Named beside
the Big Four.

In week two the client asked the right question: why does it look different for you and for me? Two people had searched the same phrase minutes apart and got different answers, one with the firm in it and one without. It is the most common reason a brand stops trusting its own AEO reporting.

Google personalises. It weights your location, your account history, the pages you have already visited. Search your own brand while signed in and you will nearly always find yourself, because you have spent years teaching Google that you are interested in you. Your own team's phones are the one place the answer is guaranteed to look good, and the one place it proves nothing.

So the first thing we build for a client is not a content plan. It is an Ideal Customer Profile, agreed with them, precise enough that any two people reading from it land on the same answer. For Steadfast it looked like this.

A larger rival started publishing comparison content: the kind that names competitors in order to be the page that ranks for comparing them. Within days it was the cited source behind almost every bullet of one head query's AI Overview, and Steadfast, which had been named there a week earlier, was reduced to a citation.

The correction was a change of intent, not of volume. The first three weeks answered questions (how a method is chosen, which databases a TPO accepts) because that is what earns a citation from nothing. Once citation was consistent, the next batch answered comparisons: best firms, most trusted consultants, who to pick for a mid-size group. That is where the buyer actually decides, and it is contested ground by definition.

The firm's own Search Console, comparing the 28 days to 5 September against the 28 before. AI Overview impressions moved from 17.1K to 22.6K; site-wide clicks from 1.04K to 1.27K and impressions from 77.8K to 95.1K. Average position barely moved, 10.2 to 10.4. The growth is new queries being entered for, not old ones climbing.

One caveat we would rather state than have you find: content was only live for 19 of those 28 days, so the “before” window is cleaner than the “after” one is complete. The true run-rate effect is larger than these percentages, not smaller, but it is also measured over less time than a month implies.

The instruction that made this engagement work came from the client, on day one of content: everything on transfer pricing, nothing else. A professional-services firm with ten service lines and a blog that touches each of them lightly is invisible in all ten, because AI answers are assembled per topic, not per brand. Concentrate a month of publishing on one practice and that practice becomes answerable.

Firm size turned out not to be the constraint. Steadfast is a fraction of EY's size and appears in the same sentence, because the model was not ranking firms. It was assembling an answer from whatever had actually addressed the question, and nobody had. That is the opening, and in a specialist category it is still wide open.