Understand what your AI visibility results mean.

A check records answers to your saved questions from the selected models. Each result belongs to a particular question, model and point in time. Together, those answers describe your brand’s visibility within that scope.

1. Set the scope

   Choose questions relevant to your market, language and buyers, then select the models you want to monitor.

2. Collect the answers

   Questions are sent separately to the selected models. The measurement request does not add instructions asking a model to favour your brand.

3. Analyse what appeared

   ShareOfAsk identifies brand mentions, competitive context and cited sources. The recorded answer remains available for you to inspect.

4. Review the next step

   Actions uses the findings and supporting evidence to suggest priorities. Review the rationale before deciding what to change.

For a discovery benchmark, ask about the category, use case and requirements that shape a shortlist. Give the model room to name relevant options.

Which cloud migration providers should a UK financial services company evaluate?

Review suggested and manually written questions for relevance and leading language. Automatic checks do not guarantee that a question is neutral.

Questions that name your brand can help examine its perception. Keep them separate from discovery questions when interpreting visibility: naming a brand in the question changes the task you are measuring.

Check the scope before comparing tiles. On Overview, AI Presence, Competitive Share of Voice and Average First Position use the latest check; Question Coverage uses the available answers from the last three months. Other views follow their displayed period and filters.

A useful recommendation should give your team something to examine. Keep the observation, its context and the proposed action distinct.

Recorded answers

Read the model’s returned answer alongside its question. Brand detection, sentiment and other extracted labels are interpretations of that answer and can contain errors.

Cited sources

Sources include citations returned by providers and links found in answer text. Open the associated pages and check what they support. A citation does not, by itself, prove why a model mentioned or recommended a brand. Source availability varies by model and response.

Suggested actions

Actions combines findings, project context and follow-up research to suggest priorities. That research is kept separate from visibility measurements. Review the evidence and proposed steps with your team; the result is a recommendation for your judgement.

Repeated checks help you see whether a pattern persists. Keep the setup stable so changes in the measurement are easier to interpret.

1. Keep questions and models consistent. Review changes to wording, language or the selected models before attributing movement to your marketing work.
2. Match the scope. Compare the same period, filters and included brands. Changing the brand pool can change competitive share of voice.
3. Look at the counts. A percentage based on fewer successful answers deserves different weight. Inspect the underlying answers when a result moves sharply.
4. Keep a record of your changes. Note the work your team completed and compare later findings with it. Read AI visibility alongside search, website and commercial data.

An observed improvement is useful evidence to investigate. It does not establish that one content change caused it, or that revenue increased.