The traditional funnel collapsed and nobody told marketing

TL;DR: Five separate research streams published over the last 18 months - from Princeton-adjacent academics to a 846,000-session SERP study to global chatbot adoption data, say the same thing in five different vocabularies: the linear marketing funnel has collapsed into a single AI-mediated moment.

Buyers research, compare, evaluate, and decide *before* they ever touch a brand-owned surface. Search intent, the planning variable every SEO and content team uses, no longer predicts behaviour. The strategic response isn't "more AI content." It's restructuring marketing around the three things that still work in a collapsed funnel: distinctive brand assets, citable substance, and a single source of truth your AI tools can read from.

Walk into any marketing planning session in 2026 and you'll still see a funnel on the whiteboard. Awareness at the top, decision at the bottom, content mapped to each stage, budget allocated against each tier.

It is the most expensive piece of fiction your team is still organising around.

I don't mean the funnel is "evolving" or "becoming more complex." I mean the moments it describes, awareness, consideration, decision, have collapsed into one moment, mediated by an AI surface you do not own and cannot see. By the time a buyer hits your site, the work that funnel was designed to do has already happened, somewhere else, without you.

This isn't a hot take. It's what five separate, well-credentialed research streams published over the last 18 months are saying when you read them together.

The five studies that prove the collapse of the traditional marketing funnel

Discovery has moved

Question searches, the moment a buyer goes from "I have a problem" to "what could solve this", have dropped 20%+ in less than five months of LLM adoption. They didn't disappear. They migrated. They now happen inside a chat window, where you have no presence, no impressions, no analytics.

Comparison has moved

Google's own data, via the largest public SERP study to date, shows users now treat the results page like a Netflix browse, scrolling back and forth, evaluating, reconsidering. Even *branded* searches do this. The brand shortcut, type the name, click the link, done, is gone. AI Overviews force everyone, including loyal customers, back into a comparison frame every time.

Decision has moved

Whatever's left of the consideration stage is happening inside the AI conversation itself, often without a click out at all. ChatGPT will summarise five vendors and recommend one. Perplexity will rank options against your stated criteria. Google's AI Mode will synthesise an answer and cite the sources beneath it - most of which won't get the click.

Intent no longer predicts behaviour

This is the finding that should rattle every content team in the building. The 20-point engagement spread that used to separate informational, navigational, and transactional intent has collapsed to under five points when AI Overviews are present. The variable your entire content strategy is segmented around, search intent, has lost most of its predictive power inside the AI layer.

Take the four shifts together and what you're looking at isn't a faster funnel. It's a collapsed one.

OLD: Awareness → Consideration → Decision → [Brand site]

NEW: [Everything, in 90 seconds, inside an AI surface] → [Brand site, maybe]

What this breaks

A collapsed funnel breaks more than people realise, because everything downstream of "we have stages" is now mis-specified.

It breaks content mapped to funnel stages. The TOFU/MOFU/BOFU planning grid assumes a buyer moves through stages at human speed with multiple touchpoints. They don't. The AI compressed all three into a single retrieval moment. Your "consideration stage" white paper is being read by an LLM, not a human, and either gets cited in the answer or doesn't.

It breaks intent-based keyword strategy. When intent stops predicting engagement, segmenting your content calendar by intent stops being a planning lever. You're optimising for a variable that no longer correlates with the outcome.

It breaks lead nurture sequences premised on multiple touches. The "average B2B buyer takes 7-13 touches" stat was already a survivorship-biased number from the pre-AI funnel. In a collapsed funnel, most of the touches happen before the buyer is identifiable - inside an AI conversation you can't track, attribute, or sequence into.

It breaks attribution. First-touch and last-touch models both assume the brand was present at the touches. If the dominant touch is an AI summary that mentions you but doesn't link out, neither model captures it. You will see this in your dashboards as "direct traffic growing, paid efficiency falling, nobody knows why."

It breaks the implicit assumption that your brand gets to participate in the consideration process at all. Increasingly it doesn't. The AI participates on your behalf, using whatever it can retrieve about you. If that retrieval is thin, generic, or contradicted across sources, the AI's version of your brand is what the buyer hears - and you never get to correct it.

This is where rising CAC is actually coming from. Not "ads got more expensive." The journey your budget was allocated against stopped existing as a sequence of separable moments.

What still works

A collapsed funnel doesn't mean everything stops working. It means a smaller set of things start mattering more. Three of them, specifically.

1. Distinctive brand assets

When AI Overviews force comparison behaviour onto every result, generic positioning gets evaluated and discarded inside the AI's comparison layer. Sharp positioning survives it.

This is the inverse of what most marketers assume. The conventional wisdom is that AI search rewards content depth and breadth. What the Growth Memo data actually shows is that AI search rewards *distinctiveness* - the ability to be recognisably different from the surrounding options in the 21 seconds of active comparison the user is now spending on the SERP.

"We help companies grow with data-driven marketing" is invisible inside an AI comparison. It evaluates to noise. "We rebuild marketing operations for B2B founders who want to fire their agency" is legible. The AI can place it. So can the human reading the AI's summary.

Brand has never mattered more in AI search. Not as logo and colour palette - as a positioning sharp enough to survive an LLM trying to summarise it in one line.

2. Citable substance

AI models surface content that contains specific, verifiable, sourced claims. The peer-reviewed GEO research from Princeton, Georgia Tech, and the Allen Institute quantified this: adding statistics produces a 40% visibility uplift inside generative engines. Quotations, 37%. Credible source citations, 30%.

Generic content, for example, the "5 tips" listicle, the unsourced advice blog, the trend piece with no numbers, is now economically negative. Not low-ROI. Negative. It can't be retrieved into AI answers, so it produces no AI-layer visibility. It also can't outrank competitors in traditional search anymore, because the Google AI Overview is eating the click-through. It costs money to produce and returns nothing in either direction.

The content engines still producing this stuff at volume are torching budget without realising it. The ones replacing it with citable, statistically-grounded, source-attributed content are showing up in answers the volume-merchants can't access.

I wrote a separate piece on the GEO research and packaged the three tactics into a free skill. Read the GEO tactics breakdown here.

3. A single source of truth your AI tools can read

Every AI surface, yours and theirs, is grounded on what it can retrieve. If your foundational brand truth lives in five locations and seven versions, every AI representation of your brand is a degraded copy of a degraded copy.

The work isn't more content or even better content - it's giving every AI tool, every freelancer, every agency, and every internal writer the same foundational source to read from. ICP, value prop, positioning, tone, content samples, competitor frame. One canonical version. Versioned. Diffable. Loaded into every AI tool as project context.

I've written about this at length as the Marketing Shared Brain. The short version: in a world where AI mediates every brand touchpoint, the single highest-leverage marketing operation you can run is making sure every AI in your stack, and every AI outside your stack that's trained on your public content, sees a consistent, sharp version of who you are.

The strategic move

Stop allocating budget against journey stages that no longer exist as separate moments. Reallocate against the three things AI-mediated discovery actually rewards:

- Distinctiveness. Cut the campaigns that sound like every competitor's campaign. Spend the saved budget on positioning sharp enough to be summarised in one line by an LLM and still be recognisably you.

- Citability. Audit every piece of content against the +40/+37/+30 GEO tactics. Cut what fails. Rebuild what survives with specific statistics, named quotes, and credible source citations.

- Source-of-truth infrastructure. Build the shared brain. Get every AI tool, freelancer, and agency reading from the same foundational files. This is unglamorous, unbillable-looking work that compounds across every dollar you spend on content thereafter.

If you do nothing else this quarter, do those three.

The honest close

The funnel was a useful fiction for a long time. It mapped reasonably well to a world where a buyer Googled a problem, read three articles, downloaded a guide, took a sales call, and bought. That world is gone. Not "transforming." Gone.

What replaces it isn't another diagram. It's a different operating model - one where marketing's job is to make sure the AI layer that now sits between your brand and your buyer has a sharp, citable, internally consistent picture of who you are and why you matter.

If you're a CMO or a founder looking at rising CAC and a content engine that produces less pipeline than it did 18 months ago, the diagnosis is probably not that you need more content or faster ads. It's that the strategy is still solving for a journey that doesn't happen anymore.

Sources

1. Growth Memo — [*Users Behave Differently in AI Overviews*](https://www.growth-memo.com/p/users-behave-differently-in-ai-overviews)

2. London Business School — [*How Gen AI is Changing Online Consumer Behaviour*](https://www.london.edu/think/how-gen-ai-is-changing-online-consumer-behaviour) (Lambrecht, Padilla et al., Feb 2026)

3. Call Your Girlfriend — [*AI Chatbot Statistics*](https://www.callyourgirlfriend.com/blog/ai-chatbot-statistics)

4. FT Strategies — [*How Generative AI is Influencing Your Customers' Behaviour*](https://www.ftstrategies.com/en-gb/insights/how-generative-ai-is-influencing-your-customers-behaviour)

5. Adobe — [*Digital Trends Consumer Report*](https://business.adobe.com/uk/resources/digital-trends-consumer-report.html)