Measure AI visibility: is your online store actually mentioned?

Finding your store in one AI answer raises an obvious question: is the optimisation working? A single response cannot establish that. It records what appeared under particular conditions. Useful AI visibility measurement needs a repeatable process.

At sjmedia, we separate three questions: is the merchant mentioned, is an owned page cited, and are there identifiable visits followed by relevant enquiries or orders? These signals are related, but should not be combined into one success figure.

Which questions should you test?

Start with actual buying decisions relevant to your audience. An ergonomic office-chair retailer might test which features suit small workspaces and which merchants serve a particular delivery area. These are illustrative topics, not measured search queries.

Separate questions without brand names from questions that already name your store. Asking about your business primarily tests how a known brand is described. A general product-selection question tests whether your offer is considered at all. Both are useful, but answer different questions.

Choose a manageable set and retain it across reporting periods. Analyse additions as a new group. Otherwise an apparent increase in mentions may simply reflect the addition of easier questions.

What should each test record?

A simple spreadsheet is sufficient initially. Record six items for each question and run:

• Question and intent: exact wording and whether it concerns product selection, comparison or merchant discovery.
• Conditions: platform, visible mode, language, region and relevant account settings.
• Timing: date and time so repeated runs can be identified.
• Result: distinguish brand mentions, specific product recommendations and links to your own URLs.
• Accuracy: check the catalogue, prices, delivery area and merchant attribution.
• Evidence: retain an answer extract or screenshot and the resulting investigation task.

Start without prior conversation content where possible. Even then, identical answers are not guaranteed. Record the visible context rather than claiming complete neutrality. Use separate German and Swiss groups where service areas or offers differ.

How should you interpret a mention rate?

Suppose your store appears in six of twenty recorded answers. Its share in that particular test is 30 percent. This hypothetical calculation describes neither all customer questions nor the number of people reached. Repeating a question does not create additional independent search volume.

Always report the denominator: question count, repetitions and platforms examined. Show changes within each question group rather than turning arbitrary answers into a general visibility index. An inaccurate recommendation is not a success just because it includes your brand.

What can Search Console and Analytics add?

Google's AI features documentation places appearances in AI Overviews and AI Mode within Search Console's general performance reporting under the Web search type. This does not provide a separate, complete measure of your store's AI reach. Source checked on 20 October 2026.

Analytics can supplement the picture with identifiable referrals from AI services and subsequent actions. Attribution is not always possible. Missing referral information, consent and movement between apps or devices can limit measurement. Test answers and measured visits are therefore separate datasets.

Check the destination pages of identifiable visits. Do customers reach an available product, a helpful guide or an outdated URL? A few relevant visits may be more useful than many irrelevant mentions. Without reliable attribution, do not assign revenue to an AI optimisation measure.

How do you turn findings into improvements?

Translate observations into specific investigations. If the delivery area is wrong, review publicly available shipping information. If an incorrect variant is recommended, examine product names, attributes and mapping. Where a useful answer is missing entirely, add it where it genuinely supports customer selection.

For tailored store requirements, we favour PrestaShop for implementation. Product data, content and extensions can be coordinated around the business. This provides practical flexibility, not an automatic advantage in AI answers. Assessment of the actual installation determines the work needed.

Our GEO fundamentals for online stores explain optimisation areas; this measurement plan helps track changes. Through GEO consulting, we connect documented questions, verifiable content and technical improvements. Retest under comparable conditions after changes, without expecting a guaranteed placement within a fixed period.