Why Expert Consultants Are Invisible in AI Search

You're brilliant at what you do. But it's not enough.

The best consultants are, almost without exception, the worst at marketing themselves - myself included. It's not a coincidence. It's a structural problem. And in the age of AI search, the cost of not doing it has just gone up.

The short version. Consultants underinvest in their own visibility for entirely understandable reasons. The same qualities that make someone genuinely good at their work - intellectual rigour, client focus, discomfort with self-promotion - tend to produce a business that is trusted by the people who already know them, and invisible to everyone else. In the verdict economy, where AI assistants now pre-select who makes the buyer's shortlist, that invisibility is no longer a minor inconvenience. It is a structural commercial risk.

Why are expert consultants invisible online?

Even if you understand the structural reasons why expert consultants stay invisible online (the modesty instinct, the 'when I have time' delusion, the platform problem) and the changes in B2B buyer journeys, there’s an even deeper problem. Most people think the ‘fix’ is to produce more content, show up more, optimise harder. It treats a reputation problem as a volume problem. And volume without authority, in the verdict economy, is noise.

The answer lies in understanding how AI search actually works and then recognising that what it rewards is not what most founders think.

AI visibility is not a position. It is a probability.

Alex Birkett, co-founder of Omniscient Digital, wrote something earlier this month that I think is the clearest framing of this I have come across. He describes AI search visibility as 'probability engineering': the task is not to hold a rank or occupy a position, but to increase the likelihood that, at any given moment, across any given AI engine running any given query, your name is in the answer.

The reason the framing matters is that AI search is structurally different from traditional search in a way that most founders have not fully absorbed. Traditional search had a deterministic quality: a keyword had a volume, a page held a position, and the relationship between action and outcome, however noisy, was legible. You could build a dashboard around it.

AI search works differently. The outputs are sampled from a probability distribution. SISTRIX tracked 82,619 prompts across three platforms and six countries over 17 weeks and found that Google AI Mode rotates 56% of its citation sources every single week, and ChatGPT rotates 74%. AirOps found that only one in five brands maintains citation visibility across five consecutive runs of the same query. The University of St. Gallen and Aurora Intelligence confirmed in peer-adjacent research published in April 2026 that even same-day repeated runs produce only 32% to 43% source overlap, confirming that the instability is inherent to how the models generate answers, not a temporary quirk.

There is no position to hold. There is only a distribution to influence.

The query fanout problem makes this more complex still. When a buyer types a prompt into ChatGPT, the model does not simply retrieve results for that prompt. It internally generates multiple sub-queries of its own before synthesising an answer. Surfer SEO research found that only 27% of these internally generated sub-queries remain consistent across multiple runs, with 66% appearing exactly once and never again. You are not optimising for a keyword. You are trying to appear across a volatile, stochastically generated set of questions you cannot directly observe.

The Pólya urn and why now matters

Birkett brilliantly uses the Pólya urn model (I too learn new things every day!) to explain why this creates urgency.

In a standard probability experiment, you draw a marble from a bag, note its colour, replace it, and repeat. Over enough draws, you converge toward the true distribution. In the Pólya urn, the rule changes: each time you draw a marble, you replace it along with one additional marble of the same colour. Early draws compound. A colour that appears early becomes disproportionately more likely to appear again. As Birkett put it: small initial advantages become structural ones.

This is what is happening in AI search right now. “AI visibility is a byproduct of genuine brand construction, not technical optimisation.”[1] The brands that appear in frequently cited sources get trained into models, retrieved more often, cited more, and become harder to displace. The distribution is forming. The window to influence it is not permanent.

For founders who have deferred visibility work, this is the uncomfortable implication: the opportunity cost is not just this quarter's pipeline. It’s the loss of a structural position in the emerging citation landscape. Every week of inaction is a week in which other marbles are being added to the urn.

Why real-world credibility is what shifts the odds

Here’s the part that surprises most founders, and that no amount of technical optimisation can shortcut.

AirOps' analysis of commercial AI search found that 85% of brand mentions in AI-generated answers originate from third-party pages, not the brand's own website. Brands are 6.5 times more likely to be cited through external sources than through their owned content. Nearly half of all AI citations come from community or user-generated platforms: forums, LinkedIn discussions, sector publications. Domain authority - the SEO metric that used to be a reliable proxy for credibility - shows no positive correlation with AI citation and is slightly inversely correlated. The always-cited pages have lower domain authority on average than the never-cited ones.

What the AI engines are weighting is not what you have published on your own site. It is the distributed evidence of your credibility across the places where your industry talks about itself. The speaking slots. The industry body memberships. The press coverage. The peer recommendations. The interviews, panel appearances, awards and the LinkedIn threads where someone else cites your thinking without being asked.

These are not extras. They are the primary inputs to the probability distribution you are trying to influence.

Birkett's framing makes this precise. The task, he writes, is to “build a dense, expansive web of proof points”, and crucially, most of that work is done for the sake of building the brand, not for AI search specifically. AI visibility follows as a consequence. You cannot manufacture it by optimising your way to it. You build it by becoming more real, more present and more coherent across the information ecosystem that AI models ingest.

This is not new advice. It is what reputation professionals have understood for decades. What has changed is the consequence of ignoring it: invisibility in AI search is now a commercial risk in a way it was not three years ago.

What this means in practice

Consistency in AI search is a lagging indicator of consistency in reputation. Which means the practical question is not 'how do I get cited by AI' but 'how do I build the kind of reputation that AI cites as a matter of course'.

For founders, that means four things in sequence:

1.    Positioning

The first is positioning: being specific and defensible about what you do, who you do it for, and why you do it better than the alternatives. Vague answers do not get cited. Specific, credible, well-evidenced answers do.

2.    Infrastructure

The second is online infrastructure: a website that accurately represents the work, a LinkedIn presence that is actively maintained, and content that demonstrates expertise rather than claiming it. Pages not updated within the last three months are more than three times as likely to lose AI citations as those that have been refreshed. A small number of well-maintained pages outperforms a large volume of stale content.

3.    IRL

The third, as the kids say, is IRL – in real life presence: the speaking slots, the association memberships, the press, the peer networks. Not because they are good for the soul, but because they produce the third-party validation that AI engines treat as evidence of authority. These are the marbles you are adding to the urn.

4.    Stamina

The fourth is sustained effort. The Pólya urn does not stop. The citation landscape continues to form. Reputable visibility is not a project to complete. It’s a rhythm to maintain.

The point is not to become louder for the sake of it. It is to make your credibility legible in the places where buyers, peers and AI systems now look for evidence. If you are already good enough to be recommended, the work is to ensure the market can see why - consistently, externally and before a buyer has already made their shortlist.

Frequently Asked Questions

Why do expert consultants underinvest in marketing themselves?

The same qualities that make a consultant genuinely expert - deep client focus, intellectual rigour, discomfort with overstatement - tend to produce a business that is well-regarded by the people who already know it and invisible to everyone else. The referral model works until it does not, and by the time it starts to slow, the founder has no established visibility to fall back on.

Does content marketing work for consultants and founder-led businesses?

Content works as part of a reputation strategy, not as a substitute for one. Buyer research consistently shows that peer recommendations and third-party validation outrank founder-produced content in the trust hierarchy. What content does well is demonstrate expertise over time, build the online presence that peer validation can point to, and signal to AI engines that the founder is a credible source worth citing. Volume without authority is noise.

How does AI search change the visibility problem for consultants?

AI assistants now pre-select the shortlist before a buyer picks up the phone. When a buyer queries ChatGPT, Claude, Perplexity, or Google's AI Overview for recommendations in a given category, the businesses that appear in the answer have a significant advantage over those that do not. AI engines select based on corroborating evidence of authority: published thought leadership, press coverage, peer validation, and a consistent online presence. Consultants who have not built that infrastructure are systematically excluded from shortlists they never know were being formed.

What does 'building visibility' actually require?

It requires four things in sequence: a clear, specific, defensible positioning statement; online infrastructure that makes that positioning findable and coherent; content and earned media that demonstrates expertise rather than claiming it; and real-world presence across the platforms, events, and peer networks where buyers gather and form opinions. The work is neither complicated nor quick. It compounds, and it needs to start before the pipeline stalls.

References

[1]Alex Birkett, Probability Engineering, Omniscient Digital, June 2026.