What is AI visibility?

AI visibility is how often, how accurately and how favourably an AI assistant mentions your brand when someone asks a question your business could answer. It is the equivalent of a search ranking for a world where the answer is generated rather than listed — and where most people never click through to a website at all.

AI visibility vs. SEO

Search engine optimisation optimises a URL for a position on a results page. AI visibility optimises a brand for inclusion inside a generated answer. The two overlap but are not the same thing, and a brand can be strong at one and absent from the other.

• Unit of measurement. SEO tracks page positions. AI visibility tracks brand mentions, the order competitors appear in, and the sources cited.
• Where the answer comes from. A search engine returns indexed pages. A model blends training data, live retrieval and third-party sources — so a roundup post on someone else's site can matter more than your own homepage.
• Visibility of the result. Rankings are public and checkable. AI answers are personalised, non-deterministic and different per model, so they have to be sampled deliberately.

The three things AI visibility depends on

In practice, whether an assistant names your brand comes down to three separate questions, and a failure in any one of them is enough to keep you out of the answer.

1. Can agents reach and read you?

AI crawlers such as GPTBot, ClaudeBot, PerplexityBot and Google-Extended have to be allowed in robots.txt, and your content has to exist in the HTML rather than only appearing after JavaScript runs. An llms.txt file and a clean sitemap give models a map of what matters on your site.

2. Can they understand you?

Structured data — Organization, Product, Offer, FAQ schema — turns prose into facts a model can quote without guessing. Clear pricing, availability, specifications, shipping and return policies published as data rather than buried in a PDF do the same job.

3. Does anyone else say it?

This is the one most brands miss. Models weight third-party corroboration heavily. Category roundups, comparison articles, review platforms, directories, community threads and press coverage are the sources assistants tend to cite when recommending vendors. If nobody outside your own domain describes what you do, there is very little for a model to retrieve.

How to measure AI visibility

1. Build a prompt set. Write 20–50 questions a real buyer would type, in their words, not yours — category questions, comparison questions, problem questions, and a few branded ones as a control.
2. Run them across models. The same prompt gives different answers on ChatGPT, Gemini, Claude, Perplexity and Copilot, so a single model tells you very little.
3. Score three dimensions. Mention rate (were you named at all), position and share of voice (who was named instead), and sentiment or accuracy (was what the model said about you correct).
4. Record the citations. The URLs a model quotes are your action list: they show which third-party pages are shaping your category.
5. Repeat on a schedule. Answers vary run to run. Only a repeated prompt set on a fixed cadence separates a real trend from noise.

What actually improves it

• Fix the readability basics first: allow AI crawlers, publish llms.txt, serve content in HTML, and add Organization, Product and FAQ schema.
• Publish the page types models quote: clear definitions, comparisons, category roundups and direct answers to real questions — not brochure copy.
• Earn third-party mentions: review platforms, category directories, roundups, podcasts, partner sites and genuine community discussion.
• Keep your facts consistent everywhere. Contradictory pricing or positioning across sources makes a model hedge or skip you.
• For commerce, go beyond being described to being usable: product feeds, an MCP endpoint and agent-readable checkout signals let assistants act, not just mention.

The deeper playbook lives in how to improve your brand's visibility in AI search.

Frequently asked questions