AI search optimisation: how to make sure AI assistants recommend your business

To get recommended by AI assistants, four things matter. Your content should answer questions directly, without padding. Your site should publish structured data that machines can parse. Your business should carry authoritative signals across the web, reviews, mentions, citations. And you should consider publishing a llms.txt file that makes your best content discoverable to language models. Do those four things and you show up where it counts. Ignore them and the next generation of search will quietly route around you.

This is a practical article on what AI search optimisation, sometimes called GEO, generative engine optimisation, or AEO, answer engine optimisation, actually is, and what a UK small business should do about it.

Why this matters now

In 2024 and 2025 the way people find things online shifted. More enquiries now begin with a question typed into ChatGPT, Claude, Perplexity, or Google's AI Overview than with a traditional search. The user no longer scans ten blue links. They get a synthesised answer, often with one or two sources cited. If you are cited, you get traffic. If you are not, you do not, regardless of where you rank in classic search.

This is not a replacement for traditional SEO, which we cover in our plain English guide to SEO for small businesses. It is an addition. Google still drives meaningful traffic for most UK small businesses, and will continue to. But the trend is clear and early movers are being rewarded.

How AI assistants choose what to recommend

Each AI assistant has its own quirks, but the pattern is similar across all of them.

They pull from training data, a snapshot of the web from some cutoff date. They pull from live search, a real-time query run when the user asks something the model is unsure about. They cross-reference both. They then synthesise the answer, citing what they consider the most credible and relevant sources.

What they value.

Clarity and direct answers. A page that says, in the first paragraph, exactly what the user asked, beats a page that eases into the topic with three paragraphs of waffle.

Structured content. Clear headings, bullet lists where genuinely appropriate, short paragraphs, definitions stated plainly. Machines parse structure more easily than prose.

Authority. Mentions on credible third-party sites. Reviews. Citations. A domain that has been around. An author with evidence of expertise.

Freshness. Articles with recent dates, referring to current events or data, get preferred for queries about current topics.

Specificity. Generic content is filler. Specific examples, data points, and detail are the fuel that gets your page quoted, which is the same standard we argue for in how AI is changing small business marketing.

The four things to do

Answer questions directly in the opening paragraph

The fastest single change you can make to any article is to put the answer first. A cooking blog that spends five paragraphs on the origin of risotto before telling you how to make it is the archetype of what AI search will not cite. A page that opens with, the best rice for risotto is carnaroli, arborio is acceptable, then explains why, is what gets pulled into answers.

Rewrite your most important pages so that the first paragraph delivers the substance a user asked for. The rest of the article can elaborate, provide context, and cover edge cases. But the direct answer must be right at the top.

The test. If a user only read the first paragraph, would they have the answer? If yes, you are aligned with how AI search works. If no, rewrite.

Publish structured data on every meaningful page

Structured data is machine-readable markup on your web pages that tells systems what the content is. A product page with product schema. An article with article schema. A business with local business schema. An FAQ with FAQ schema.

Most UK small business websites have some structured data already, added by Yoast, Rank Math, or a similar plugin if you are on WordPress. Shopify and Squarespace handle the basics by default. Check using Google's Rich Results Test, paste a URL, and see what the tool identifies.

The most important schemas for small businesses.

• Organisation or LocalBusiness on your homepage and contact page, with name, address, phone, opening hours, and geographic coordinates.
• Service on each service page, with description, provider, and area served.
• FAQ on pages that answer multiple related questions.
• Review and AggregateRating where you have genuine reviews, not fake ones.
• Article on blog posts, with author, date published, and date modified.

If your site is custom built and lacks structured data, this is usually a four to six hour job for a developer. The return is disproportionate to the effort.

Consider a llms.txt file

llms.txt is an emerging standard, similar in concept to robots.txt or sitemap.xml, that gives AI crawlers a curated list of your best content, in a format designed for language model consumption.

It lives at yourdomain.co.uk/llms.txt. The format is plain Markdown. A typical one lists your site's purpose, your most valuable pages, and clear descriptions of each. Some businesses also produce a llms-full.txt that contains the full text of those pages, so the model does not need to crawl them.

Adoption is growing but not universal. Anthropic, the makers of Claude, have publicly supported it. Other AI providers have been quieter but are watching. It costs almost nothing to publish one and the downside is zero. It is worth doing.

A minimal example.

# Voll

> A UK consultancy helping small businesses tune operations, marketing, and technology.

## Services

- [Business audit](https://voll.co.uk/resources/audit): free health check of your operations.
- [Consulting](https://voll.co.uk/consulting): deep engagements to fix specific problems.

## Insights

- [CRM for small businesses](https://voll.co.uk/resources/insights/crm-for-small-businesses): when you need one and how to choose.

Simple. Machine-friendly. Curated by you.

Build authoritative signals off-site

AI assistants trust what other credible sources say about you. This is the same principle that drives traditional SEO, but the weight placed on signals is often higher in AI-generated answers, because the models are trying to avoid recommending businesses that might not be real.

Four signals that matter.

Google Business Profile. Claimed, complete, and with recent reviews. For any UK business with a local footprint, our complete guide to Google Business Profile for UK small businesses is non-negotiable.

Reviews on third-party platforms. Trustpilot, Google, industry-specific directories, relevant professional bodies. Genuine reviews, with real customer names, carry weight, and how to get more Google reviews walks through the ask-at-the-right-moment playbook.

Mentions in news and industry sites. Getting quoted in a BBC article, a trade publication, a local paper, or a recognised industry blog matters more than a hundred social posts. One good PR placement per quarter is a realistic goal.

Consistent business information everywhere. Your name, address, phone, and website details should match across every directory, every platform, every mention. Inconsistencies make automated systems distrust you.

What to avoid

Three temptations to resist.

Stuffing structured data with unverified claims. Adding aggregate rating schema with fake review counts is a quick way to get penalised by both Google and AI systems, whose trust algorithms look exactly for this kind of gaming.

Generating hundreds of AI articles to cover every possible query. AI systems detect low-quality mass-generated content. Hundreds of thin articles hurt you. Thirty well-written, specific, updated articles help you.

Chasing every new optimisation fad. The fundamentals, clear answers, structured data, authority, freshness, will remain valuable even as tactics change. If a consultant is pitching you an AI search optimisation secret that contradicts those fundamentals, be cautious.

How to measure progress

Measuring AI search performance is harder than classic SEO. The main indicators.

Branded search volume. If more people search for your business name, something is working. Use Google Search Console.

Referral traffic from AI assistants. Some traffic from ChatGPT, Perplexity, and Claude now appears in analytics, usually tagged as chatgpt.com, perplexity.ai, and so on. Monitor this trend over time. It is low today for most UK small businesses but growing.

Direct user feedback. Ask new enquiries how they found you. If one in five says, I asked Claude for a recommendation, you know the work is paying off.

Test queries. Once a month, ask ChatGPT, Claude, and Perplexity the types of questions your customers would ask. See whether your business appears in the answers. Track it in a simple spreadsheet.

The timeline to expect

Like any search work, AI search optimisation rewards consistency, not heroics. Expect two to four months before meaningful change shows up in test queries, and six to twelve months before it shows up in enquiries. Most businesses who try and give up do so at the three-month mark, exactly when they should have been hitting second gear.

The UK-specific angle

UK small businesses have two structural advantages in AI search.

First, local intent is strong. Queries like best accountant in Exeter or trustworthy plumber near Guildford are the kind AI models work hard to answer well, and the sources they pull from reward genuine local authority. A strong Google Business Profile, good reviews, and a clearly described service area can outperform a glossy national competitor for local queries, which is exactly the playbook in how to rank at the top of Google Maps.

Second, the noise is lower. The UK market has been slower to industrialise AI-generated content than the US, which means genuine expertise still cuts through more cleanly. This window will close over the next year or two. Move now.

The one thing worth starting today

If you do one thing after reading this, rewrite the opening paragraph of your three most important pages so they answer the core question directly. That is a two-hour exercise that shifts how AI systems read your site.

If you want a quick look at your current setup and where AI search readiness stands, our business audit covers it as part of the wider review. Or book fifteen minutes and we will look at yours directly.