For twenty years, B2B marketing doctrine has been the same: win the middle of the funnel. Get into the shortlist. Nurture the consideration set. Own the moment when a prospect moves from "aware of five vendors" to "seriously evaluating two."
Entire departments are built around this job. Content marketers write comparison guides. Product marketers craft battlecards. Demand gen teams build nurture tracks, score engagement, grade leads. The middle of the funnel is where preference is earned, where brands actually sell, and where the modern marketing stack was designed to live.
That middle of the funnel just left the building.
The part most people are missing
According to 6sense's 2025 Buyer Experience Report, based on a survey of roughly 4,000 B2B buyers, 94% now use LLMs during their buying process. Forrester had LLM adoption at 89% by mid-2024. This is not a trend. It is the median experience.
The intuitive reaction is that ChatGPT is killing the top of the funnel: ads not clicking, organic traffic down, people doing their research in an AI instead of on Google. That is happening, but it is not the real story.
The same 6sense research shows that B2B buyers still average 16 interactions with the vendor they eventually choose. That number is statistically identical to 2023. Before LLMs exploded. Before any of this.
And here is the counterintuitive part. Buyers are engaging sales three and a half weeks **earlier** than they did a year ago: 26.4 weeks into their journey versus 30 previously. They are not hiding from sales. They are rushing toward sales, because they need humans to validate what they already learned from a machine.
Same number of touchpoints. Sales engaged earlier. Top-of-funnel traffic softer. What could possibly fit all three?
The middle of the funnel collapsed.
What LLMs actually replaced
Think about what the middle of the B2B funnel used to be. A prospect visited three vendor websites. Downloaded two whitepapers. Read a dozen blog posts. Watched a product demo video. Scanned G2 reviews. Joined a nurture track for six weeks. Compared pricing pages. Maybe attended a webinar. Somewhere in all that, they moved from "aware" to "sold on one of you."
That used to take weeks. It produced your best data. It was the stage you instrumented most carefully, because it was the stage you won or lost.
Now, the same work takes twenty minutes. Here are the kinds of prompts your buyers are typing into ChatGPT right now:
Your buyer does not read your whitepaper. ChatGPT reads it for them, extracts the relevant bits, compares it to two competitors, and hands back a summary. Your six-week nurture track gets compressed into a single conversation that lasts less time than it took you to write the first email in the sequence.
The 6sense research confirms this: LLM use peaks in the middle of the buying journey, when buyers are synthesizing, comparing, and building shortlists. Not at the start. Not at the end. The middle.
This is the Silent Middle. The stage where preference is still being formed, but where you have zero visibility into what is happening, what your competitors are being told about you, or which buyers you are losing before they ever knew your name.
Why your dashboard will not catch this
Your top-of-funnel metrics are tracking. Your bottom-of-funnel metrics are tracking. The middle metrics are the ones that got weird.
Pipeline quality looks higher. That is because the buyers who used to enter the middle of the funnel and fall out during comparison are now falling out inside an LLM conversation you never saw. The filter did not go away. It moved upstream.
First-touch attribution is getting unreliable. A prospect's first touch in your CRM is often their last touch in reality, after a dozen AI-mediated interactions you had nothing to do with.
Direct traffic is rising for reasons no one can explain. Some of it is buyers who got your URL from a chatbot and pasted it into a browser. No referrer. No UTM. No trail.
None of this is a measurement problem you can fix by upgrading your analytics. It is a visibility problem. Entire stages of the buying journey stopped producing data, and no amount of attribution modeling recovers data that was never generated in the first place.
What determines whether the AI recommends you
When a prospect asks ChatGPT "what is the best Google Ads agency for B2B SaaS," the answer does not come from your SEO ranking. It comes from a model trained on whatever the internet has said about you.
The inputs that matter are different from the ones your team has been optimizing:
Third-party authority. G2, Clutch, Reddit, Quora, industry publications that mention you by name. LLMs pull from these sources when forming a view of your brand. If you do not exist in these places, you do not exist in the AI's world.
Structured, citable content. LLMs favor content that states things clearly. Definitive claims. Named frameworks. Specific numbers. If your content hedges with "it depends" and "there are many factors," the AI will quote someone else who gave a straight answer.
Entity clarity. The AI needs to understand that your company, your founder, your product, and your content are the same entity. A clean Wikipedia page, consistent profiles, and schema markup on your site make that easier. Fragmented brand presence across platforms makes it harder.
Recency. LLMs with live web access pull recent content. If your last real blog post is from 2024, you are losing to competitors who published last week.
Your whitepapers are training data now, whether you like it or not. So is every blog post, every product page, every press release, every comment you ever left on a Reddit thread about your category.
What this actually means for your budget
The demand gen function was built to own a stage of the funnel that has largely moved into a room you cannot enter. That is a strategic problem, not a tactical one. Buying more mid-funnel content or running another nurture experiment is like adding lanes to a highway losing traffic to a route nobody told you about.
Top-of-funnel spending on awareness is probably performing better than your dashboard can prove. That brand campaign you cannot attribute may be the reason an LLM knows your name when a buyer asks. You will never prove it with last-click.
Mid-funnel content still needs to exist, but its job has changed. It used to be read by humans scoring you against alternatives. Now it also needs to be crawlable, structured, and quotable enough that an AI can summarize it accurately when asked. The audience is half human, half machine, and the machine is the one making the shortlist.
And the biggest shift: some of the budget that used to fund the middle of the funnel belongs in places that do not show up in any ad platform. Building G2 presence. Earning mentions in industry publications. Creating content that AI models will cite. Maintaining a Wikipedia page. Clean schema markup. None of this fits in a Google Ads dashboard. All of it determines whether you show up in the Silent Middle.
The uncomfortable part
Most B2B marketing teams are not structured for this, and most are not even aware the ground has shifted. Your SEO team optimizes for Google's crawler. Your content team writes for humans who browse. Your analytics team measures clicks. Meanwhile, the stage where you used to earn preference is now a closed-door conversation between your buyer and a machine, and the machine is choosing favorites based on inputs nobody on your team is managing.
The brands that figure this out first will not just win more deals. They will win deals that their competitors never even knew were in play, from buyers who made a decision inside a conversation nobody in the industry was in the room for.
The middle of the funnel did not disappear. It went silent. And if you are still running the same demand gen playbook you ran in 2023, you are spending real money to show up in a room nobody's using.
The question is no longer how do we optimize the middle of the funnel. The question is how do we get quoted in the conversation that replaced it.