Earlier this year we ran a 90-day GEO programme with a UK security firm. They had a strong service offering, dozens of happy clients, and a Google ranking position that hovered around page two for their most important commercial keywords. When we ran the initial AI-visibility audit, the picture was bleak: zero citations across ChatGPT, Perplexity, Claude or Google AI Overviews for the buyer-intent queries that mattered most. Competitors were named confidently in every one of them.
The diagnostic was illuminating. The firm's website was professional but vague about the specifics. The schema markup was thin. There was no consistent description of the business across the web. Articles, where they existed, were generic and unattributed. The entity profile was, to AI eyes, a blur.
We rebuilt the foundations first. Tightened the canonical description of the firm. Implemented full Organisation and LocalBusiness schema, including credentials, areas served, and service lines. Got the language consistent across the website, the social profiles, and the major directory listings.
Then the engine started. Ninety AI-optimised articles a month, briefed from the citation gaps the audit had identified. Each one engineered for entity clarity and citation eligibility. Each scored against a 100-point internal quality bar before publishing.
The first citations landed in week three — Perplexity surfaced the firm for "construction site security companies UK" with a direct quote from a published guide. By day 45, the firm was being cited across three engines. By day 90, it was named in answers to seven of the top-ten buyer-intent queries we'd identified at the start.
The traffic followed, but the more meaningful result was the enquiry quality. Inbound leads now arrive citing AI by name: "ChatGPT recommended you for commercial security." Those leads convert at materially higher rates than cold inbound did before.
Three lessons stand out. The first is that entity clarity does more heavy lifting than people expect — most of the foundation work paid back inside thirty days. The second is that velocity compounds: each cited article makes subsequent citations easier, because the model's confidence in the entity grows. The third is that citation rate is the right metric to manage to. Rankings and traffic followed automatically once that metric was moving.
We are now running variations of the same playbook across other UK security firms. The category is competitive, but AI search is still early enough that determined entrants can win meaningful share. Two years from now, that window will be much narrower.