[00:00]
Ashley: Welcome to Podcast7, a new interesting episode today. I am Ashley.
Ray: And I am Ashley's co-host Ray. Welcome everyone. Today we are telling a ghost story.
Ashley: Oh, but uh not the kind with chains and haunted houses. This is a ghost story for B2B marketers.
Ray: You know that feeling when you're absolutely certain someone is in the room with you? You can feel their presence. You know they're looking at your stuff, but you just can't see them.
Ashley: Spooky, but uh very accurate. And that's basically the state of B2B demand gen right now. We've been looking at this massive stack of research reports from Forrester, Gartner, and some really dense privacy policies from LLMs. The headline is just unsettling: the identity graph, the foundation for everything, is going dark.
Ray: It is. But I want to be careful here because going dark sounds like the lights are just cutting out completely. It's a bit more nuanced. The identity graph isn't dying; it's retreating. It's moving behind walls that marketers just aren't allowed to climb. We are trading observable behavior for invisible synthesis.
Ashley: Okay, so before we hit the panic button, let's ground this. We need to make sure everyone is visualizing the same thing, because "identity graph" is one of those buzzwords people nod at but might not fully hack.
Ray: Right. For the last 10 years, if you were in B2B, you lived and died by this graph. It’s a data architecture that links identifiers to build a cohesive picture of buyers at both the contact and account level.
Ashley: Think of it like a digital fingerprint that spans devices. Layer one is the individual identity—connecting Jane Doe to her work email, laptop, and mobile phone.
Ray: And layer two is the account mapping layer. This is where the money is for B2B. It connects Jane to Acme Corp through IP addresses and corporate domains. The whole promise was seeing Jane’s journey across the open web.
[02:15]
Ashley: And the key word there is "open web". If Jane was searching for CRM software on Google, we had signals because the browsing was observable. But the center of gravity has shifted.
Ray: It really has. Research from Forrester shows that in 2024, 89% of business buyers used AI in the buying process. By 2025, that's projected to hit 94%.
Ashley: 94%! That’s pretty much everybody. It looks like AI has become the interface for information, from summaries at the top of Google to sidebars in Microsoft Edge.
Ray: Which brings us to our thesis: we are moving from observable web browsing to "conversational AI-mediated research". And the scary part? We can't see into those conversations. That is the ghost in the room.
Ashley: That’s the concept of "intent enclosure". It’s not just a blind spot; it’s a black box.
Ray: Let's unpack that. A lot of us assume we can just track AI like we track Google search, but the sources show these are walled gardens on steroids.
Ashley: Take OpenAI. They are testing ads in ChatGPT now, but they explicitly state that conversations remain private from advertisers.
Ray: So if Jane from Acme Corp asks ChatGPT to compare the security protocols of Vendor A versus Vendor B—the most valuable intent data imaginable—it is locked inside the box. You don't know who she is, you don't know her company, and you don't know the context.
[04:45]
Ashley: And it gets even more restrictive with Microsoft Copilot. Because Copilot is integrated into Microsoft 365, the data lives inside the company's private "tenant".
Ray: Exactly. If Jane asks Copilot for software help, that isn't search data Microsoft sells; it’s internal telemetry belonging to her employer. You can't track a pixel in a private Teams meeting summary.
Ashley: It turns what used to be third-party intent data into first-party internal data for the buyer. We have zero visibility.
Ray: It makes the "cookie apocalypse" look like a total red herring. It doesn't matter that Google kept cookies. AI chatbots don't use cookies. Gartner even predicts traditional search volume will drop 25% by 2026.
Ashley: The cookies are safe, but the users are leaving the bakery. So, what can we actually see?
Ray: The source material proposes a new signal taxonomy for the next 1 to 3 years. We have to stop looking for the old stuff and recognize the new buckets.
Ashley: Bucket one is LLM-native signals—the actual prompts. These are basically opaque to external marketers.
[07:30]
Ray: Bucket two: LLM advertising signals. You might get a click from a sponsored unit in Perplexity or ChatGPT, but you lose the conversational context. You don't know if they spent 20 minutes arguing with the bot before clicking.
Ashley: Bucket three is "answer engine visibility". This is the new SEO—or GEO, Generative Engine Optimization. It's about your brand's presence in the model’s latent space. Are you being cited?
Ray: And bucket four: first-party deep signals. This is the gold mine where the identity graph still works better than ever.
Ashley: Because the top of the funnel is dark, the moments when they hit your site—authenticated sessions, trial signups, API docs—become critical.
Ray: Exactly. Buyers are doing their sense-making before they ever visit you. By the time they land on your page, they aren't there to learn; they're there to validate what the AI told them.
[10:15]
Ashley: Before we engineer for this, a quick word for our partners. [SPONSOR] If you're a marketer figuring out AI demand gen, check out Demand7 at demand7.ai. For the engineering side of AI execution, hit up GTM7 at gtm7.ai.
Ray: Their dashboards for tracking "share of answer" are incredible for visualizing that invisible layer.
Ashley: So back to GEO. In the old world, you optimized for keywords for a click. In the new world, you optimize for synthesis. You want the LLM to understand your product so clearly that you are the primary recommendation.
Ray: But there's tension. Publishers are suing AI companies for crawling content. If you block crawlers to protect your IP, you disappear from the answer engine.
Ashley: For B2B vendors, the move is to be crawled. You aren't selling ads on a blog; you're selling software. You have to feed the machine so it knows your pricing and features.
Ray: This leads to "Pattern B": building your own AI inside your stack using RAG—Retrieval Augmented Generation.
[13:00]
Ashley: Rag grounds a model like GPT-4 in your own library of documents and manuals. When a buyer asks your on-site co-pilot a question, you capture that prompt and can connect it to their identity.
Ray: You regain the visibility you lost on the open web. It’s like bringing the walled garden inside your own fence.
Ashley: This also addresses the "attribution bifurcation". Inside walled gardens, you rely on their platform metrics. Outside, you move to incrementality experiments and account-level progression.
Ray: So practical strategy time: we have to "re-anchor" the graph. If the top of the funnel is invisible, you anchor identity on high-confidence moments downstream.
Ashley: Like trial activations or visiting security and compliance documentation.
Ray: Exactly. Casual browsers don't read SOC2 reports. Someone auditing your API rate limits is screaming, "I am ready to buy".
[15:45]
Ashley: And you’re tracking a committee, not an individual. Forrester says typical buying decisions now involve 13 internal stakeholders.
Ray: If the CTO hits the API docs, the CFO hits pricing, and the compliance officer hits security, that constellation of signals is your truth.
Ashley: Your content needs to be "validation-ready". No fluff—give them ROI models, implementation guides, and specs.
Ray: It's a shift from fighting for attention to fighting for verification. The identity graph has retreated to the validation phase.
Ashley: Which might be better. We stop chasing vague interest and start servicing actual intent.
Ray: I want to leave everyone with a provocation today.
Ashley: If your buyer completes 94% of their research inside a platform that refuses to share their identity with you, is your traditional top-of-funnel identity graph already obsolete?
Ray: Are you measuring a funnel that doesn't exist anymore?
Ashley: Think on that. Continue the conversation at podcast7.ai.
Ray: Thanks for listening
