On 1 September 2026 the Crane Index™ scored 2,291 UK websites and only 93 cleared 70 out of 100. That leaves the top of almost every industry table empty, so an ordinary, attentive company can lead its field on the measure AI engines use to read a business right now, while the work is still cheap.
Only four sites in a hundred read cleanly enough to be named with confidence
A score of 70 on the Crane Index means an AI engine can read a site and state its products, prices and terms back in a sentence without guessing; below 70 it hedges or drops the business when a buyer asks it to recommend someone. Across the 2,291 sites the Index could read on 1 September 2026, 93 cleared that bar.
This is a retrieval question before it is a marketing one, and the evidence says structure decides it: Meltwater's May 2026 analysis of 9.5 million AI citations across six models found that large language models lift structured, expert-led, specific passages as direct answers rather than whatever ranks highest. What drags a score down is rarely exotic: a price that lives only inside an image or a PDF, opening hours that disagree between two pages, a product range written in brochure language a machine cannot parse. The fix is structured data and clear entity markup of the kind Google itself documents, so a machine can read the fact and trust it.
Every count here is out of the sites the Index could actually read, never the number approached. A site a machine cannot read at all gets no score, not a low one. So these 93 are the best of the sites already legible, and even among those, clearing 70 is uncommon.
The biggest budgets are not the ones AI reads most cleanly
None of the 40 FTSE 100 sites the Crane Index could read on 1 September cleared 70, and none of the 77 readable UK Ecommerce 100 sites did either. The companies with the largest sites, the biggest teams and the deepest budgets are not the ones AI engines read most cleanly. Clean machine reading is a different discipline from spending, and those budgets have not been pointed at it.
In most fields, first place is a contest between two names
In legal services, 2 of the 133 readable sites cleared 70. In transport and logistics, 2 of the 110 readable sites did. In fields of well over a hundred real competitors, the number reading cleanly enough for an engine to name them with confidence is two.
Ask an assistant which firm handles a dispute like yours, and it can only name a firm whose site states its practice areas, fees and jurisdictions in machine-legible terms, ideally with FAQ and service markup. Ask which carrier can move a particular load, and it can only shortlist a site that states its routes, load types and coverage as structured facts rather than sales copy. In most fields that shortlist is drawn from a handful of names, because a handful is all that has done the work.
Technology and software proves the score is learnable, not luck
Technology and software is the exception: 17 of its 123 readable sites cleared 70. That settles a question the rest of the market keeps asking. A high score is not luck and it is not gated by budget. Software got there first because stating what a product does in structured, machine-legible terms, using schema types such as SoftwareApplication and clean API documentation, is close to how it already documents itself.
The window is open because the work is new, and it will close as the field catches up
The scores are low because the work is new. Most sites were built to be read by people, and by Google's old ten-blue-links index, not by an engine that has to state your products and terms back in a sentence. The structured-data and entity work that lets a machine do that cleanly is deliberate and largely undone. That is why an ordinary business can lead: the leader's seat in most fields is empty because almost nobody has sat in it. Software has already started closing its own window. The advantage goes to whoever does the work while the seat is still empty.
The decision is whether to take the lead while it is still cheap
Find out where you score on the measure AI engines use, then decide whether leading your field is worth taking now. This is AI search work, closer to engineering than marketing: making your pages state, cleanly and consistently, who you are, what you sell and on what terms. You do not need the biggest budget in your field to win it, because on this measure the biggest budgets are not winning. You need to decide it matters, before the two names already reading cleanly become twenty.
