Skip to content
Matrex Analytica

What a new leader publishes in the first ninety days decides how AI describes them later

In its 2026 analysis of around five million cited URLs, Surfer found that domain authority had almost no correlation with how often a page was cited by AI engines. The same study reported that 38% of citations in Google's AI Overviews came from the first 100 words of a page, and that 67.82% of cited pages did not rank in the top 10 for the query or its fan-outs.

Read those three findings together and a simple point emerges for anyone stepping into a visible leadership role. The signals that cause an AI system to describe a person a particular way are not the ones a traditional launch plan optimises for. They are specific, textual, and set early.

Why the first quarter does more work than it should

A new leader arrives with a thin public record. Search engines and language models fill that gap with whatever they can find: an appointment release, a first interview, a couple of early posts, a bio on the corporate site. Those documents are small in number, high in prominence, and often written in a hurry. Whatever framing they carry becomes the default description that later content has to argue against.

This is not a claim that ninety days are magic. It is a claim about scarcity. When there is little to summarise, each new document is a large share of the total. Once a leader has been publishing for two years, one more post barely moves the summary. In the first quarter, it does.

What the research actually says, and what it does not

Surfer's work is platform-agnostic generative-engine research, not a study of executive-name queries. Applying it to a leadership transition is an interpretive step. With that caveat, three findings carry over cleanly.

First, domain authority is not the lever. A page on a modest site can be cited as often as one on a large publisher if its content answers the question directly. The corporate site, a personal essay and a trade column can compete for the sentences a model repeats.

Second, the first 100 words of a page do disproportionate work. In Surfer's sample, more than a third of AI Overview citations came from that opening block. A bio or announcement that buries the point under throat-clearing is not just weaker writing. It is a weaker signal.

Third, ranking is no longer the gate. Two-thirds of pages cited in Google's AI Overviews did not sit in the top 10 for the query or its fan-outs. The pages that get quoted are the ones that answer precisely, not the ones that won the old contest.

Surfer also cites Yext research, reported second-hand, suggesting that 86% of AI citations come from brand-managed sources. Treat the number as reported rather than independently verified, but the direction is consistent with everything else in the piece. The surfaces a leader controls are the ones models prefer to quote.

The practical move

If the first ninety days set the summary, the useful work is not another announcement. It is a small, deliberate stock of texts written to be quoted.

A short list is enough. One biography page on the corporate site that answers, in its first paragraph, who the leader is, what they are responsible for, and the two or three positions they hold on the questions their market cares about. One essay or interview that states those positions in full sentences a model can lift. One or two external pieces, in outlets a model already treats as reliable, that use the same phrasing. The consistency across those surfaces is the signal. The repetition is the citation.

Two cautions. The texts have to say something: a bio padded with adjectives is not quotable, because there is no claim to quote. And the phrasing has to be stable. If the corporate site, LinkedIn and a conference programme describe the same person three different ways, the model has to pick, and it will usually pick the shortest and least flattering version.

What this changes for the people around a new leader

For communications teams, the launch window is not primarily about coverage volume. It is about seeding a small number of precisely worded documents in the surfaces the organisation controls, and making sure the handful of external pieces that follow use compatible language. A press release that reads well to a journalist but contains no quotable sentence about the leader's position is a missed opportunity, not a successful placement.

For the leader personally, the first long-form piece is worth more time than the next ten. The draft that goes out in week six is still shaping the summary in year three.

For boards and investors, the question "what has the new chief executive said in public?" is a reasonable governance question in month four, not an impatient one. The public record being set now is the one the market will quote back later.

The honest limit

None of this is a guarantee. The Surfer research measures citations in general, not citations of executive names. The ninety-day frame is a reasonable reading of how thin public records fill in, not a measured inflection point. Models change, and the specific signals they prefer will shift.

What does not shift is the underlying mechanic. Systems that summarise people pick up the sentences those people, and the organisations around them, publish in the places they control. A new leader who treats the first quarter as an indexing event, not a goodwill tour, writes those sentences on purpose. The one who does not lets someone else write them by accident.

Source: Surfer, "9 ways to get cited by AI" (September 2026).

Source: https://surferseo.com/blog/llm-citations/