March 29, 202611 min read

The AI Dark Funnel: How Buyers Make Decisions Through Neural Networks

What the AI dark funnel is, why traditional analytics can't see decisions made through ChatGPT and Alisa, and how to measure and optimize this hidden customer acquisition channel.

Vladislav Puchkov

Founder of GEO Scout, GEO optimization expert

In 2026, every third buyer makes a decision after consulting a neural network. But this path doesn't show up in your analytics. This is the AI dark funnel — a blind spot where competitors are stealing your customers.

What is the AI dark funnel

AI dark funnel is the part of the customer journey that takes place inside neural networks and is not tracked by standard analytics tools.

A typical scenario:

1. A user opens ChatGPT or Alisa
2. Asks: "Which email marketing service is best for small business in Russia?"
3. AI names 3-5 services with descriptions of their advantages
4. The user remembers the first brand on the list
5. Types its name into Google or the address bar
6. Lands on the website — but analytics records it as a direct or organic visit

The connection between the AI recommendation and the visit is lost. You don't know the customer came because of (or instead of) a neural network recommendation.

The scale of the problem: numbers you can't ignore

The last row is key. Only 10-15% of visits initiated by an AI recommendation arrive with a referrer from the AI service (e.g., chat.openai.com). The remaining 85-90% are masked as direct or organic traffic.

This means: AI's real impact on your sales is 6-10x greater than what analytics shows.

Why traditional analytics is blind to the AI dark funnel

Google Analytics and Yandex.Metrica

Classic analytics systems work with referrers and UTM tags. They answer the question "where did the user come from" but not "why did they come."

When a user gets a recommendation from ChatGPT and then searches for the brand on Google, analytics records an organic visit for a branded query. Technically correct, but essentially misleading. The real decision source is ChatGPT, but this doesn't appear in the data.

CRM systems

CRM records the lead's entry point: a form on the site, a call, a request. But it doesn't know that before filling out the form, the customer spent 20 minutes chatting with Perplexity, comparing services in your niche.

Attribution models

Neither last-click nor multi-touch attribution accounts for AI as a touchpoint. The AI recommendation is step zero of the funnel — it exists before the first measurable touch.

How AI influences decision-making: three models

Model 1: Direct recommendation

The user asks AI directly: "Which CRM should I choose for a real estate agency?" AI names 3-5 options, the first of which receives disproportionate attention. It's analogous to the first position in search results, but without the ability to scroll further.

Model 2: Choice validation

The user is already considering 2-3 options. They ask AI: "Which is better — Bitrix24 or amoCRM for a team of 10?" AI forms an opinion that often becomes decisive. If your brand isn't among the options, you've lost before the comparison even begins.

Model 3: Niche research

The user asks a broad question: "How to automate marketing in 2026?" AI talks about approaches and mentions specific tools along the way. Brands that AI mentions in the context of an expert answer gain trust by association.

All three models share one thing: if your brand isn't in the answer, you're not part of the decision-making process.

Anatomy of the AI dark funnel: the customer journey

Let's examine the real customer journey with and without the dark funnel.

Classic path (visible)

Search query → Click on a link → Website → Lead form → CRM

Every step is tracked. The marketer knows the source, query, and landing page.

Path through the AI dark funnel (invisible)

Question to ChatGPT → AI recommendation → Brand memorized →
→ Direct visit or brand search → Website → Lead form → CRM

In CRM, this lead looks like they "came on their own" or "found us on Google." But the actual decision point was a ChatGPT response that happened minutes or days before the visit.

Even worse: the lost customer

Question to Alisa → AI recommends a competitor → Customer goes to the competitor

You won't even know this customer existed. They won't appear in analytics or CRM. They simply won't come — because AI sent them to another brand.

How to measure the AI dark funnel

Direct measurement is impossible — that's the essence of the "dark" funnel. But indirect methods exist that together paint a reliable picture.

Method 1: AI visibility monitoring

The primary tool. If you know how often AI mentions your brand for target queries, you understand the funnel's scale. GEO Scout monitors brand visibility daily across 9 AI providers: ChatGPT, Claude, DeepSeek, Gemini, Perplexity, Grok, Google AI Mode, Google AI Overview, and Yandex with Alisa.

Key metrics for assessing the dark funnel:

• Mention rate — in how many responses you're mentioned
• Position — how prominently (1st place vs 5th)
• Recommendation rate — whether AI recommends you directly
• Share of Voice — your share among competitors

Method 2: Correlation analysis

Compare AI visibility dynamics with direct and branded traffic dynamics. If mention rate in AI grows and 2-3 weeks later direct traffic increases, the correlation confirms the AI dark funnel's influence.

Method 3: "How did you hear about us" surveys

Add to your website forms and onboarding process: "Where did you first hear about us?" with the option "AI / neural network recommendation." In practice, 10-25% of respondents choose this option when it's available.

Method 4: Branded query analysis

Growth in branded queries on Google and Yandex without corresponding growth in advertising or PR is an indirect sign of the AI dark funnel. A neural network named your brand, and the user googled it.

Who wins in the AI dark funnel

The AI dark funnel creates a "winner takes all" effect. Brands that AI mentions first receive a disproportionate share of attention.

This is harsher than SEO. In search results, there are 10 results on the first page plus a second page. In an AI response — 3-5 brands and no "next page." More on the importance of position in the article why the first position in AI matters more.

Optimization strategy for the AI dark funnel

Step 1: Determine the scale of the problem

Start with an audit: is your brand mentioned in responses from key AI providers for target queries? If not, you're completely invisible in the AI dark funnel. If yes — what's the position and tone?

Step 2: Launch systematic monitoring

Manually checking 30 prompts across 9 providers means 270 queries daily. GEO Scout automates this process and provides daily data for analysis.

Step 3: Optimize content for citation

AI cites expert content with facts, figures, and structure. A detailed guide is in the article GEO optimization for websites. Key actions:

• Add JSON-LD markup to key pages
• Fill content with specific numbers and case studies
• Create FAQ sections with answers to customer questions
• Ensure presence on external platforms

Step 4: Work with different providers

Each AI provider is a separate channel. Alisa relies on Yandex, ChatGPT on Bing, Perplexity searches the web in real time. Optimization for one provider doesn't guarantee visibility in another. More on the differences in the article why visibility differs between AI providers.

Step 5: Use the Command Center for prioritization

Monitoring data alone doesn't tell you what to tackle first. The GEO Scout Command Center analyzes all metrics, competitive gaps, and technical audit results, then generates a prioritized action list — from the most impactful to less urgent. Each action is tied to specific prompts and providers.

Step 6: Close the measurement loop

Track the correlation: AI visibility growth → branded and direct traffic growth → conversion growth. This is the only way to quantitatively assess the AI dark funnel's impact on business results.

AI dark funnel across industries

The AI dark funnel's impact is uneven. In some niches it's already critical, in others it's just building up.

More on GEO for specific niches: GEO for SaaS, GEO for e-commerce, GEO for B2B, GEO for local business.

What happens if you ignore the AI dark funnel

A scenario for a brand that doesn't work on AI visibility:

1. Quarter 1: competitors begin GEO optimization. AI starts mentioning them more often
2. Quarter 2: your Share of Voice drops. Competitors capture a larger share of AI recommendations
3. Quarter 3: direct and branded traffic stagnates, even though SEO positions are stable
4. Quarter 4: the gap becomes critical. AI confidently recommends competitors, while your brand is mentioned "among others"

Companies that attribute stagnation to "the market" or "seasonality" often miss the real cause — the AI dark funnel is redirecting customers to competitors.

Checklist: working with the AI dark funnel

Diagnostics

• Brand visibility checked in responses from ChatGPT, Alisa, Perplexity, DeepSeek
• Mention rate and position assessed for target queries
• Visibility compared with key competitors
• Direct and branded traffic dynamics analyzed over 6 months

Measurement

• Daily AI visibility monitoring set up through GEO Scout
• "How did you hear about us" question added to forms with "AI / neural network" option
• AI referrer tracking configured in analytics
• Correlation analysis launched: AI visibility vs direct traffic

Optimization

• Key pages optimized for AI citation (facts, JSON-LD, FAQ)
• Expert content created for target niche queries
• External presence strengthened (media, review sites, directories)
• Provider-specific work configured (Alisa, ChatGPT, Perplexity)

Management

• Command Center used for GEO task prioritization
• Weekly analysis of AI metrics dynamics
• AI visibility included in marketing reporting
• Quarterly KPIs set for mention rate and Share of Voice