If you want to know where your product’s specification-marketing has holes, stop asking your agency. Open ChatGPT, type the prompt an architect or QS in your category would type today, and read the answer. That’s your audit. It takes about ten minutes and tells you more than most quarterly reviews.
Here’s why that works now, and didn’t a year ago. The evidence trail that always won specifications) published case studies, BIM objects, technical data sheets, standards compliance, trade-press mentions, third-party project references) is the same evidence AI reads to decide who to name first. If those assets are thin, gated behind a login, or missing from the parts of the internet that get indexed, you’re invisible to the model. And you’re invisible to the specifier who was going to Google you next. The problem isn’t new. AI just made it measurable.
So run the prompt. A real one. Not “list the top brands of X” – specifiers don’t type that. They type something closer to: “I’m scoping acoustic ceiling systems for a school retrofit in Gauteng, mid-range budget, Class A rating, easy maintenance access. What should I be considering, and what’s the evidence for each?” Read who gets named. Read in what order. Read what proof the model cites: is it a case study on a real school? A named standard? A BIM library entry? A trade article? Then check whether any of that proof is yours. If your product isn’t there, or is there on thin evidence, you now know the gap. Not as a hunch. As something you can hand to the person who owns your marketing collateral.
What the gaps are actually telling you is worth being clear about, because it’s easy to misread. The model isn’t rejecting your product. It’s reading what’s publicly indexed. If your best case study is a PDF locked behind a lead-capture form, the model can’t read it – and neither can half the specifiers who won’t fill in the form to see it. If your BIM object lives only on your own site and not in the libraries specifiers actually use, it isn’t in the answer. If the last trade-press mention of your product on a live project was three years ago, that’s a signal too. None of this is exotic. It’s the specification-marketing infrastructure that’s always mattered. What’s changed is that it now has a second job on top of persuading humans, it has to be legible to the systems those humans use to shortlist.
That opens up a spend argument worth putting to your team. A single well-placed trade-press case study (real project, real specifier, real result) now earns its keep three ways. The specifier reads it in a magazine or newsletter (PR). Google indexes it and it turns up in searches (SEO). The model cites it when someone asks the question your product answers (AI visibility). Most marketing teams still budget for those as three separate line items, often with three separate agencies. They’re not three jobs. One good asset does the whole job. Teams that spot this earlier tend to spend less and surface more.
On Monday, this is what we’d do. Pick your top three product categories – the ones where getting specified matters most to the number. Write one prompt per category in the voice of a real specifier: a project brief, a constraint, a request for the evidence behind each option. Run each prompt twice, in ChatGPT and one other model, so you’re not reading one system’s quirks as a pattern. Note which of your products got named, on what evidence, and where the gaps are. That’s your audit brief: hand it to whoever owns your case studies, BIM library, technical documents and trade-press outreach as the priority list for the next quarter. The answer to “why aren’t we getting specified in category X” was sitting in a prompt the whole time.
Want more like this? Our monthly newsletter covers practical tips, real client stories, and what’s changing in construction sales and marketing: Sign Up Here.