughly a third in a single year.
ChatGPT reached 900 million weekly active users in 18 months. Google took 15 years to reach comparable scale.
23% of Americans bought something via an AI agent last month. Conversational commerce hit $290B in U.S. spend in 2025.
92% of B2B buyers enter the purchase process with a vendor already in mind. The buyer no longer comes to you first.
Holding companies have cut tens of thousands of marketing roles since 2024. The 'AI-augmented marketer' framing does not match the headcount data.

Decision brief

Generative AI is shifting discovery from a click economy toward an answer economy. Brands will earn visibility when their expertise is explicit, well-evidenced, attributable, and easy for search and answer engines to extract, compare, and cite.

Decision criteria

The page resolves one identifiable executive question directly.
Claims have named sources, dates, limitations, and an accountable author.
The site connects the answer to a coherent entity and topic architecture.

Evidence to check

Non-branded search impressions and qualified visits by topic.
Citations or mentions in AI answers and assisted research journeys.
Engaged reading, newsletter conversion, and downstream business conversations.

What I would do Monday

Take one strategic topic you want to own. Publish a direct answer, your distinctive position, the evidence behind it, and the decision a reader should make next, all on one canonical page.

Questions executives ask

Is SEO dead in 2026?: Search engine optimisation as a standalone discipline is collapsing into a sub-discipline of generative engine optimisation. Gartner forecasts a 25% reduction in traditional search volume by 2026, and AI Overviews has cut click-through rates on informational queries dramatically. Hiring an 'SEO Manager' in 2026 is staffing for a 2019 discovery layer. The work is now about being cited inside AI answers, not ranked on a results page.
Will AI replace marketers or augment them?: The 'augmentation' framing is corporate cope. WPP, Publicis, Omnicom and IPG have collectively removed tens of thousands of roles in 2024 to 2025 against flat-to-declining organic growth. Microsoft's own pilots show 29% faster task completion and roughly nine hours saved per user per month. The honest read is that AI is replacing routine marketing work and concentrating remaining roles around brand, agent platforms, and judgement. Companies that maintain entry-level pipelines through 2028 will hold a senior-talent advantage by 2032.
What is agentic commerce and why should a CMO care?: Agentic commerce is direct purchase delegated to an AI agent. ChatGPT's Instant Checkout went from zero to over one million Shopify merchants in approximately six months. 23% of Americans bought something via an AI agent in the past month, per Morgan Stanley AlphaWise. The strategic risk is that brands without product feeds compatible with the Model Context Protocol or Agentic Commerce Protocol will be excluded from the buying flow before they ever appear in the consideration set.
What is generative engine optimisation (GEO)?: GEO is the discipline of being cited and recommended inside AI answers from ChatGPT, Gemini, Perplexity, Microsoft Copilot and Google AI Overviews. It depends on structured data, schema markup, evidence-backed claims, FAQ formats, and earned third-party citations on Reddit, YouTube, Wikipedia and reputable trade publications. It is the natural successor to SEO and the prerequisite for visibility in the new discovery layer.
What should a CMO do in the next 90 days?: Five things. Audit AI visibility on the top 50 buyer queries monthly across the major AI surfaces. Stand up a GEO function. Catalogue every AI use across creative, media and CX before EU AI Act Article 50 enforcement on 2 August 2026. Audit consent infrastructure and lift first-party data coverage above 40% of audience reach. If you sell physically, ensure product feeds are MCP and ACP compatible.
What are the biggest risks no one is pricing in?: Three sit in the high-probability, high-impact quadrant. A model-collapse event in which AI-generated content degrades the training corpus and triggers a quality regression. A regulatory shock from the EU AI Act, U.S. state-level rules, or antitrust action against Google. And the platform-tax dynamic in which two or three agent operators capture pricing power over advertisers in the way Google and Meta did at the search-and-social layer. Boards that do not commission scenario planning are allocating capital on the assumption that the next ten years will resemble the last ten. They will not.

Canonical URL: https://juanbeltran.ch/blog/the-generative-ai-reset