What AI search visibility actually means for a UK SME

Ask ChatGPT to recommend an accountant in Leeds, or a roofer in Coventry, and it will name a handful of firms. It will not name most of them. The businesses it does name did not necessarily pay for that placement, and in most cases they did not deliberately earn it either — they simply happen to have the characteristics these systems draw on.

Those characteristics can be worked on. That is what AI search visibility means.

Why it is not just SEO

Traditional search optimisation aims at a ranked list of ten blue links. The mechanics are reasonably well understood: relevance, authority, technical health, links. AI-generated answers work differently in three ways that matter commercially.

1. There is no list — there is one answer

Position eleven in Google still gets some traffic. Not being named in a generated answer gets none. The distribution is far less forgiving.

2. Corroboration matters more than optimisation

A model does not rank your page. It builds a representation of your business from many sources — your site, directories, press coverage, Companies House, review platforms — and names you when those sources agree that you exist, do the thing being asked about, and operate where the asker is. Inconsistency between sources is one of the most common reasons a business is not named.

3. Structure determines whether you can be quoted

These systems extract passages that stand alone. A page that opens with a direct answer, defines its terms, and cites dated figures is quotable. A page that opens with "In today's fast-moving digital landscape…" is not.

The most common cause of invisibility

It is unglamorous: many websites block the crawlers. Over the last two years a great deal of advice circulated recommending that sites disallow GPTBot, ClaudeBot, PerplexityBot and Google-Extended in robots.txt, on the reasoning that models should not train on your content for free.

That is a legitimate position for a publisher whose product is the content itself. For a business whose product is a service and whose content exists to attract customers, blocking those crawlers means opting out of being recommended. Many businesses have done it without realising, because it arrived in a plugin default or an agency's standard configuration.

This is worth checking today. Visit yourdomain.co.uk/robots.txt and look for those user-agents.

What actually moves it

• Crawler access. Explicitly permit the assistants you want to be found in.
• Entity consistency. Identical business name, address and phone number across your website, Google Business Profile, Companies House, LinkedIn and every directory you appear in. Discrepancies make you harder to resolve as a single, real entity.
• Citation-ready structure. Direct answers in opening paragraphs, defined terms, comparison tables, dated statistics with sources, genuine FAQs.
• Presence in sources models read. Trade press, regional business press, industry directories. This is digital PR, and it has become considerably more valuable than it was.
• Published specifics. Prices, service areas, sectors served. Vague sites are hard to recommend confidently.

How to measure it

Write down thirty questions a potential customer might plausibly ask — "who are the best commercial electricians in Birmingham", "which agency should I use for B2B SEO in the UK" — and run each through ChatGPT, Gemini, Perplexity and a Google search that triggers an AI Overview. Record whether you are named, how you are described, and who appears instead.

Repeat monthly. That is the baseline, and without it any claim of improvement is unverifiable.

How long it takes

Technical corrections land immediately. Changes in what assistants say typically take three to six months, because the underlying corpus updates slowly and corroboration accumulates gradually. Anyone offering faster is describing something else.

The honest summary: this is early, the mechanics are less settled than SEO, and nobody can guarantee a third party's model will name you. What can be guaranteed is measurement, correction and reporting — which is considerably more than most businesses have today.