Your AI visibility tool found the gap. Closing it is a media decision.

An AI visibility tool tells you something true. Ask it how the engines answer the questions your buyers ask before they choose, and it comes back with the brands that get named, the fact that you are not among them, and the pages each answer was assembled from. That screen is worth paying for. It is also where every product in the category stops.

What the software hands you next is a list of fixes for your own website, and that is the wrong end of the problem. The gap it found lives inside articles published by people you do not employ, on sites you cannot edit. Closing it means putting a claim you can stand behind in front of the journalists who write them, which is a commissioning decision made on a media budget, not a ticket for whoever owns your content management system.

The score is the alarm, the source list is the brief

The software hands you two things, and most teams read only the first. There is a score, some version of how often you turn up in the answers being tracked, and there is a list of the publications those answers were built from.

Sit with the second one for a minute. A piece of software has just told you, from live answers rather than from a media pack, which publications decide what an assistant says about your category this month. A planner would normally assemble that list by argument over a quarter and still be talked out of half of it. Our own AI visibility report ends on the same list, for the same reason: it is the only output that tells you where to spend.

Your website is the only thing a software company can task you on

So every recommendation points at your own domain: question-shaped headings, cleaner markup, a machine readable file at the root, more pages on the topics you want to own. Some of that is worth doing on its own merits. None of it reaches the pages the answer is made of.

Split the questions in two and the reason shows. Your own pages carry the answer when a buyer types your name, and that is the question you had already won. On the question that decides a shortlist, where somebody describes a problem and asks who solves it, your domain is close to absent, and the pages standing in its place belong to publishers, trade titles and review sites. Every fix on that list is work inside the half you had already won.

Narrow the ask to one topic before you approach a publisher

Treat it as a shortlist of partners rather than a finding in a report, then narrow the ask. A brand message that covers everything you sell gives a publication nothing to write and a model nothing to quote. One topic, the one your buyers ask the most questions about and the one where your answer holds up best against a rival's, gives both of them something to work with.

The mechanism itself is old. A publisher's own writers produce a piece to your brief, under their editorial rules, and it runs on their site as their journalism. Branded content has worked that way for years. What is new is where the article ends up, because the model summarising your category reads that publication too, and it reads it every time the question comes round.

That changes the shape of what you buy. One article in one publication is a placement, and a planner already knows what a single placement does. The same claim carried at once by titles with no connection to each other is a position, and only the second version gives a model a reason to treat it as where the category has landed. So budget it as a burst across publishers rather than a line on one, expect it to sit on the media plan and not inside a search retainer, and brief the titles together, which is what one brief to many partners is for.

Most considered categories have no incumbent in the answer yet

In most considered categories it is not settled yet. Semrush with Kevin Indig, in Search Engine Land, July 2026, read 1,094 categories in ChatGPT between January and June 2026 and found that 15.2% of categories have a brand the models clearly recommend. The other 85% are unowned. That sample covers the United States, so take it as evidence the ground is open rather than as a count for your own market. Read the analysis.

An assistant summarising a topic can only work with what has been written about it, and in travel, insurance, appliances and B2B software, very little has been written in the form a model can quote: specific, attributable, comparable against a rival. There is no incumbent to displace. There is an absence.

Absences fill. They fill the moment a competitor supplies the checkable claim an assistant can repeat, and from then on it is the one on hand every time the question is asked. That is the whole of the timing argument and it is the only urgency in this worth acting on. Almost nobody in your category is being outranked today. Somebody is about to be first.

Set the measure before a word is commissioned

Measuring it is the same discipline that has always separated the content that gets renewed from the content that gets queried, and what we wrote about setting the measure before the campaign starts applies here without amendment.

Two rules keep the measure honest. Track the same questions before and after, phrased the way a buyer would phrase them. And never accept one blended visibility number as the result, because a blended number averages the questions that name you, where you were always going to appear, with the open questions where a shortlist is decided. It moves when the easy half moves. Ask for the open questions on their own. Our report reads the four kinds of question separately for that reason. There is no single figure to hide behind.

No one controls what a model says, so no honest supplier will promise you a mention in it. What can be stated is movement: these are the questions, this is how the brand appeared on them before the work started, this is how it appears now. Trust built through third-party editorial is not a new outcome for this channel, only a newly measurable one.

Get the source list, then commission against it

Every decision you take before you have that source list is a guess about which publications matter in your category. Our free AI Visibility Report puts your buyers' real questions to the engines and returns the brands they name, the language used about each one, and the publications each answer leans on. Then the part software cannot do for you. We commission editorial across a network of more than 600 trusted media brands in Australia, the UK and the United States, so the titles on that list are ones we can book.