What is AI visibility, and how do you measure it?

AI visibility is how often, and how favourably, an AI assistant mentions or recommends your brand when someone asks a question you want to be the answer to. It is the search ranking of the assistant era, with one difference: there is no page one. An assistant writes a single answer, names a handful of sources, and the buyer rarely looks further.

Why one check is not a measurement

Ask ChatGPT the same question five times and you can get five different answers. Assistants sample. A brand that appears in two of five answers has a visibility of about 40 percent, and a single check would have told you either "yes" or "no". Any measurement worth acting on repeats each prompt and reports a range.

The second trap is reading only for a mention. "Brand X is an option, though users report billing problems" is a mention. So is "Brand X is the one most teams choose." A proper reading records the position in any list, the sentiment, the descriptors used, what the brand is recommended for, and the criticisms. Those are the things you can change.

The four numbers that matter

1. Share of voice: of all brand mentions across the answers to your prompts, what share is yours, per assistant.
2. Sentiment: positive, neutral or negative, and the words behind it.
3. Cited sources: which pages the assistants pointed to. These are the sites to earn a mention from, and the pages of yours that already work.
4. Trend: the same prompts, weekly, so a change in the model or in your content shows up.

What moves it

Assistants synthesise from sources they trust, so the work is less about ranking and more about being quotable and being cited.

• Answer first. Open pages with a direct, complete answer in one or two sentences, then the detail. Assistants lift those sentences.
• Structure for questions. Question headings, short definitions, lists and tables. Structured data (FAQ, Product, Organization) helps the crawlers that feed the assistants.
• Let the crawlers in. GPTBot, ClaudeBot, PerplexityBot and Google-Extended must be allowed in robots.txt, and the content must be in the HTML, not rendered only by JavaScript. Publish an llms.txt that points to your best pages.
• Earn citations. Find the sources assistants already cite for your topics and get mentioned there: reviews, comparisons, industry publications, directories.
• Fix the criticisms. If every assistant repeats the same complaint, the complaint is the work.

Measuring it in Serpified

Serpified asks ChatGPT, Gemini, Claude, Perplexity and Google's AI Overviews your prompts several times each, reads every answer per brand, and reports share of voice, sentiment, descriptors, cited sources and a ranked action list. The site audit checks crawler access and server-side content; the content tools score pages for answer shape. Runs are weekly, with an alert when a number drops.

Start with ten prompts your buyers actually ask. That is enough to see where you stand and what to do first.