The session that starts too late
Every analytics tool you own measures the same thing: what happens after someone arrives on your website. Sessions, pageviews, time-on-site, scroll depth, conversion rate. The entire discipline of web analytics is built around the moment of arrival.
But B2B buyers don't begin their journey when they open your homepage. They begin it when they first feel the problem. That might be a capacity crunch in Q3. A bad experience with their current engineering provider. A conversation with a CFO who asked an uncomfortable question about project delivery costs.
In the old model, that early-stage research came to your website eventually. They'd google "what is consulting," find your glossary, read a few articles, gradually build familiarity. Your analytics would capture all of it — imperfectly, but directionally. First touch, last touch, multi-touch. The funnel was porous and messy, but it was at least on your domain.
In 2025, the early buyer journey migrated almost entirely off your domain — into a conversation with a machine that synthesizes your content, your competitors' content, and a decade of internet knowledge into a single paragraph.
The question your prospect asked ChatGPT at 11pm was: "What's the difference between external consulting and building an in-house team?" The answer cited no single source. It compressed what might have been a three-session research journey into ninety seconds. And your analytics captured none of it.
Why seven phases — not three
The traditional funnel — Awareness, Consideration, Decision — was always a gross simplification. It was good enough when the research journey happened on Google and you could at least see the keyword signals. TOFU content ranked for problem-awareness queries. MOFU content ranked for comparison queries. BOFU content converted.
This architecture collapsed for a simple reason: LLMs don't expose keyword signals. When a buyer asks an AI assistant about your category, the query disappears. You get a branded visit, eventually, but the chain of reasoning that produced it is permanently invisible.
To build content that performs in this environment, you need a more granular model of what buyers are actually thinking at each stage of their research. Not three buckets — seven distinct mental states, each with specific information needs.
The Seven Information Situations — Buyer Mental States
the Problem
the Term
Solutions
Options
Criteria
Providers
to Decide
This is not an academic taxonomy. It is a practical content architecture framework. Each of these seven situations requires a fundamentally different type of content — different structure, different vocabulary, different function. And critically: each maps to a different type of LLM prompt.
The separation of Compare Options (comparing service types) from Compare Providers (comparing providers) is not a subtle distinction — it's decisive. A buyer asking "consulting vs. in-house team" is in a completely different mental state than a buyer asking "which consulting firm is right for my company." Your content must serve both. Most companies confuse the two and build content that serves neither particularly well.
The decision phase, disaggregated
The decision phase deserves particular scrutiny, because it's where the most damaging simplification occurs. "Decision content" is typically understood as testimonials, case studies, and contact forms. This misses at least five distinct buyer situations:
Trust verification — the buyer has chosen a category and is now asking: is this provider serious, stable, reliable? They're looking for third-party validation. Awards. Customer volume. Years in business. Compliance signals.
Fit assessment — the buyer needs to know if the provider's solution fits their specific profile. Industry. Company size. Project scope. A manufacturing company has entirely different engineering needs than a software startup. Generic provider content fails both.
Claim verification — the buyer has heard something about the provider and needs to verify it. Often this is an LLM-generated claim: "I was told rhinegold is part of a regional banking group." Is that accurate? What's the ownership structure? These questions need clear, findable answers on your site.
Criteria clarification — the buyer doesn't yet know what to look for in a provider. They need help understanding what distinguishes a good consulting partner from a mediocre one. This is not comparison content — it's framework content.
Switching readiness — the buyer currently has a bank relationship and is weighing whether and how to introduce an independent financing partner alongside it. This is a specific, high-stakes situation that almost no provider addresses directly.
What LLMs absorb — and why it matters for your content
Language models don't index pages. They absorb patterns. When a buyer asks an LLM "what are the advantages of external consulting over building in-house," the model draws on the accumulated weight of every relevant document it has encountered — prioritizing sources that were clear, well-structured, frequently cited, and conceptually precise.
This creates a feedback loop that should alarm any B2B marketer. Your definitional content — glossary pages, explainer articles, FAQ sections — is precisely the content type that LLMs absorb most readily and reproduce most faithfully. It is also the content type that receives the least direct traffic benefit from this absorption, because LLMs answer definitional questions directly, without referral.
The content that performs best as LLM training signal — definitional, precise, well-structured glossary and explainer content — also receives the least direct traffic from LLM citations. You are feeding the model that diverts your traffic.
This is not an argument against creating definitional content. It is an argument for understanding that definitional content now serves a different primary function: establishing conceptual authority in the model's knowledge base, which surfaces as brand citation in later-stage provider queries.
The implication is counterintuitive: a company with 150 precise, well-structured glossary pages is building LLM authority even as its organic traffic from those pages declines. The authority materializes not in glossary clicks, but in the likelihood of brand mention when a buyer asks "which consulting provider should I consider for my mid-sized manufacturing company."
The implication for your content strategy
Traditional SEO optimized for ranking. The signal was: does this page appear on page one for this query? The success metric was click-through rate from search results pages.
Content strategy in the LLM era requires a different optimization target. The signal is: does this content become part of the model's understanding of this category? The success metric is citation rate — how often does the LLM mention your brand in contextually relevant responses.
| Content situation | Old optimization target | New optimization target |
|---|---|---|
| Problem awareness | Rank for symptom keywords | Provide conceptually precise problem framing the LLM will absorb and reproduce |
| Category education | Drive traffic to glossary pages | Establish definitional authority that generates brand citation in category queries |
| Option comparison | Rank for "X vs Y" comparison queries | Provide comparison frameworks the LLM uses when constructing its own comparisons |
| Criteria clarification | Rarely targeted — no search volume | Critical: LLMs answer selection criteria questions constantly; your framework becomes the answer |
| Provider comparison | Landing pages and case studies | Trust signals, specific USPs, and verifiable claims that LLMs can accurately reproduce |
The underlying shift is from competing for clicks to competing for conceptual real estate in the model's representation of your category. This requires a content architecture built around semantic precision — not keyword density, not engagement metrics, but the clarity and completeness with which your content covers the information needs of each buyer situation.
In the next episode, we'll look at the measurement challenge: how do you detect LLM citation patterns, what does a GEO audit actually measure, and what early data tells us about which content types drive brand visibility in AI-generated responses.