Your pipeline is dying before sales ever touches the lead

You think you have a conversion problem. It might be earlier than that. Buyers are asking AI about you first, and getting an outdated answer instead of you.

A few weeks ago, I typed the name of a company into ChatGPT.

The description wasn't wrong.

It was off.

It sounded like a company that had existed at some point in its history, but not the company that exists today. Part of the answer came from an old positioning. Part reflected a product the company had already outgrown. Another part sounded suspiciously close to one of its competitors.

Nothing in the answer was false.

That's what stayed with me.

If the answer had been obviously wrong, a buyer would question it. Instead it was plausible enough to become someone's first impression.

That's when I noticed something I hadn't thought about before.

Not really about that company. About what a stranger typing your company's name into the same box might currently be reading.

There was always another conversation coming

For years, marketers have worried about how the market understands their company. We invest in positioning, messaging, analyst relations, customer stories, and category design because we know perception shapes buying decisions.

But we also relied on something we rarely named out loud.

There was always another conversation coming.

If an analyst misunderstood the company, someone booked a briefing. If a prospect compared you to the wrong competitor, sales corrected it on the first call. If the market was still using an old description, product marketing spent a quarter changing the narrative.

The understanding was never perfect. But it was negotiable. Humans asked questions. Humans clarified. Humans resolved the ambiguity between what a company said about itself and what the market believed.

That correction loop quietly protected companies from a lot of bad interpretations.

AI doesn't remove it. It changes when it happens.

Not because AI creates misunderstanding. Because it increasingly becomes the place where the first interpretation happens, before anyone from your company is in the room.

Where does that understanding come from?

This is why I think most conversations about AI optimisation start in the wrong place. We ask how to rank. How to earn citations. How to appear in AI generated answers. Fair questions, but they assume AI already knows what kind of company it's looking at.

Where does that understanding come from?

The more I sat with that question, the clearer it became this wasn't really an AI question. It was a marketing question AI had made impossible to postpone.

Curious whether this had been studied before, I looked into how market categories form. Markets don't discover categories. They converge on them, through the accumulated language of companies, analysts, journalists, and customers, until a shared definition settles into place.

Categories aren't found. They're negotiated. And AI doesn't invent them, it inherits them, not from your website or your newest positioning deck, but from the accumulated description the market has already produced: analyst reports, reviews, customer stories, comparison articles, job postings, product pages, all feeding the same understanding.

AI isn't learning directly from your company. It's learning from how the market has collectively described your company.

The market writes. AI reads. Buyers decide.

AI inherits the market's interpretation of your company before your company enters the conversation.

AI has no independent understanding of your company.

It reconstructs one from the evidence that forms the most coherent picture.

What the test showed

That observation isn't only theoretical. I ran a structured test on two enterprise SaaS companies that had each completed a major repositioning within the previous eight months. Without retrieval, every AI system I tested consistently described the companies as they had existed before the change. The moment retrieval was enabled, the new identity appeared almost completely. The transition wasn't gradual. It was binary. AI's internal knowledge inherits historical market consensus. Left on its own, it does not update until a newer model is trained. What changes is whether the model is allowed to look beyond that internal knowledge before it answers. Retrieval overrides the stale version almost immediately once fresh evidence is available. The model's own internal knowledge does not.

Which means correction hasn't disappeared. It now depends entirely on whether coherent, current evidence already exists for AI to find, not on whether a human eventually gets the chance to explain it.

The full case study, including the platforms tested, the exact prompts used, and where each AI system's understanding actually came from, is here: AI inherits identity late. It corrects fast. The gap in between is the risk.

Visibility is not the first competition

Once I realized AI inherits categories rather than inventing them, a second question followed: if categories are inherited, who is actually shaping what gets inherited?

Here's where this stops being ordinary category strategy.

Anyone who has worked in marketing knows how hard it is to shift a first impression once it settles. That same difficulty now shows up one step earlier, before a human is even involved.

I used to believe the main competition was visibility: who shows up most often in AI answers. I'm no longer sure that's the first competition.

Over the past months I've started seeing the same pattern repeatedly. One company dominates AI visibility reports. Another quietly becomes the vocabulary the market uses to describe the problem. You notice it in analyst reports, comparison pages, review sites, and eventually in procurement conversations. By the time buyers start evaluating vendors, the criteria themselves often reflect one company's framing more than another's visibility.

The company with more citations isn't always the company that gets chosen. Sometimes it's the company whose language everyone else, including its competitors, has already started using to describe the category.

Visibility determines whether you're seen. Definition determines what you're seen as.

Those are different competitions, and only one of them shows up in a citation count.

Why Semantic Identity Systems

The more I watched how AI described organisations, the more convinced I became that visibility was rarely the deepest issue. The deeper issue was whether the market had produced one coherent identity for AI to inherit in the first place.

That's ultimately why I started working on Semantic Identity Systems, a discipline for auditing and aligning how AI classifies a company before that classification ever reaches a buyer. It wasn't designed to optimise AI outputs. It was designed to shape the identity AI inherits before those outputs exist.

The goal isn't to teach AI something new about a company. It's to reduce the contradictions AI has to resolve when reconstructing that company from what the market has already said.

SEO helps search engines find you. GEO helps AI reference you. Both matter. Neither answers the earlier question.

What company is the market currently teaching AI that you are?

That's the question Semantic Identity Systems is built to answer.

When category strategy starts to matter

AI hasn't changed the importance of category strategy. It has changed when category strategy starts to matter.

For decades, companies could rely on human conversations to shape a buyer's first interpretation, not just repair a later one. That's what's shifting. The buyer's first interpretation increasingly forms before anyone from the company has the chance to participate.

Which means the companies that win won't necessarily be the ones producing the most content or collecting the most citations. They'll be the companies whose identity the market has taught AI to reconstruct consistently, accurately, and with confidence, long before the conversation with a buyer ever starts.

Companies used to manage messaging. Now they increasingly need to manage the evidence AI inherits before any messaging reaches a buyer.

Every recommendation begins with an interpretation.

Every interpretation begins long before anyone asks AI the question.

The first conversation about your company may already be happening without your company.