AI search optimisation for scaling businesses: what actually works

Several, if not most, of our clients have started to flag the same problem. The digital metrics they used to rely on as performance indicators are no longer as helpful. Sometimes website traffic is down while sales or enquiries hold steady, or even improve. Sometimes traffic is stable while sales and revenue are down. Either way, what the correlation between what your website reporting says and what the business is actually doing performance-wise have come apart.

This is not a reporting glitch in the matrix. In 2026, a growing share of how buyers find and evaluate brands happens inside AI tools that analytics cannot see. People are asking ChatGPT, Perplexity and Google’s AI Overview the questions they would once have typed into a search box and getting back a shortlist of recommended brands. Users who receive a recommendation from a large language model are around 2.5 times more likely to visit that brand’s website. But roughly 55.9% of those visitors arrive by typing the brand name in or searching for it, rather than clicking a referral link. 

That blind spot cuts both ways. If your brand is the one AI tools are recommending, you may see fewer casual sessions but much higher-quality ones: visitors arrive at the end of their research, view roughly twice as many pages and stay twice as long, because the evaluation is done. Sales or pipeline holds up, or improves, even as generic informational traffic falls away. If a competitor is being recommended instead, the mirror image applies. Traffic looks steady because low-intent visits are padding the number, while the ready-to-buy audience is going to whichever brand the AI selected.

What is actually happening

Buyers are no longer browsing lists of blue links to compare options, whether they are B2B decision-makers building a shortlist or B2C consumers researching a category. They are delegating that work to AI assistants and answer engines. Those tools do not rank pages one at a time by keyword. They read across the web and return a shortlist of four or five brands they judge best answer the question.

Being ranked first in traditional Google search no longer guarantees you appear on that shortlist. The AI is drawing on the whole body of evidence about your brand: what analysts say about you, what reviewers say, what your customers say in industry forums, what your own site says. It forms a view long before anyone clicks anywhere.

A few numbers to illustrate the scale of this shift:

• Only around 1% of users click a link inside a Google AI Overview.
• Brands are routinely losing 20% to 30% of organic traffic to zero-click results.
• Over half of users now spend longer inside AI apps than they do in traditional search
• Traditional branded search volume has fallen by as much as 35% in a single year as queries move to LLMs.

The wrong way to respond to this change

When channels stop working, the default reaction in most marketing teams is a scramble for a checklist of fixes. Add schema markup. Reformat everything into Q&A blocks. Buy a handful of brand mentions on low-quality sites in the hope of pushing your name into more citations. This is the same short-termism that has damaged digital marketing for a generation: treating the symptom on the surface and leaving the underlying issue in place.

We firmly believe the true weakness is that a lot of brands now look and sound the same

Over the last decade, businesses drifted towards blander, more uniform digital content because it fed algorithms and proved a return per click. AI has made this problem worse, because any competitor can generate lookalike, good-enough copy and creative from the same training data at almost no cost. If your content, ads or website merely matches the industry average, that sameness is what gets you compressed into the summary and left uncited.

Structured data around undifferentiated content in particular changes nothing. A highly authoritative site that simply restates the category consensus doesn’t get summarised and cited. The pages that do get named tend to introduce their own data, their own case studies, their own frameworks. 

The technical implementation is just plumbing around the edges. Originality is what actually wins citations and ultimately the customer.

What actually drives AI visibility

Three things matter, in this order:

Being talked about elsewhere first

LLMs weight what independent third parties say about your brand far more heavily than what your own website says about you. That means reviews on sites like G2 and Trustpilot, discussions in industry forums and on Reddit, coverage in trade press, mentions in analyst reports and any presence you have on Wikipedia. Unlike traditional SEO, a mention without a hyperlink still carries weight, because AI is reasoning over the words themselves. If your category has a settled default set of four or five names and you are outside it, no amount of on-site optimisation gets you in. 

One EdTech scale-up we work with is building its PR programme around a proprietary careers dataset, because owning the source of a category-defining statistic is what earns the mentions that then earn the citations.

Being clear about what you do and who you serve

If a machine cannot instantly categorise your business, it is unlikely to consider you. Generalist businesses used to survive in a manual search world by buying attention or being close enough to the answer. 

AI is designed to remove risk and effort for the buyer and it filters undifferentiated brands out first. 

You need absolute clarity on three things: who you serve, what they value and why you are the right answer for them. Descriptions leaning on interchangeable adjectives like trusted, experienced or innovative give the machine nothing to work with and it will treat you as the average.

Being deep on something specific

Generic information is now effectively free and instantly available. The advantage runs to the brand that owns a specific, authoritative footprint on a well-defined topic, rather than a shallow presence across twenty. This can be more than one topic, but each one has to be genuinely niche. 

A B2B software business we work with, in a category where several larger horizontal players cover the same broad ground, has historically published content across the full range of what their platform can do. We’ve moved to do the opposite: identify the two areas of advanced functionality where a serious buyer would actually feel the difference against a horizontal competitor and make those the only territories the brand competes for online. Everything else stays supported inside the product but stops fighting for attention outside it. The commercial logic is that being cited as the specialist on the two things that matter most to their buyer wins more of the right shortlists than being one of ten names on a broader one.

Don’t be tempted to delegate this to one team to fix

The instinct in most businesses will be to hand this to the web team or an SEO agency, or ask marketing to produce more content. Both underestimate what has to shift and where the intelligence to shift it actually lives.

AI recommendations get won upstream of the purchase decision, in the questions your buyers ask before they ever enter the funnel. The people who know what those questions really are, what proof lands, and what a serious buyer will challenge sit in sales and account management and customer success as much as they sit in marketing. The people who know which capabilities are genuinely defensible, which are commodity, and which are worth investing in to widen, sit on the commercial side. Marketing owns the surfaces where the answers appear, and it cannot build the substance on its own.

The businesses we see doing well on this are treating it as a cross-functional programme rather than a marketing campaign. Sales feeds in what buyers are actually asking and what breaks their confidence in a shortlist. The commercial side is honest about where the real difference is, and where there is not one. Marketing turns that into the assets that stay useful: proprietary data, first-party case studies, credible third-party mentions, and a website that reads as the clear source of truth about what you do. One of our software clients has responded to this by commissioning a benchmark study on the operational metric that best distinguishes their product from the horizontal AI incumbents in their category, on the principle that a defensible difference has to be proved with numbers rather than asserted with adjectives.

The upside is that the moat compounds. Once you are the brand AI recommends, buyers arrive already convinced, sales cycles shorten, and top-of-funnel spend does more work on its own. A competitor cannot close the gap by increasing PPC or refreshing the homepage. They have to redo the strategic work underneath and that takes time and effort. 

What to do in the next 90 days

Start with the strategic question, not the technical one. What is the specific, defensible answer you want to be and what can you credibly claim that a competitor cannot? This can be more than one thing, but each thing has to be genuinely niche. Your existing positioning work is doing a new job here. It now has to be clear enough for a machine deciding who to cite and not only for a buyer reading your homepage.

From there, go and look. Ask ChatGPT, Perplexity and Google’s AI Overview the questions your buyers or customers would actually ask. See who gets named. If it is not you and a competitor is not genuinely more differentiated, that is your gap. If a competitor has earned the clearer position on merit, that is a more useful problem to know about now than in twelve months.

From there, three practical moves:

• Audit where your brand shows up on the wider web today: in reviews, in industry publications, in community discussions, in analyst commentary. Prioritise the credible third-party surfaces where AI tools are most likely to be reading and invest in earning presence there before you touch your own site.
• Identify the original proof you already have (customer outcomes, proprietary data, distinctive frameworks) and start publishing it in a form the web can actually cite. That means visible, plain HTML with clear structure, not gated PDFs or slide decks.
• Cut anything on your own site that is undifferentiated, me-too content. It is not neutral filler; it actively weakens the machine’s picture of what you specifically stand for.

You must resist the urge to be visible on everything. A brand that shows up vaguely across ten topics loses to one that is the unambiguous answer on one or two. This is a positioning decision before it is a content or SEO decision and it needs to sit with the leadership of the business rather than the marketing team on its own.

A few final thoughts 

Hopefully it’s obvious the answer to AI visibility sits with clarity applied across your own assets and distributed through third party proof. 

The three drivers earlier in this piece (being talked about elsewhere first, being clear about what you do and who you serve, being deep on something specific) all pull in the same direction and none of them are buildable on unclear positioning. 

Positioning is a multi-dimensional exercise. It means naming the sameness in your category as the commercial problem in its own right, mapping the whitespace on several axes at once (where trust actually forms, what the category has stopped competing on, which parts of the buyer’s journey no one owns, which proprietary assets you already have that a competitor cannot cheaply copy) and then choosing where to compete on that map rather than everywhere on it.

AI visibility now forces discipline: the machine needs the parameters by which it can tell you are actually different and it flattens anyone who has not given it any. If your positioning has been running on convention rather than conviction, that is where the work is and it is what decides whether AI recommends you or the business next door.

Luckily, this is exactly what we help businesses fix.