AI visibility

AI visibility is how often, how prominently and how accurately a brand appears in the answers that AI assistants give. It is measured by asking a fixed set of buyer questions on platforms such as ChatGPT, Perplexity, Gemini, Doubao or Tencent Yuanbao, then counting how often the brand is mentioned, recommended first and described correctly.

It is the AI-era counterpart of search visibility. In classic search, visibility means where your pages rank. In AI answers there are no rankings to check. The assistant writes a reply, and your brand is either part of it or missing.

Why AI visibility is its own problem

Strong search rankings do not automatically carry over. An assistant builds its answer from a mix of what the model learned in training and the pages it retrieves at answer time (its citation sources). Several things can break the link between good SEO and good AI visibility:

• Key pages that AI crawlers cannot read, because of robots.txt rules or client-side rendering.
• A brand name shared with another company, so the model credits your work to someone else.
• Facts that differ between your site, your profiles and press coverage, so the model has no reliable version to repeat.
• Content that sits in an ecosystem a given assistant cannot reach, such as WeChat Official Accounts for most assistants outside Tencent’s own (see Tencent Yuanbao).

The three dimensions of AI visibility

A useful definition separates three questions.

1. Presence: does the brand appear? Measured by mention rate: answers that correctly mention the brand divided by valid answers. “Correctly” matters, because a mention of a same-name company is not a mention of you.

2. Prominence: is the brand recommended first? Measured by first-recommendation rate: answers that place the brand first divided by valid answers. On recommendation questions, readers tend to look only at the first one or two options, so this is often closer to commercial impact than mention rate. A brand can appear in every answer and never be first.

3. Accuracy: is the description right? Checked by reviewing what each answer actually says, plus name confusion rate: the share of answers using your name that describe a different organization. A high mention rate combined with high confusion means your visibility is benefiting someone else.

Reporting only one of these can mislead. Presence without accuracy is not visibility you want.

How to measure AI visibility

Measurement is only as good as its method. The approach used in GEO work rests on a few rules.

Build a fixed question set. A question set is the list of prompts you test with, and it is the denominator of every metric above. Mix question types:

• Category questions that do not name the brand (“which firms offer…”)
• Decision queries (“how do I choose…”, “is it worth…”)
• Long-tail questions about specific situations
• Branded questions (“what does [company] do?”)
• Disambiguation questions involving likely same-name entities

Write them in buyer language, not your own marketing copy, or the results will be inflated.

Freeze the rules. Record the exact wording, the platforms, whether web search is on, how many times each question is sampled, what counts as a valid answer and how “first” is judged.

Track platforms separately. Each assistant reads different sources, so do not average them into one score. Differences between platforms tell you where your content assets are concentrated.

Measure a baseline, then retest. Run the set once before making changes, then repeat under identical conditions at regular intervals and log what happened in between (see baseline and retest). AI answers vary between runs and models update continuously, so a single snapshot proves little.

How to read the numbers

A question set is a constructed sample, not a record of what real users ask. AI visibility metrics therefore show direction of change under fixed conditions, not market share. Changes can come from your own work, a model update, a competitor’s activity or new third-party coverage, and these cannot be fully separated. State conclusions as observations (“under this method, in this period, mention rate moved from X to Y”), not as proof of cause.

Breaking results down by question type is often more useful than the headline number. Strong long-tail visibility with weak category visibility means the content works but the model does not yet see you as a player in the category.

How AI visibility is improved

Improving AI visibility is the work of generative engine optimization (GEO), also called LLM SEO or, as an umbrella, AI search optimization. It usually proceeds in layers:

1. Technical access: crawlable pages with text in the HTML.
2. Entity clarity: a well-defined brand entity with consistent facts, structured data and sameAs links.
3. Answerable content: pages that answer real questions in the first sentence.
4. Outside confirmation: third-party evidence that matches your own claims.

None of these guarantee a mention in any particular answer; they raise the odds and make results more stable.

Where to start

An AI visibility audit combines a baseline measurement with technical and entity checks. You can request a free GEO audit from Moments, or read how we approach this work as an AI SEO agency.