Your Buyers Asked ChatGPT About You. Here Is What It Said.

A simple exercise reveals the problem faster than any deck: take the 40 questions a company's buyers most plausibly ask an AI assistant ('best platforms for X', 'how should we evaluate Y', 'alternatives to Z') and run them across the four major LLMs. The results follow a pattern that surprises almost every leadership team.

Why this became urgent

For fifteen years, being findable meant ranking in Google. That assumption quietly expired. A growing share of B2B evaluations now begins with a question typed into an AI assistant, and the answer arrives as a short, confident list of vendors. There is no page two. There is no tenth blue link that still gets a trickle of clicks. Either your brand is inside the answer or the buyer never learns you exist. This is the shift we describe in the buying journey timeline: the research phase has moved into a surface you do not control and cannot buy your way into with media spend. The companies that treat AI answers as a first-class channel this year will compound an advantage that is very hard to claw back later, because models keep citing the brands they already know.

The three outcomes

Most brands simply do not appear. Not on the shortlist, not in the comparison, not even as an 'also consider'. A smaller share appear but are characterised inaccurately, positioned in the wrong category, credited with a competitor's weaknesses, or described using messaging from two rebrands ago. Only a minority are present and represented correctly.

How to run the audit yourself

You do not need tooling to get a first read. Write down the questions your buyers actually ask at each stage: category discovery ('what tools exist for X'), evaluation ('compare A and B for a mid-market team'), validation ('is A credible for enterprise'), and alternatives ('cheaper options than A'). Aim for thirty to forty prompts. Run each one in ChatGPT, Perplexity, Gemini, and Claude, in a clean session with no history. Record three things per prompt: did you appear, where in the list, and what the model actually said about you. Repeat the run monthly with the same prompt set so the numbers are comparable. The first run takes an afternoon and usually reshapes the next quarter's marketing conversation more than any campaign review.

Why this happens

LLMs form opinions from the public record: third-party mentions, review sites, community threads, analyst coverage, and the consistency of your own narrative across the web. If that record is thin, stale, or contradictory, the model fills the gap with whatever is available, usually your loudest competitor's framing.

The sources that carry the weight

Not all mentions count equally. Models lean on sources that look authoritative and get repeated: comparison and review platforms, industry publications, practitioner communities, well-maintained documentation, and structured data on your own site. A brand with ten thoughtful third-party mentions in respected places will usually beat a brand with a thousand pages of its own blog content that nobody else cites. That is why the fix is not 'publish more'. It is closer to public relations than to content production: earn descriptions of your company, in other people's words, on surfaces the models trust. Your own site still matters, but mostly as the canonical record that keeps every other description consistent.

Misrepresentation is the quiet killer

Invisibility at least tells you where you stand. Misrepresentation is worse because it works against you silently. If a model describes you as an SMB tool when you sell to enterprise, you are being filtered out of the exact conversations you win. If it attributes a pricing model you abandoned two years ago, buyers arrive with the wrong anchor or do not arrive at all. Stale descriptions usually trace back to old press releases, outdated review profiles, and abandoned pages that still outrank your current story. The correction is unglamorous: update the record everywhere it lives, retire or redirect stale pages, and give the models a consistent, current description to converge on.

What to do this quarter

Start by measuring. Establish your citation rate across a fixed prompt set and track it monthly, the way you track branded search. Then close the gaps in order of leverage: correct your structured data, refresh the third-party sources models cite most, and publish authoritative content that answers category questions directly. Visibility compounds, but only after you can see it.

A 90 day correction plan

Week one to two: run the audit, fix your own house (structured data, a clear category statement on the homepage, consistent descriptions across every profile you control). Week three to six: target the third-party record. Update review-site profiles, pursue two or three credible industry mentions, and make sure your best comparison content answers the questions buyers actually ask. Week seven to twelve: publish one substantial, citable asset and promote it where practitioners in your category talk. This is the same authority work we unpack in the content flywheel, and it is a core module of our AI visibility practice. None of it is exotic. What makes it work is sequence and consistency, not volume.

Can you not just pay for placement?

The question every executive asks next: is there an advertising product that skips the work? Mostly, no. The major assistants do not currently sell placement inside organic answers, and where sponsored results are being tested they arrive clearly labelled, in a context where buyers went specifically to escape advertising. Even if paid slots mature, the economics will mirror search: ads rent attention while the organic answer keeps compounding, except here the 'organic ranking' is the model's learned opinion of your brand, which no budget can rent. That is uncomfortable news for teams hoping to buy their way in this quarter, and excellent news for teams willing to build the record, because the moat is made of exactly the slow work competitors keep postponing. The earlier you start, the longer the head start you are defending.

How you know it is working

Watch four numbers together: citation rate on your fixed prompt set, accuracy of how you are described, branded search volume, and the share of inbound conversations that mention an AI answer. They move in that order. Citations improve first, description accuracy follows, and demand signals lag by a quarter or so. If you want an outside baseline before investing, this exact audit is the first deliverable of our Revenue Diagnostic: where you appear, how you are characterised, and what to fix first. However you run it, start measuring now. You cannot manage a channel you have never looked at, and your buyers are already asking.