Our view
Old rules, new surface
The spending happened. The return, for the most part, did not.
56%
of CEOs report neither increased revenue nor decreased costs from AI in the last twelve months. Only 12% report achieving both.
Source: PwC 29th Global CEO Survey, January 2026
78%
of UK businesses are now using AI in some capacity, yet the vast majority have yet to see any return on that investment.
Source: Studio Graphene / Censuswide survey of 500 UK senior decision-makers, 2026
79%
of organisations report facing challenges adopting AI, and only 29% see significant ROI from generative AI.
Source: Writer, Enterprise AI Adoption Survey 2026
It is tempting to read those numbers as evidence that the technology is overpromised. We think that is the wrong conclusion. The problem is not the models; it is that companies bought tools instead of changing how they are found. Internal efficiency projects are real, but they are not a growth channel, and almost none of them touch the question of whether a buyer encounters your brand at the moment they are deciding.
Meanwhile, a meaningful share of commercial research has quietly moved into assistants. People ask which supplier to use, which platform handles a given requirement, which firm is credible in a particular sector — and they receive an answer that names two or three companies. Visibility inside that answer is a distribution channel. Almost nobody is treating it like one.
Nothing was thrown away
There is a widespread assumption that generative search represents a clean break — that the last two decades of search practice were made obsolete the moment assistants began answering questions in prose. That assumption is wrong, and it is expensive.
An assistant answering a question about your category is doing what a search engine has always done: identifying which sources are credible, which are relevant, and which can be summarised without embarrassment. The output format changed. The underlying judgement did not.
Structure is how machines read intent
Clear information architecture, consistent entity definitions, and correct structured data were never cosmetic. They are the difference between a machine understanding what your company does and a machine guessing. Assistants are considerably less forgiving of ambiguity than a human browsing a website, because they have no patience for navigation and no tolerance for a page that buries its answer beneath three paragraphs of positioning.
Most sites we audit fail here first. Not because the content is bad, but because nothing on the page states plainly what the business is, who it serves, and on what terms.
Authority is still earned elsewhere
Assistants weight sources that other credible sources reference. That is the same signal that has underpinned ranking since the beginning, and it remains largely outside your direct control — which is precisely why it works. Earned citations, credible third-party coverage, and consistent presence in the places your category is genuinely discussed will move visibility further than any amount of on-site optimisation.
Answer the question, then stop
Content that gets cited answers the question directly, in the first hundred words, in language a reader would recognise. It is specific, it commits to a position, and it can be quoted without qualification. Very little marketing content meets that bar, which is why so much of it never surfaces.
None of this is novel. It is the same discipline, applied to a surface that happens to be new.
Want this applied to your brand?
The first step is finding out where you stand right now. Start a conversation and we will tell you honestly what we see.
Start a conversation