The discovery call used to be where a vendor first learned what a buyer actually cared about. The buyer would show up, explain their problem, and the seller would start qualifying. That dynamic has shifted. Buyers are showing up to discovery calls having already formed a view of the category, compared approaches, and in many cases, pre-selected a shortlist. They didn’t get there through your website.
They got there through an LLM.
What buyers are actually doing
The research behavior that used to happen across multiple sessions — search queries, review sites, analyst reports, peer recommendations — is increasingly collapsing into a single conversation with an AI assistant. A buyer with a problem to solve opens ChatGPT, Claude, or Perplexity and starts asking questions.
Not keyword searches. Questions.
“What’s the best way to manage X for a company our size?” “What are the tradeoffs between approach A and approach B?” “Which vendors are worth looking at for Y, and what should I watch out for?”
The LLM synthesizes an answer from everything it knows — vendor content, documentation, analyst coverage, review platforms, community discussions, third-party comparisons — and delivers a confident, structured response. No blue links. No results page. No clicking through to twelve different sites. This is AI Search — and it leaves no trackable signal for the vendors being researched. This is zero-click research at its most consequential for B2B vendors.
McKinsey found that B2B decision-makers now use an average of ten channels in their buying journey, up from five in 2016. AI assistants are the fastest-growing of those channels, and the only one that generates zero visibility for the vendor being researched.
The queries that matter most
Not all LLM queries are equal from a vendor perspective. The ones that shape buying decisions fall into a few distinct patterns. Together they define your query coverage — whether you appear across all buyer journey stages or only in a narrow subset.
Category education queries.
“How do companies approach X?” or “What’s the difference between Y and Z?” The buyer is orienting themselves in a space they don’t fully understand yet. The LLM’s answer shapes their mental model of the category — including which frameworks, terminology, and vendors get treated as legitimate.
Vendor discovery queries.
“Who are the leading vendors for X?” or “What platforms do companies use to solve Y?” This is the shortlist moment. The LLM names two, three, maybe four vendors. Everyone else doesn’t exist in that conversation. Each mention is an LLM citation that shapes the buyer’s consideration set. Where your company appears in that list is your LLM Position — and it matters.
Comparison queries.
“How does Vendor A compare to Vendor B?” or “What are the limitations of Product X?” The buyer is stress-testing options they’ve already identified. This is where positioning accuracy matters — what the LLM says about your product’s weaknesses is just as important as what it says about your strengths. This is your AI Brand Presence in action.
Implementation queries.
“What does a typical deployment look like for X?” or “How long does it take to get value from Y?” These are buying queries, not research queries. A buyer asking about implementation timelines has already cleared the awareness stage. They are evaluating whether the effort is worth it.
Compliance and technical queries.
“Does Vendor X support SOC 2 in multi-region deployments?” or “What’s the security model for Y?” In regulated industries or enterprise contexts, these queries happen before a buyer will engage with a vendor at all. If your documentation isn’t public and structured, the LLM either gets it wrong or skips you.
The demographic reality behind the shift
Nearly three-quarters of B2B buyers are now Gen Z or Millennials. This is the generation that grew up on YouTube, Reddit, and social platforms — not corporate websites. They treat AI assistants the way previous generations treated Google: as the starting point for any unfamiliar problem.
They enter buying processes more prepared than any generation before them. They involve more stakeholders, move faster through evaluation, and have less tolerance for sales friction. Sending them a brochureware website and waiting for a form fill is a disqualifying signal. By the time they reach your website, they’ve already decided whether you’re worth their time.
89% of B2B buyers have adopted generative AI as a primary source for self-guided information across all buying stages. That adoption happened in under two years.
What this means for how you get found
The buyer research happening inside LLMs is invisible by design. No pixel fires. No cookie gets set. No session registers in your analytics. A buyer can research your category, compare you to three competitors, form a strong preference, and never touch your website — and you would have no idea it happened.
This is the dark funnel. It existed before AI in the form of Slack conversations, subreddit threads, and peer recommendations. AI has industrialized it. The volume of pre-purchase research happening outside your visibility is growing faster than most revenue teams have registered.
The practical implication is straightforward: the content, credibility signals, and third-party presence you build today determine how AI systems describe and recommend you to buyers you’ll never see coming. Your review site profiles, analyst mentions, community presence, and thought leadership aren’t just brand plays. They are your AI citation pipeline. By the time traditional buyer intent signals register in your CRM, the shortlist is already formed.
Buyers are already there. The question is whether your company shows up when they ask.
For a detailed map of how buyers actually move through each stage — from problem recognition to post-decision — see What is the AI Buyer Journey? For a deeper look at how shortlist formation works specifically, see How B2B Buyers Use AI to Build Vendor Shortlists. For more on how to build presence in this channel, see What is the AI Demand Channel? and What is AEO?
B2B buyers open ChatGPT, Claude, or Perplexity and ask natural language questions about their problems and solution options. They run category education queries to orient themselves, vendor discovery queries to build a shortlist, comparison queries to stress-test options, and implementation queries to evaluate effort. The entire process happens inside AI conversations without visiting any vendor website.
Five query types shape B2B buying decisions: category education queries (orienting buyers in an unfamiliar space), vendor discovery queries (building the initial shortlist), comparison queries (stress-testing shortlisted options), implementation queries (evaluating deployment effort), and compliance or technical queries (validating fit in regulated industries). Vendors invisible in any of these query types miss that stage of buyer influence entirely.
LLM research generates no trackable signals. No pixel fires. No cookie gets set. No session appears in analytics. A buyer can research your category, compare you to competitors, and form strong preferences entirely inside AI conversations — and you would have no idea it happened. By the time any traditional intent signal reaches your CRM, the shortlist is already formed.
LLMs cite vendors based on the quality and structure of their public content, third-party validation through analyst mentions and review sites, consistency of positioning across all surfaces an LLM might read, and recency of content. Vendors with ungated technical documentation, strong review site presence, and consistent messaging across their digital footprint appear more frequently and more accurately in AI responses.
Vendors cannot intercept LLM research — they can only shape what AI systems say about them. This means structuring content for AI extraction through AEO, building third-party citation presence through analyst relations and review sites, ungating technical documentation so AI can access it, and measuring Share of LLM regularly to track whether AI responses include and accurately represent your company.