The Dark Funnel: Why Direct Traffic Spikes When AI Search Visibility Grows

By Sean Chaudhary · Published Oct 13, 2026

We get some version of this call every month. A client opens GA4, sees direct traffic up 150% or 200% over the last quarter, and asks the obvious question: where did all these people come from?

So we do the boring work first. We ask about every channel that could send people straight to the site. Direct mail. Email blasts. Radio, TV, billboards. A conference booth, a podcast appearance, a press hit. Most of the time the answer is none of the above. The team didn’t change anything, and the spike is still sitting there in the report.

The explanation, in nearly every case we’ve looked at, is the same. Their brand started showing up in AI search, and AI search is terrible at handing out credit.

What Is the Dark Funnel?

The dark funnel is the part of the buying journey that happens where your analytics can’t see it. The idea has been around in B2B marketing for years: word of mouth, private Slack groups, peer recommendations. AI search pours fuel on it.

It helps to think about it in three layers:

Framework adapted from Gaetano DiNardi’s dark SEO funnel.

The influence happens in step one. Your analytics only sees step three, and it labels it “direct.” The same Search Engine Land article cites Wynter research finding that up to 84% of B2B buyers now use AI for vendor discovery, and 68% start in AI tools before they ever touch Google. If that’s where the shortlist gets made, that’s where the attribution breaks.

Why Direct Traffic Spikes When AI Visibility Grows

Two things are happening at once, and they stack.

GA4 does have an AI channel now, and it’s worth setting up. It only catches visits that arrive with a referrer intact, which is a small slice of the real number.

How We Use PostHog to Pull AI Visits Out of the Dark

PostHog earns its place in our stack for one reason: the dark funnel. If AI referrals landed cleanly in GA4, we wouldn’t need it. They don’t, so we built our own AI traffic tracking in PostHog to recover as much of that lost attribution as we can.

Here’s how it works. A small classifier runs before the first event fires and checks two signals on every visit: the referrer, and the UTM source. ChatGPT, for example, adds utm_source=chatgpt.com to outbound links, so that tag survives even when the referrer gets stripped and GA4 would have called the visit direct. When either signal matches an AI engine (ChatGPT, Perplexity, Gemini, Copilot, Claude, Grok, DeepSeek or Meta AI), the visitor gets stamped with the engine that sent them. We run PostHog behind our own domain so ad blockers don’t drop the data.

That stamp follows the visitor everywhere. We can see which engines send buyers, which pages AI tools cite, and how many AI-referred visitors go on to request a quote. We also record each visitor’s first-touch source on every form fill, so a lead’s real starting point stays on file.

It’s far from perfect. The signals vary wildly: each AI engine, app and browser handles referrers and UTM parameters differently, many send nothing at all, and the platforms change their behavior without notice, so the classifier needs regular updates. And the buyer who reads an AI answer today and types your URL next week leaves no trail for any tool to find. That’s why PostHog is one layer of the measurement stack, sitting next to AI citation tracking, branded search and pipeline. Together they close most of the gap a single analytics tool leaves open.

The Pattern Across Our Clients

One account with a weird month would be easy to wave off. We see the same shape on nearly every client whose AI search visibility is rising: AI citations climb, branded searches climb, and direct traffic climbs right alongside them.

The quality of those visits tells the same story. For the marine audio brand, the engagement rate of direct visitors jumped from 47.2% to 64.9%. That jump comes from pre-sold buyers who already met the brand in an AI answer.

How We Rule Out Every Other Explanation

Calling every direct spike “AI” would be lazy. Before we credit the dark funnel, we work through a checklist with the client:

• Offline and outbound campaigns. Direct mail, email sends, TV, radio, billboards, trade shows, conferences, sponsorships, podcasts and press. If any of them line up with the spike, they get the credit.
• Tracking changes. A new consent banner, a broken UTM, a migration that dropped referrers, or internal traffic that stopped being filtered.
• Bots and spam. Junk traffic shows up as single-page, zero-engagement sessions from odd locations. Real buyers read, scroll and click.
• Landing pages. Typed visits land on the homepage. AI clicks with lost referrers often land deep on guides, service pages and the exact pages AI tools cite.
• Branded search. If direct traffic and branded queries in Search Console rise together, people are looking for you by name.
• AI citations. We line up the direct trend against AI citation and visibility data. When they climb in the same weeks, the story writes itself.

When every other channel is ruled out and direct is still up 2× or 3×, the dark funnel is the explanation left standing.

How to Measure the Dark Funnel

If your best buyers now arrive through a door analytics can’t label, last-click reports will undersell AI SEO every single month. Here’s what we put in front of leadership:

That last one matters most. A traffic chart is an argument. A pipeline number ends it.

What the Dark Funnel Is Worth

Take the 3PL fulfillment provider above. When they started with us, they had a one-page website that had been live for about 20 years. Eighteen months later, they book 15 to 25 qualified meetings a month, with 23 in September alone and meetings up 27% quarter over quarter.

Each new client is worth high six to low seven figures to them. At that value, September’s 23 meetings represent roughly $20 million in sales-opportunity pipeline. A big share of the buyers behind those meetings would show up in GA4 as “direct.”

Their CEO put the patience part well in his review: “SEO is not something you try for 90 days and abandon when the phone does not immediately start ringing.” The dark funnel is why. The first signals show up in AI answers and branded searches, months before anyone can draw a clean line from a click to a deal.

Frequently Asked Questions

What is the dark funnel?

The dark funnel is the part of the buying journey your analytics can’t track: AI answers, peer recommendations, communities and private conversations. In AI search, buyers discover a brand inside ChatGPT, Perplexity or Google’s AI features, verify it with a branded search, then return directly to convert. Analytics usually credits that visit to “direct.”

Why is my direct traffic suddenly spiking in GA4?

First rule out offline campaigns, email sends, tracking changes and bots. If none of those explain it and your AI search visibility is rising, the likely cause is the dark funnel: AI clicks that arrive without a referrer, plus buyers who discovered you in an AI answer and came back by typing your URL.

Does traffic from AI search show up as direct traffic?

Often, yes. Some AI platforms pass a referrer and land in GA4’s AI channel. Many clicks from AI apps, copied links and some AI search features arrive with no referrer and are labeled direct. Buyers who return later by typing your domain are always counted as direct.

What is the B2B dark funnel?

In B2B, buyers research and shortlist vendors long before they fill out a form. More of that research now happens in AI tools. The B2B dark funnel is that invisible research phase, and it’s why high-value B2B leads often show up with no clear source.

How do you prove AI SEO results if the traffic is dark?

Track AI citations and visibility, branded search, direct users and their engagement, then tie it to pipeline: quote requests, meetings booked and a self-reported “How did you hear about us?” field. When those climb together and no other channel explains the lift, the dark funnel is doing the work.

Can PostHog track traffic from AI search?

Partly. A PostHog setup that checks both the referrer and UTM tags can identify AI-referred visits GA4 files as direct, then follow those visitors through to a quote request. Visits from buyers who later type your URL still can’t be tagged, so PostHog works best alongside AI citation tracking, branded search data and pipeline reporting.