The complete guide to AI visibility audits.

When your customer asks ChatGPT, Claude, Perplexity or Gemini for a business in your category, are you the answer they get? For most companies in 2026, the answer is no. This guide explains what an AI visibility audit is, how to run one, what it should reveal, and what to do with the findings. Written for founders, not agencies.

01. What an AI visibility audit actually is

An AI visibility audit is a structured test of how often, and how favourably, a business appears when potential customers ask AI assistants for recommendations in that business's category.

The audit treats the AI assistant as a buyer would: it asks the questions a real customer would ask, captures the answers exactly as they are returned, and produces evidence of where the business stands today.

A good audit answers three questions:

• How often does the business appear when its customers ask the question that should surface it?
• When the business does not appear, which competitor does?
• What specific structural changes would change that?

It is the AI search equivalent of a search engine optimisation audit, but the rules are different. AI assistants do not rank URLs. They synthesise answers from sources they trust. Trust is built by a different set of signals than Google rankings.

02. Why this matters in 2026

According to Forrester research published in 2025, AI tools are now the most common channel buyers use to research purchases, ahead of vendor websites and analyst reports. ChatGPT alone reports more than 600 million weekly active users. Perplexity, Claude and Gemini collectively add hundreds of millions more.

The shift matters for two reasons. First, AI assistants give a single answer or a short list, not ten blue links. There is no second page. Second, the businesses AI assistants recommend are almost entirely different from the ones Google ranks first. Many businesses winning at SEO are invisible to AI search, and vice versa.

The implication is uncomfortable: a business can be the market leader in a category and still be invisible to the channel its customers are increasingly using.

03. How AI search differs from Google search

Google answers a query by indexing pages, matching keywords, and ranking results. AI assistants answer a query by synthesising a response from sources they consider authoritative. The mechanics are different in four ways that matter:

1. AI assistants break questions into sub-queries. A question like "best accountant in Manchester for a small SaaS" is decomposed into smaller searches for "Manchester accountants," "accountants for SaaS," and "small business specialists." The business needs to show up across all the sub-queries, not just the headline one.

2. AI assistants rank by mention frequency, not page position. A business mentioned by three independent sources is more likely to be cited than a business with a top Google ranking but no third-party coverage.

3. AI assistants prefer structured, fact-dense content. Long pillar pages with clear headings, lists, and direct answers outperform shorter pages with the same keywords. Schema markup helps the model parse who the business is.

4. AI assistants trust some sources more than others. Reddit, Wikipedia, established industry publications, and structured directories carry more weight than blog content, social posts, or paid placements.

04. GEO, AEO, LLMO: the terminology

The industry has not yet settled on a single term. Three terms are in common use:

• GEO (Generative Engine Optimisation) covers all the work of being visible in AI-generated answers.
• AEO (Answer Engine Optimisation) focuses specifically on appearing in direct answers from tools like Google AI Overviews and ChatGPT.
• LLMO (Large Language Model Optimisation) is sometimes used for the same work, with emphasis on appearing in the model's training data rather than its real-time retrieval.

The terms overlap and most practitioners use them interchangeably. What matters is the underlying work: making a business one of the sources AI assistants cite when answering buyer questions.

05. How to check your AI visibility in 15 minutes

Before paying anyone, a business owner can run a fast diagnostic themselves. The result will not be complete, but it will reveal whether the situation is critical or merely poor.

Open ChatGPT, Claude, and Perplexity in separate browser tabs. Ask each one the same five questions:

1. "Who are the best [your category] in [your city] right now?"
2. "Can you recommend a [your service] for a [type of customer]?"
3. "I am looking for [specific problem your business solves]. Who should I talk to?"
4. "What are the top three [your category] in the UK?"
5. "Have you heard of [your business name]?"

Record the answer to each. Count how many name the business by name versus how many describe categories without naming anyone.

This is the 15-minute version. A proper audit tests fifty or more buyer-intent questions across four assistants, captures screenshots, and analyses patterns. The fast version is enough to know whether the deeper work is justified.

06. What a full audit should include

A serious AI visibility audit produces evidence and direction. It should contain, at minimum:

Audits that miss any of these components are usually a thin lead magnet for a retainer pitch rather than a usable deliverable in their own right.

07. What an AI visibility audit costs

Pricing in the UK in 2026 falls into four bands:

Most small businesses do not need the enterprise tier. The productised tier is enough to identify the gap and build a roadmap that a single founder can act on without an agency.

08. How to fix the gaps once you find them

Three categories of work move the needle. Most businesses need all three, in order:

Own site signals

Publish authoritative pillar pages that directly answer the buyer-intent questions identified in the audit. Add structured data (Organization, Service, FAQ, Article schemas) so AI models can parse the business cleanly. Keep facts consistent: name, location, services, prices, hours. Inconsistencies confuse the models and reduce citation rate.

Third-party mentions

AI models trust patterns of mentions across independent sources. The work here is unglamorous but durable: getting listed in industry directories, contributing answers on Reddit and Quora that include the business by name, securing guest posts on established publications, and earning podcast or interview appearances.

Real-time presence

Perplexity and Google AI Overviews retrieve information at query time rather than from training data. They favour fresh content. A regularly updated blog, an active Reddit profile, recent press, and current public profiles all feed this layer.

09. How long it takes to show results

Timelines differ by mechanism:

• Perplexity and AI Overviews: changes can appear within days to weeks because these systems pull live content.
• ChatGPT and Claude: citations from training data refresh over three to six months as new model versions are released.
• Compound effect: the biggest gains come at month four or five once both training-data and real-time signals reinforce each other.

Anyone promising AI visibility "in two weeks" is either limiting the work to Perplexity or overstating what is realistic.

10. The four mistakes most businesses make

1. Treating AI visibility as an SEO project. The disciplines overlap but do not replace each other. Keyword stuffing, link farms, and thin landing pages that scored short-term Google wins are exactly the signals AI models discount.

2. Optimising for the home page only. AI assistants synthesise from many sources. The home page is one source among many. A single pillar guide often outranks a polished home page for AI citations.

3. Ignoring Reddit and Quora. These platforms are heavily weighted in training data and real-time retrieval. A handful of substantive answers signed by the business owner over six months can shift visibility measurably.

4. Measuring once. AI models update. Competitors publish. Visibility today does not predict visibility in three months. The businesses that win run a fresh audit every quarter.

11. Frequently asked questions

What is the difference between an SEO audit and an AI visibility audit?

An SEO audit tests how a business performs on Google's ranking systems. An AI visibility audit tests how a business performs as a cited source for AI assistants. The two share some signals (authority, fresh content, structured data) but diverge on others (AI models weight independent mentions heavily; Google weights link profiles).

Do I need a separate audit for each AI model?

No. A useful audit tests the same query set across all four major assistants (ChatGPT, Claude, Perplexity, Gemini) because the underlying work of being a trusted source improves all of them. The audit reveals which assistant is your weakest, not which one to optimise for in isolation.

Will my business eventually appear in ChatGPT on its own?

Possibly, if your business gets organic mentions across the kinds of sources AI assistants weight heavily. Most do not. The audit identifies whether a business is on that trajectory or invisible to it.

Is this a one-off cost or an ongoing one?

The audit itself is one-off. The work to act on it can be done internally or with help. Curiosity Limited offers the audit as a fixed-price product, with optional implementation help on request.