Your buyer is researching you right now. You cannot see them. They are not on your website. They are not in your CRM. They have not downloaded your pricing PDF or triggered any of your MQL rules. They are asking ChatGPT, Perplexity, or Claude about vendors in your category. And based on what those models say, they are building a shortlist. Then they are going direct to the two or three names that made it.
This is the AI dark funnel. It is new in 2026. It is not visible to Bombora, 6sense, or any traditional intent data tool. It is also where 94 percent of B2B purchase journeys now touch, according to 6sense's 2025 research. If you do not have a presence inside AI search, you are effectively invisible to a huge and growing slice of your market.
This is not a hot take. It is the single biggest shift in B2B buyer behaviour since content marketing went mainstream in 2010. And most marketing leaders are still running playbooks built for a world where buyers started on Google.
What is the AI dark funnel and how is it different from the traditional dark funnel?
Start with the distinction, because it matters.
The traditional dark funnel is anonymous buyer activity that happens outside your tracked channels. Slack communities where your buyers compare notes. Word of mouth between two heads of revenue over a coffee. Group chats full of operators sharing "who's actually good at this." Podcasts on the commute. A conversation at a conference. We wrote the broader case for this in our earlier piece on the B2B dark funnel, and the concept is not new. Chris Walker popularised it around 2021. It has been reshaping demand generation thinking ever since.
The AI dark funnel is something genuinely new. It is anonymous buyer activity that happens inside large language models. ChatGPT. Perplexity. Claude. Google AI Overviews. Gemini. Copilot. When a Director of Revenue Operations at a Series B SaaS company sits down to scope a new tool, they do not always start in Google any more. They open ChatGPT, type "best B2B intent data tools for mid-market SaaS," and read the answer. That conversation happens at their desk, on their phone, in a browser tab. It leaves no trace your marketing stack can pick up.
The useful way to think about it is this. The traditional dark funnel is the bit of the buying journey that happens between humans, off-platform. The AI dark funnel is the bit that happens between a human and a machine. Both are anonymous. Both shape shortlists. But the tools and tactics for winning in each are different.
Why this deserves its own category
Intent data platforms were built for a specific shape of the web. IP addresses, cookies, bid stream data, content syndication, B2B topic surges. The whole stack assumes the buyer interacts with a piece of content somewhere on the open internet, leaving a signal that a vendor can buy back. That model has been straining for years. GDPR, third-party cookie deprecation, and the steady growth of private channels have all nibbled at its edges. The AI dark funnel is the bigger problem. When a buyer's research happens inside a chat interface, none of the existing signals fire. There is no IP activity, no content interaction, no surge topic, no form fill.
We estimate that somewhere between 30 and 50 percent of early-stage vendor discovery for B2B software now begins inside an LLM. The 6sense number of 94 percent includes any touch at any stage. Our own client data suggests first-touch AI discovery is growing by roughly ten percentage points a quarter. It is moving fast.
How do B2B buyers actually use AI in the buying process?
Plenty of noise gets written about this. Here is what we see when we interview our clients' buyers, pulled from 44 discovery calls we have run in the first quarter of 2026.
Vendor shortlisting. The single most common use. "What are the best B2B intent data tools in 2026?" "Who are the leading ABM platforms for mid-market?" The buyer wants a named list. They get one. They save the names in Notes or a shared doc, and this becomes the first version of their shortlist.
Feature comparison. Once they have three or four candidates, they ask targeted questions. "Does Apollo do intent data at enterprise tier?" "Can Clay do account enrichment at scale without breaking?" These queries pull the buyer out of vendor websites and into a neutral place where they can compare claims against one another.
Pricing discovery. Pricing has been the worst-kept secret in B2B for a decade because vendors hide it behind "Request a demo" buttons. Buyers have worked around this using G2, Reddit and LinkedIn. Now they ask ChatGPT directly. "How much does 6sense cost per year for a mid-market company?" The answer may not be exact, but it is close enough to inform whether the buyer books a call or moves on.
Trust validation. "Is [vendor] any good?" is now one of the most common Perplexity queries for B2B categories. The model pulls from G2 reviews, Reddit threads, LinkedIn posts and review sites, then synthesises a verdict. A single strong Reddit thread from 2024 can dramatically swing how a vendor gets characterised in 2026.
Use case validation. Before a buyer even looks for vendors, they often ask whether the category can solve their problem at all. "Can a sales engagement platform help with multi-threaded account selling?" This is upstream of vendor discovery and is arguably the highest value queries to be cited in.
Category education. "What is signal-based selling?" "What is the AI dark funnel?" When buyers are early-stage, they use AI to get a fast briefing. If you wrote the category-defining content, you get cited. If you did not, you are absent from the conversation before it even turns to vendors.
The pattern across all six use cases is the same. By the time a buyer interacts with your website or your sales team, the frame has already been set by an AI conversation you were not part of. Lenny Rachitsky's 2026 research on buyer behaviour lines up with this, as do the benchmarks from Averi and SEMrush's AI citation studies.
What does ChatGPT actually say about you?
The mechanics are straightforward and worth understanding. When ChatGPT, Perplexity or Claude answers a vendor question, the model does one of three things. It either pulls from pre-training data (older, slower to update), retrieves fresh content from the live web (ChatGPT Search, Perplexity, Claude with web search), or synthesises a view from both. In all three cases, specific sources get cited or quoted.
The source types that get cited most often in B2B answers are not random. They fall into a clear hierarchy.
- LinkedIn content. SEMrush's 2026 AI citation study found 89,000 LinkedIn URLs cited in AI responses across major B2B categories. Individual posts from named operators carry disproportionate weight because the models treat them as expert opinion rather than marketing copy.
- Original research and data reports. Perplexity cites original research in 73 percent of B2B answers. If your competitor publishes a "State of X 2026" report with proprietary data, it will get referenced across hundreds of buyer queries for the next twelve months.
- Comparison content. "Vendor A vs Vendor B" pages get disproportionately cited because they answer the exact question buyers are asking. If you publish comparisons that name real competitors, you become the source of truth for that comparison.
- Transparent pricing pages. AI models reward structured data. A pricing page with tiers, numbers and feature lists gets extracted cleanly and quoted back to buyers. A page that says "Contact us for pricing" gets skipped entirely.
- G2, Reddit and niche industry forums. User-generated commentary forms the trust layer. AI models treat sentiment here as ground truth.
- Industry publications. Gartner, Forrester, HubSpot, SaaStr, First Round Review, OpenView's Growth Unhinged, Lenny's Newsletter. These carry high trust weights in model evaluations.
The practical test you can run today takes ten minutes. Open ChatGPT. Ask "What are the best [your category] companies in 2026?" Ask it three more variants. "Who is the leading [your category] vendor for [ICP]?" "What is the best [your category] tool for [pain point]?" "[Your category] vs [competitor category]." See whether you appear. See what the model says about you. See who gets cited more.
If you are not in the top five names the model lists, you are not in the shortlist. If the things the model says about you are thin or outdated, your market is being mis-briefed on your behalf.
Why do intent data tools miss the AI dark funnel?
This section matters if you are currently paying for Bombora, 6sense, ZoomInfo intent, Demandbase or similar. None of these tools can see AI dark funnel activity, and the reasons are structural rather than a product gap one of them will close next quarter.
Intent data platforms operate on four signal types. IP-based browsing activity on known publisher networks. Content consumption from gated B2B content syndication. Bid stream data from programmatic ad exchanges. Topic surges derived from aggregated content interactions. Every one of these signals requires the buyer to interact with the open web in a way that leaves a traceable identifier.
When a buyer opens ChatGPT at 9.14 am on a Tuesday and asks about your category, none of the four signals fire. The buyer's IP address is sitting on OpenAI's infrastructure, not on a publisher you can trace back to a target account. No content consumption event exists. No bid stream data because the buyer is not browsing ad-served pages. No topic surge because the interaction is private to the chat.
The practical result is that the buyer completes their early research phase, builds a mental shortlist, and only surfaces to your intent platform when they visit your actual website to book a demo. By that point the shortlist is set. The research has been done. You are reacting to a buyer who has already made up their mind.
A handful of new vendors are trying to close the gap. SEMAI, Profound, Otterly and a few others promise "AI visibility" dashboards that track citation volume, share of voice and sentiment across ChatGPT, Perplexity, Google AI Overviews and Claude. The category is genuinely useful but nascent. Expect product gaps and fast evolution for the next 18 months.
The AI dark funnel is the single biggest blind spot in the current B2B marketing stack. Every dashboard you have is measuring the half of the funnel that still leaves fingerprints.
How do you actually win inside AI search?
Here is the practical framework we use with ORRJO Intelligence clients. Seven moves. None of them are silver bullets. All of them compound.
- Publish original research with clear data points. LLMs cite specific stats. A sentence like "73 percent of B2B buyers used AI in their last purchase" gets extracted and quoted. Vague opinion pieces do not. Run your own survey, even a small one. Publish the numbers with methodology. Make the report easy to link to and easy to parse. This is the single most valuable move you can make.
- Optimise for direct answer extraction. Lead with the answer. Use H2s as buyer questions, not clever headlines. Put your conclusion in the first sentence of each section. The models reward content that answers the question cleanly rather than building towards it over three paragraphs. This is a style shift that takes most teams a month to internalise.
- Build comparison content that names competitors. "X vs Y" pages get cited at a disproportionate rate because they are the exact form of query buyers enter. Name the competitors. Be fair. Be specific. If you win on one dimension and lose on another, say so. Models reward balanced analysis and penalise obvious puffery.
- Invest in LinkedIn thought leadership from named individuals. LinkedIn has 89,000 URLs cited in AI responses according to SEMrush's 2026 study. The highest cited URLs are posts from founders, operators and analysts with strong personal brands. A founder posting three times a week with genuine insight outperforms the company page posting seven times a week on autopilot.
- Make pricing transparent and easy to extract. If your tier structure is hidden behind a form, you lose the pricing query entirely. Publish tiers, prices and what is included. If your real pricing model is usage-based, publish the rate card. "Contact sales" pages are a direct donation to your more transparent competitor.
- Cover category-defining terms in depth. Be the answer for "what is [your category]" and "what is [concept your category solves]." Write the single most comprehensive explanation that exists. This page becomes the default citation for thousands of downstream buyer queries.
- Earn citations from industry publications. Gartner, Forrester, HubSpot, SaaStr, Lenny, First Round, OpenView. These sources carry high trust weights with LLMs. Guest posts, research collaborations, podcast appearances and named quotes in their articles all compound.
None of this is magic. It is content strategy with the knob turned sharply towards "what would an AI model find useful" and away from "what would our SEO agency optimise for." The overlap is substantial but the priorities differ. If you want to go deeper on the broader methodology, see our GTM research agency guide and the State of GTM Research 2026 report.
What kinds of companies are winning the AI dark funnel right now?
Patterns we observe from tracking citation volume across 30 B2B categories for our clients.
Original research publishers are winning. Lenny Rachitsky's newsletter, First Round Review, OpenView's Growth Unhinged, Clay's data reports, 6sense's State of B2B Buyer research, Gong's Labs analysis. All of these get cited across dozens of adjacent buyer queries because the underlying data is proprietary and quotable.
Companies with transparent pricing pages are winning. Attio publishes tiers and numbers. Clay shows usage-based rates openly. Apollo has clear pricing with specific feature cut-offs. When a buyer asks ChatGPT about cost, these names come up with specifics. Competitors that gate pricing get either skipped or summarised inaccurately.
Long-form technical content with named frameworks is winning. Companies that coin a term, document a methodology and publish it in depth get cited every time a buyer asks about that concept. Naming matters. A named framework is easier for a model to extract than a generic "how to" post.
Companies with strong LinkedIn presence from named individuals are winning. The pattern is clear. Category-leading voices post consistently, get quoted in AI answers, drive inbound to their companies. The brand is the individual and the individual amplifies the brand.
Category-definition content is winning. Whoever wrote the canonical explainer for a concept tends to own the citation real estate around it for years. Being first to publish the clear definition compounds.
Notably losing: companies with thin websites, gated content, hidden pricing, no LinkedIn presence from named individuals, and no original research. If that description fits three or more of your competitors, this is your window.
How should you measure AI dark funnel exposure?
Measurement is still rough. That does not mean skip it. Here is what we recommend.
Run quarterly AI search audits. Ask ChatGPT, Perplexity, Claude, Gemini and Google AI Overviews the exact questions your buyers are asking. Record where you appear, how you are described, who gets cited alongside you. Do it every three months. Track the trend, not the snapshot. We include this in every ORRJO Intelligence engagement because the baseline changes fast.
Track citation volume. Which of your URLs are being cited by AI engines? Tools like SEMAI, Profound and Otterly give you this directly. DIY by sampling ten buyer queries a week and logging the cited URLs. After a quarter you have a usable dataset.
Use AI visibility tools if your budget allows. The category is early but functional. Profound and SEMAI both have credible dashboards in 2026. Pricing starts around 400 dollars a month for a team tier. Not a bad investment if you are spending six figures on other marketing measurement.
Add AI search to your self-reported attribution. When a prospect books a call, ask "how did you first hear about us?" and include "ChatGPT, Perplexity, Claude or another AI tool" as an option. Our client data suggests between 8 and 18 percent of inbound prospects now name an AI tool when this option is offered. A year ago it was under 2 percent.
Watch the direct traffic pattern. AI answers often send buyers straight to your domain by name, bypassing the search box. A sustained rise in direct traffic with high intent (demo requests, pricing page visits) with no equivalent rise in organic search is often AI dark funnel activity leaking into your normal analytics.
The best signal of all is qualitative. When a prospect joins your first call and mentions that they "asked ChatGPT about you" before booking, write it down. Frequency matters. One mention is anecdote. Ten mentions a month is a channel.
What is ORRJO doing about this?
Honest positioning. We do not sell an AI dark funnel product. We do not have a dashboard to flog. What we do have is a research-led engagement, ORRJO Intelligence, where AI search visibility is now a standard component alongside ICP work, competitive analysis and messaging audits.
Inside the 14-day engagement, our team runs a full AI visibility audit. We query ChatGPT, Perplexity, Claude and Google AI Overviews across 40 to 60 buyer-style questions for your category. We benchmark your citation volume and narrative against your named competitors. We identify the specific gaps: where a competitor owns the answer, where your content could plausibly get cited if it existed, where an AI model is mis-characterising you. Then we recommend specific content moves to close the gap. Original research ideas. Comparison pages. Category-definition pieces. LinkedIn distribution strategy. Pricing transparency decisions.
This is not a separate product. It is part of how competitive intelligence has to work in 2026. You cannot audit a competitor without auditing how AI search describes them. The two are now the same question. We also publish our own approach publicly for exactly the reasons we recommend to clients: original research like the State of GTM Research 2026, long-form pillar content like our signal-based selling guide and competitive intelligence playbook, and consistent LinkedIn posting from named individuals on our team.
If you want the same analysis run on your category, a 45-minute briefing call with our team is the place to start.
Frequently Asked Questions
Are traditional intent data tools now obsolete?
No, but they only cover half the funnel. Intent platforms like Bombora and 6sense still capture valuable signal around content consumption, IP-based browsing and bid stream activity. What they do not see is buyer activity inside ChatGPT, Perplexity and Claude, which now accounts for a significant slice of early-stage research. Treat intent data as one lane of a multi-lane motorway, not the whole road.
How do I get ChatGPT to recommend my company?
You get cited by earning the same signals that rank high in search but filtered through what LLMs prefer. That means original research with clear data points, comparison content that names competitors, transparent pricing pages, LinkedIn posts from named individuals, and coverage in industry publications that AI models already trust. There is no paid placement or algorithm to game. You earn citations by being genuinely useful.
Does SEO still matter if buyers use AI?
Yes. AI models use the same web index. Perplexity, ChatGPT Search and Claude all crawl or reference web content when answering vendor questions. Pages that rank well in Google tend to get cited in AI answers too. The difference is that AI answers reward long-form depth, clear data, and direct answer extraction more than traditional SEO keyword density. Good SEO is still the floor. AI-ready content is the ceiling.
What is AI visibility?
AI visibility is the emerging category measuring how your brand appears inside LLM answers and AI search results. Tools like SEMAI, Profound and Otterly track citation volume, sentiment, share of voice and competitor benchmarks across ChatGPT, Perplexity, Claude and Google AI Overviews. It is the AI dark funnel equivalent of rank tracking.
How long before AI search overtakes Google for B2B research?
Gartner projects that 25 percent or more of B2B research will move to AI assistants by 2027. That does not mean Google dies. It means the first touch moves upstream of the search bar. By the time a buyer lands in Google to check a specific vendor, their shortlist is often already set by the answer they got from ChatGPT or Perplexity an hour earlier.