The Notion Problem: Losing the Enterprise Conversation

Consider a hypothetical. A company is the best known product in its category. Everyone in the market can name it. Its brand recall is close to universal, its funding is enormous, and its logo is on the laptop of every operator you know.

A buyer at a 4,000 person company opens ChatGPT and asks which tool their compliance team should use for documentation with SSO, audit logs, and data residency in the EU. The market leader may or may not appear in that answer. Its market position has surprisingly little to do with it.

Call this the Notion problem, using Notion only as a familiar stand-in for a large, well-known workspace tool. The pattern applies to any category leader, and it is a pattern rather than an audit. I am not claiming anything about what any model says about any specific company today. I am describing a structural risk that scale creates.

Retrieval does not know who is winning

Traditional search had a strong bias toward incumbency baked into it. Domain authority, link graphs, brand queries, and click behavior all reward the biggest player. Being famous was, mechanically, a ranking factor.

Answer engines work differently at the point of assembly. A model receiving a specific question does a retrieval pass over documents, pulls candidate passages, and synthesizes. The unit of competition is the passage, and the question it has to satisfy is narrow. Does this chunk of text directly address SSO, audit logs, and EU data residency for a compliance team?

Brand strength is an input to that process, mostly through the training data and through the authority of the sources that get retrieved. It is not the deciding factor at the moment of assembly. Specificity is.

Fame gets you into the training data. It does not get you into the answer for a question you never wrote a page about.

Why leaders write worse content

There is a predictable arc to how content quality degrades as a company grows, and it works against retrieval at every step.

• Positioning goes abstract. Early on you write about the specific problem you solve because that is all you have. At scale you write about vision, category creation, and the future of work. Abstract copy matches nothing.
• Breadth replaces depth. A leader serves many segments, so the site flattens into one page per broad use case. The narrow buyer question, which is where purchase intent lives, gets no page at all.
• Brand governance flattens claims. Legal and comms review removes the concrete numbers, the direct comparisons, and the plain admissions of what the product does not do. Those are exactly the sentences retrieval likes.
• Docs and marketing split. The precise answers move into a documentation subdomain written for existing customers, phrased in internal vocabulary, and disconnected from how prospects ask.

Meanwhile the challenger is doing the opposite. It has one segment, one wedge, and a strong incentive to write a page for every objection a buyer might raise. It publishes the comparison the leader will not publish. It answers the security questionnaire question in public. It names the competitor.

The enterprise conversation is the vulnerable one

This asymmetry hits hardest at the top of the market, and that is the part worth paying attention to, because enterprise deals fund everything else.

Enterprise buyer questions are long, compound, and constraint-heavy. Data residency. Retention policy. SCIM provisioning. Admin controls per workspace. Procurement wants a comparison against two named alternatives. Security wants a specific answer about encryption at rest. Each of those is a narrow retrieval query, and each one is winnable by whoever wrote the clearest public paragraph about it.

Large companies often have excellent answers to all of these. They are just locked inside sales collateral, PDFs behind forms, and trust centers that require a login. A gated answer cannot be retrieved. From the model's perspective, it does not exist.

Where consideration sets get formed

The stakes are higher than one answer, because the shortlist is where the damage happens. If a buyer asks an assistant to narrow a category to three options and you are not in those three, you never get evaluated. There is no impression, no click, no chance to win on product.

Traditional search leaked at every stage but always gave you a shot at appearing. Answer engines compress ten blue links into a short list, and short lists are brutal. Second place on Google still gets clicks. Fourth place in an AI shortlist gets nothing, because it is not in the shortlist.

What the leader should actually do

The instinct is more brand spend, which does nothing for retrieval. The work is duller and more effective.

Start by enumerating the real buyer questions. Not keywords. Questions, in the compound, constrained form buyers actually type into an assistant. Sales calls and support tickets are a better source for these than any keyword tool.

Then check which of those questions have a public page with the answer near the top. Most will not. Write those pages, keep them ungated, and let them be specific enough to be uncomfortable, including where the product is a poor fit. Honest limitations get cited because they read as trustworthy and they resolve exactly the query a skeptical buyer is running.

Publish your own comparisons. If you do not, the only comparison pages in the retrieval pool will be written by challengers, and those pages will define you in every answer where the question is competitive.

The uncomfortable summary: category leadership is a lagging indicator, and answer engines read leading ones. The company that answers the question wins the question, no matter who is bigger.