# The Funnel Just Got a Middleman By Wes Pomeroy. Published April 26, 2026. https://wespomeroy.com/writing/ai-search-and-the-marketing-funnel/ Pull up Google and ask it something. Read the answer. Then notice that you didn't actually click anything to get it. That's the leading edge of a change that's coming fast — and it's already started. For most of marketing's modern life, the deal was simple: customers searched, they landed on our sites, they explored what we sold, and eventually they bought. Marketers measured the whole path, and our owned pages — landing pages, category pages, product detail pages — were the central means by which we did the job. Now there's a layer beginning to form in the middle. It shows up as Google's AI summary at the top of the search page, as the Apple Mail preview that replaces your carefully written preheader with its own AI summary, as the Gmail thread digest that lets you skip reading a long message, and increasingly as the browsers and personal AI agents that go fetch information on your behalf. They all do the same job: they read content for you, and then they tell you what it said. That layer is the middleman. It sits between every piece of content you produce and the human you produced it for, and it's beginning to eat into the part of the funnel marketers used to own most directly. First, what hasn't changed. The job is the same: people still need to be made aware of your products, given reasons to buy them, and have their concerns addressed. They still spend money, and the outcomes you're measured on — pipeline, revenue, conversion — don't move. What's changing is the path, along with the metrics that have been used to track it. Here's where things actually stand today. AI-referred traffic — people who click through to your site from a ChatGPT or Gemini answer — is still tiny. Conductor's 2026 AEO/GEO benchmarks put it at about 1% of total website visits on average, growing slowly. Organic search from Google still drives roughly a quarter to a third of arrivals for most marketers. So the AI shift isn't a traffic story yet. But it doesn't have to be. AI is already shaping what people see *before* they click. Conductor's same study found that AI Overviews now appear in 25% of Google searches on average, and that share climbs to nearly half (48.7%) in verticals like Health Care. And Pew Research reports that 34% of US adults have used ChatGPT — including 58% of adults under 30. The traffic hasn't shifted dramatically yet. What's shifting is *where customers form their understanding of you in the first place*. The biggest hit lands in the middle of the funnel — where people are figuring out what to buy. "Which carrier has the best family plan," "iPhone 17 vs Galaxy S26," "what's actually included in this tier": those are exactly the queries the middleman is best at answering inside its own summary. An increasing share of those visitors are getting their answer in the AI overview before they ever land on your plan page or PDP — and the share has only been climbing. Top of funnel was always partly off-property — paid awareness, organic, social. Bottom of funnel is mostly safe: pricing, configurators, account stuff, anything with current data the AI can't bluff. But the middle is where good marketing content has historically done most of its work, and the middle is where the middleman is hungriest. There's real good news buried in all this, though. Of the visitors who do show up on your site, a growing share have already done their homework with the AI before arriving, which means they came to verify, configure, or buy rather than to browse. So as traffic volume starts to drop in the affected segments, the quality of who's left tends to go up. HubSpot reports that organic traffic for their customers has dropped 27% year-over-year. Separately, Semrush's June 2025 study of AI search behavior found that AI-referred visitors convert at 4.4x the rate of traditional organic search visitors. Less volume, better intent — that's the direction the path is heading. One thing worth flagging before we get to the playbook: most of these numbers come from tool vendors with a stake in the story being big — Conductor, Semrush, HubSpot. Methodologies are disclosed and the sample sizes are large, but treat the magnitudes as directional, not definitive. The direction is clear; the precise magnitude isn't. So what does a marketer do? ## 1. Get visible inside the middleman, not just on your own site. When someone asks ChatGPT about your category, what does it say? Does your brand come up? Are competitors getting mentioned ahead of you? You need to know the answers, the same way you've always wanted to know where you ranked on Google. A new category of tools has emerged to check exactly this for you (Profound, Otterly, Peec AI, Ahrefs Brand Radar, Semrush AI Toolkit). Pick one, run the questions your customers actually ask, and watch where you show up, where you don't, and who's beating you to it. ## 2. Publish where the AI looks, not just where Google looks. A big share of what the AI quotes back to people doesn't even come from polished marketing content. According to Semrush's June 2025 analysis of 150,000 AI citations, about 40% of the sources AI uses come from Reddit, 26% from Wikipedia, and 23% from YouTube. If the AI is increasingly forming opinions about your category by reading Reddit threads, you'd better have a real, credible presence in those threads — not fake accounts, but actual people from your team having actual conversations in good faith. Same goes for keeping your Wikipedia page accurate, posting on YouTube, and putting your people in the places where your category gets talked about. ## 3. Invest in the channels the middleman doesn't sit in front of. Email, your app, the communities you host, podcasts, your own video. These are places where you and the customer have a direct line, without an AI in between summarizing the conversation. They're the only spaces the middleman doesn't already sit in front of. Expect the budget conversation to start shifting in their direction. ## 4. Track what the AI is saying about you, and whether it's right. This is the one most teams haven't started yet. When the AI describes your products, does it get them right? Wrong pricing, made-up features, comparisons to competitors that aren't even accurate — these happen all the time, and you only find out if you're actually checking. The new job isn't just "are we visible" — it's "are we visible AND accurate." Treat the things AI gets wrong about your brand the way you'd treat a press correction, except now you're doing it at scale. ## 5. Build the new measurement story before your dashboard goes dark. The numbers we've been reporting up the chain (visits, click rates, the last thing a customer touched before buying) are going to start looking bad even when the marketing's working fine. The new dashboard lives in three places. The first is how often you appear in AI answers, which the tools above will give you. The second is bigger, looser business signals — branded search volume (are more people Googling your name?), overall pipeline and revenue trends, and brand awareness studies. These have always mattered; click-tracking just gave us such a clean precision feed that the bigger signals could quietly take a back seat. As precision coverage shrinks, the bigger signals matter more again. The third is on your own site, but with a twist — track how well each visitor converts, not just how many visitors come through. Fewer people showing up, but the ones who do are buying. That's what the new version of success starts to look like. Stand this up while the old numbers still look healthy. The marketers who get caught flat-footed are the ones who wait until the chart points down to start building the new one. ## The job hasn't changed. The road is changing. We've spent two decades getting good at the old version of this — owned pages we controlled, click paths we could see, metrics that traced every dollar to a conversion. That whole system assumed you could watch the customer move through the funnel. The middleman is what's starting to break that assumption, because more and more often, the customer is forming part of their understanding of you in a place we don't have admin access to. The marketers who do well in this next chapter are the ones who get their heads around it before the numbers force them to. Not because the sky is falling — pipeline and revenue still work, and most of today's visitors are still arriving the human way. It's because the instruments we've been using are starting to stop telling us the story we're used to hearing, and the rate of change isn't slowing down. Better to learn the new instruments while there's still room to make mistakes on them. ## Extra reading - [Conductor](https://www.linkedin.com/company/conductor-inc-): [The 2026 AEO / GEO Benchmarks Report](https://www.conductor.com/academy/aeo-geo-benchmarks-report/) — current AI referral traffic share by industry (1.08% average; IT and consumer staples leading) - [Semrush](https://www.linkedin.com/company/semrush): [We Studied the Impact of AI Search on SEO Traffic. Here’s What We Learned.](https://www.semrush.com/blog/ai-search-seo-traffic-study/) — the source of the 4.4x conversion rate finding and the 40% Reddit / 26% Wikipedia / 23% YouTube citation breakdown - [HubSpot](https://www.linkedin.com/company/hubspot): [Introducing HubSpot AEO: The answer to showing up in AI search engines](https://ir.hubspot.com/news-releases/news-release-details/introducing-hubspot-aeo-answer-showing-ai-search-engines) — primary source for the 27% YoY organic traffic decline among HubSpot customers - [LeadWalnut](https://in.linkedin.com/company/leadwalnut): [How AI Search Changes the Marketing Funnel (Not Just Discovery)](https://www.leadwalnut.com/blog/ai-search-changing-marketing-funnel) — the case study where organic traffic dropped 18% and demo requests went up 12%, and the CMO couldn’t tell why - [Search Engine Land](https://www.linkedin.com/company/search-engine-land): [The funnel flip: Why AI forces a bottom-up acquisition strategy](https://searchengineland.com/ai-funnel-bottom-up-acquisition-strategy-474877) — by [Jason Barnard](https://www.linkedin.com/in/jasonmbarnard/), on how AI engines build recommendations bottom-up, not top-down - [Cloudflare](https://www.linkedin.com/company/cloudflare): [Introducing pay per crawl: Enabling content owners to charge AI crawlers for access](https://blog.cloudflare.com/introducing-pay-per-crawl/) — the new infrastructure layer for content owners who want to charge AI crawlers - [Contentful](https://www.linkedin.com/company/contentful): [The funnel is shifting upstream: What AI-driven discovery means for marketing leaders](https://www.contentful.com/blog/funnel-shifting-upstream-ai-driven-discovery-marketing/) — on the “decoupling of visibility and traffic”