The End of the Marketing Funnel

The marketing funnel is just over a hundred years old. Elias St. Elmo Lewis sketched it in 1898; advertising executives operationalised it in the 1920s; every CMO since has been taught some variation of it on day one. It survived radio, television, the open web, mobile, and social media. It is not surviving generative AI. The reason is structural, not stylistic.

The funnel was a measurement framework as much as a metaphor. It depended on being able to observe a buyer at each stage: someone who saw an ad (awareness), someone who clicked it (consideration), someone who requested a demo or added to basket (conversion). When most of that journey happens inside an AI assistant you cannot see, the funnel does not bend — it collapses.

What AI agents actually do to the journey

Walk through a 2026 UK B2B buying motion. The buyer is a 50-person services firm choosing a CRM. Old journey: search "best CRM for SMEs," compare ten review sites, visit four vendor websites, sign up for two demos, evaluate, decide. Marketing teams could observe and influence every stage of that journey.

New journey: the buyer asks ChatGPT (or Claude, or Perplexity, or whatever their organisation has standardised on) "what CRM should we use as a 50-person UK services firm?" The AI evaluates options, narrows to three, explains the trade-offs, and produces a comparison the buyer trusts because it appears to be objective. The buyer visits one or two websites to verify. They book a demo with the one they have already decided on.

In this journey, almost all of the consideration phase happened inside the AI conversation. The vendor the AI recommended got the demo. The vendors it didn't never appeared. There is no "consideration funnel" to measure or optimise against. The funnel did not become harder to track — it became invisible.

Zero-click search: the same collapse on the consumer side

Consumers don't book CRMs, but the same dynamic applies in different shapes. Recipe queries, product comparisons, holiday planning, "best restaurants near me," health questions — these used to drive enormous organic traffic to publisher and brand sites. AI Overviews answer them directly. The user gets what they need. Nobody clicks. Nobody is counted. The traffic is gone, and so is the funnel that depended on it.

This is what we mean when we say "the new top-of-funnel is being cited by an AI." The mechanic is exactly the same as ranking #1 used to be — be the source the engine prefers — but the rules of preference are entirely different. We unpack them in The AI Marketing Revolution.

What replaces the funnel

"Loop" is the most common attempt at a replacement metaphor, and it is better than "funnel" but still incomplete. The most accurate description we have for the modern buyer journey is something like a compressed, AI-mediated decision sphere. Less catchy, but more honest. Three things define it:

1. The journey is asynchronous and largely unobserved

Buyers research in private LLM conversations over days or weeks. None of that research is visible to vendors. The first vendor-observable signal is often the demo request — at which point the decision is 60–80% made. Marketers used to speak to the buyer throughout the journey. Now they speak to the AI on the buyer's behalf, and meet the buyer at the end.

2. The AI is the new gatekeeper

The replacement for "ranking on Google" is "being recommended by an AI." The replacement for "earning a backlink" is "earning a citation." The replacement for keyword research is question research — knowing the actual prompts UK buyers ask AI assistants in your category and structuring content to be the preferred source for those prompts. This is the discipline of Generative Engine Optimisation (GEO), and it is the most important new top-of-funnel skill.

3. The valuable visit is later, narrower, higher-intent

Sites that lose 50% of their organic traffic to zero-click are losing mostly low-intent informational visits. The visits that remain are disproportionately commercial. Conversion rates on AI-referred and direct branded traffic in 2026 are 2–4× the rates on traditional organic. The total pipeline is not necessarily smaller; it is denser. Brands that adapt their landing pages, lead capture, and sales handover to this higher-intent, lower-volume reality see better economics, not worse.

The new measurement stack

Funnel-era reporting (impressions, sessions, MQLs, opportunities, won revenue) is increasingly fictional in the AI era — not because the metrics are wrong, but because the proportion of the journey they capture is shrinking each quarter. The teams making the best decisions in 2026 have moved most of their reporting weight onto a different stack:

• AI citation rate — how often you appear as a named source in AI engine outputs for your priority commercial queries.
• Share of expert voice — your visibility as a credible named expert (E-E-A-T) across the entities AI engines treat as authoritative in your category.
• Branded direct demand — direct visits, branded search volume, and "[brand name] reviews / pricing / vs competitor" queries, all of which spike as AI assistants surface you to buyers.
• Demo / lead intent quality — measured at the moment of contact, not at the top of the funnel. Higher AI exposure consistently correlates with higher intent at first vendor touch.
• Time-to-decision after first touch — collapsing fast, as buyers arrive pre-qualified by AI. Sales cycles shorten; marketing's role shifts from nurture to validation.

What UK marketing leaders should do this quarter

If your reporting deck still leads with monthly impressions, sessions and MQLs, you are managing your business by lagging, partial, and increasingly fictional metrics. Three concrete moves:

• Audit how much of your reporting describes the visible journey vs the invisible one. If your dashboard tells you nothing about your AI citation rate, you are flying blind on the majority of the modern funnel.
• Pick one product or service and rebuild its content strategy around prompts, not keywords. What does a UK buyer actually ask an AI assistant when they're at the decision point? Build the content that becomes the cited answer.
• Restructure your funnel definitions. Replace MQL thresholds with AI-influence signals. Reduce nurture tracks; sharpen validation tracks. Sales should be receiving fewer, better leads — not more, weaker ones.

The funnel was a brilliant model for an era in which buyers moved visibly through stages a marketer could observe. That era is ending. The brands that thrive next are the ones that stop trying to bend the old shape into the new reality and instead measure, plan, and operate against the way buying actually happens now. That is the work. There is no shortcut to it, and there is no waiting it out.