AI Visibility: How to Track Your Store in AI Answers

AI visibility measures how often AI answers mention your store and how they describe it. Track it with a fixed set of buyer questions, repeat those questions across the same assistants and record mentions separately from citations. Your store’s analytics then show which recorded AI referrals reach a product page or lead to an order.

For a small store, a spreadsheet and a monthly check are enough to start. Automated tools help when you need more questions, several markets or frequent comparisons with competitors. Keep the question set stable so a change in your report reflects the answers you collected.

What AI Visibility Means for a Store

AI visibility is how often and how favourably AI tools mention your brand and products in answers to relevant questions. Watch Google AI Overviews and AI Mode, plus ChatGPT, Perplexity, Gemini and Copilot. Record each surface separately because they answer questions differently.

AI brand visibility combines plain mentions with recommendations and citations. A mention names your store. A recommendation suggests it as a place to buy. A citation links to one of your pages, sometimes without recommending your store at all.

Search engine optimisation (SEO) tracks your place among search results and the traffic those results attract. An AI answer can name only a handful of stores or give an answer without citing your site. A strong organic ranking therefore doesn’t establish that shoppers hear about your brand in an assistant.

Separate visibility from visits. Your store can appear in an answer that nobody clicks, while a citation can send a shopper straight to a product page. Measure both to answer the founder’s question about brand presence and the trading team’s question about sales.

Share of Model and AI Share of Voice

Share of model is the percentage of relevant AI answers that name your brand. This measure is also called AI share of voice, although vendors use different denominators for similarly named metrics. Define your calculation in the report before comparing results.

Use this calculation for a fixed prompt set: answers naming your brand divided by all completed answers, multiplied by 100. Count an answer once even if it repeats your brand name. Record a failed run separately and retry it instead of counting it as an answer without your brand.

In a hypothetical check, you ask 20 questions across five assistants once each. That gives 100 answers. If 18 name your store, your share of model is 18%. If a competitor appears in 32 answers, its rate is 32%. Both stores can appear in one answer, so these rates don’t need to add to 100%.

Track the same measure for two or three competitors month by month. Keep branded questions separate from unbranded buying questions. A prompt that explicitly asks about your store measures how AI describes it, while an unbranded prompt tests whether AI chooses to mention it.

How to Track AI Overviews

Track AI Overviews by recording whether Google shows an overview for each query and whether it names your store or cites your site. Rank trackers can automate those observations. Semrush Position Tracking, for example, reports AI Overview presence and whether a domain appears as a cited source.

Google says traffic from AI Overviews and AI Mode is included in Search Console’s overall Web performance data. That report doesn’t give you a separate total for each AI surface. Use Google’s performance guidance to interpret those clicks alongside your own answer records.

For a manual check:

1. Select a short list of buying questions from customer enquiries and your search data. Include an Australian delivery or location requirement when it belongs in the question.
2. Search each query using the same device type and location. Record those settings so the next check uses the same conditions.
3. Record whether an AI Overview appears. If it does, log your brand’s mention and save the cited URLs, including competitors’ URLs.
4. Save a screenshot or the answer text with the query and observation time. Keep regular organic results separate from the overview.
5. Repeat the list monthly. Compare overview availability as well as your citation rate.

An overview that doesn’t appear is a separate result from an overview that omits your store. Google says AI Overviews don’t trigger for every query. If a tracker records a citation but your manual search doesn’t, compare its location and collection time before treating either observation as an error.

Tracking Mentions in ChatGPT and Other Assistants

Track assistant mentions by asking the same buyer questions and recording each answer’s treatment of your store. Build the prompt set from questions customers send your team and terms they use in onsite search. Include the product comparisons they ask about. Keep the wording specific enough to describe a buying need.

For example, a hypothetical outdoor retailer could track “Which Australian stores sell waterproof hiking jackets for summer storms?” A separate branded question could ask whether that retailer delivers to regional addresses. The first measures discovery. The second tests how accurately the assistant describes the store’s policy.

Start a new conversation for each prompt so earlier messages don’t supply the brand name. Record the assistant, available model label, search setting, location and whether you’re signed in. Generated answers can differ between runs because the response and retrieved sources change. Personalisation and conversation history can also change the context.

Use a row for each prompt, assistant and run. Record these fields:

Field	What to record	Why it helps
Brand mention	Yes or no, with the exact wording	Measures whether your store appears
Recommendation	Whether the answer suggests buying from you	Separates a recommendation from a passing reference
Mention position	First named store, second named store or later	Shows prominence within that answer
Citation	The URL of each linked page on your site	Measures citations separately from plain mentions
Tone and accuracy	Positive, neutral or negative, with the supporting excerpt	Shows whether the description helps or misleads shoppers
Competing stores	Names and positions of the same comparison brands	Supports a consistent benchmark

Use these fields for Gemini as well as the other assistants. Citation position and brand mention position are different measurements, so label them separately. Inspect an unfavourable answer’s wording before assigning tone: a correct warning about an unsuitable product is different from a false delivery claim.

Repeat the same prompts across the same assistants every month. For questions that produce unstable answers, collect several runs and retain every result. Compare mention rate, recommendation rate and citation rate per assistant before combining them into a total.

An AI visibility score is a vendor’s summary of its selected observations. It can weight prominence or use a broader prompt database instead of your own questions. A score from one tool isn’t interchangeable with another tool’s score. Keep the underlying counts and prompt coverage beside it.

Finding LLM Traffic in GA4

Find large language model (LLM) referral traffic in Google Analytics 4 (GA4) through the session source data, then group the observed AI sources into a custom channel. You need Editor access or higher to create that channel group. Use Google’s custom channel group instructions for the settings.

1. Open the Traffic acquisition report and select Session source / medium. Inspect the sources that actually arrive at your store and identify AI assistant referrals.
2. In Admin, open Data display, then Channel groups. Create a new channel group by copying an existing group.
3. Add a channel named AI assistants. Set its Source conditions to match the assistant sources observed in your report.
4. Place the AI assistants channel ahead of general Referral in the channel order. GA4 assigns traffic to the first matching channel.
5. Save the group and use its session channel dimension in an acquisition report. Check that known assistant referrals now appear in AI assistants.

Use your property’s observed source values instead of assuming every assistant sends the same hostname from every app. If known referrals remain in Referral, inspect the Source condition and channel order. If the new channel is empty, check the reporting period and whether any matching referrals exist.

Some AI-assisted visits have no usable referrer. A shopper might copy a link or type your store’s address after reading an answer. Missing source information can also affect attribution. Google explains that direct traffic has no clear referral source. An increase in Direct doesn’t prove an increase in AI visits.

AI Attribution: Linking Sales to AI Referrals

AI attribution connects orders with sessions that began through a recorded AI assistant referral. Start with the AI assistants session channel and compare its sessions with purchases and purchase revenue. Your store needs working ecommerce event collection before this comparison can show orders.

Separate a session-based order count from credit assigned by an attribution model. A shopper can first visit through an assistant, return through email and then buy. Different reporting scopes can credit those visits differently, so state which report and measure you use.

In a hypothetical report, 50 recorded assistant-referral sessions include four purchases. The session purchase rate is four divided by 50, or 8%. That describes the recorded sessions. It doesn’t establish that AI caused four extra sales or reveal how many shoppers read an AI answer without clicking.

A later direct visit can hide the original influence, especially across devices. Ask an optional post-purchase question about how customers found you if that suits your checkout. Report those responses as self-reported AI influence, separately from measured referral orders. Treat broader AI-influenced sales as an estimate with stated assumptions.

If referrals appear but purchases don’t, check purchase event collection and the reporting scope before concluding that the visits don’t convert. Keep order counts and revenue consistent with the store’s own reporting rules.

Running an AI Visibility Audit

To audit your store’s AI visibility, record where your brand appears and where competitors appear instead. Collect 20 to 40 real buyer questions, then run them across the assistants you want to monitor. A small business can begin with ten questions and a spreadsheet, then expand the set as it learns which questions matter.

1. Group the questions by buying task, such as choosing a product or checking delivery. Separate branded questions from questions that don’t name a store.
2. Ask each assistant under recorded settings. Save each answer and cited page so another person can inspect the observation.
3. Log your store’s mentions and recommendations. Record the same fields for two or three competitors and retain every citation to your site.
4. Calculate rates for each assistant and buying task. Flag incorrect product details and questions where competitors consistently appear instead.
5. Write a report with the prompt set, completed-run count and comparison rates. Attach an ordered list of affected pages or product data, with the answer excerpt behind each item.

Prioritise a false stock or delivery statement over a harmless difference in wording. A report entry might say that an assistant cited an old returns page while describing a policy your store no longer uses. That entry gives the page owner a specific URL and claim to investigate without treating every missing mention as a defect.

Verify the audit by replaying a sample of the prompts under the recorded settings. The new answers needn’t match word for word. Your saved evidence must let a colleague trace each reported rate back to the rows counted. Retain the baseline when you change the prompt set, so the next report explains the change in coverage.

Choosing an AI Visibility Tool

The right AI visibility tool tracks your buyer questions in the assistants and locations you need. Compare subscription structure and checking allowances as well as the headline feature list. A broad database and a custom prompt tracker answer different questions.

Tool type	What it tracks	Coverage and limits to compare	Cost structure
Manual spreadsheet	Your own prompt set, mentions and saved citations	Any assistant you can use directly, with recorded Australian settings	Staff time and any assistant subscription
SEO rank tracker with AI reporting	Google queries that trigger AI Overviews and cited domains	Search location, device, keyword allowance and result snapshots	SEO subscription and keyword or campaign limits
Custom prompt tracker	Scheduled answers to the questions you select	Supported assistants, location, prompt allowance and refresh frequency	Subscription plus included checks or prompt allowance
Broad AI visibility database	Brand presence across a vendor’s collected questions	Topic coverage, country coverage and the questions behind each score	Database access, platform packages or bundled subscription

Ahrefs Brand Radar combines a broad AI visibility index with custom prompt tracking. Its documentation lists AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot for custom checks. Each prompt uses a check for each platform, location and refresh. Its Lite allowance lists 150 checks per month, so refresh frequency changes how many questions you can track. The custom prompt documentation explains the location selector and checking schedule.

Semrush’s Prompt Tracking documentation lists ChatGPT Search, Google AI Mode and Gemini. The standalone AI Visibility Toolkit includes 25 custom prompts. Australia is included in its supported Volumes and Topics countries. Tracking is desktop-only, so it doesn’t describe a complete mobile shopping experience. Its broader toolkit reports cover additional AI surfaces, which must be distinguished from those available for custom tracking.

For Australian monitoring, compare the exact location available for each assistant and each report. Australian topic data, an Australian search location and Australian billing are separate properties. Choose the tool that covers your buyer questions and intended locations without paying for a large database your brand barely appears in.

Benchmark your store and two or three competitors on the same unbranded prompt set. Keep assistant coverage and the number of runs equal. Compare mention rate and citation rate by product category, then review the saved answers behind any large change. Export the baseline before switching tools because a new prompt database can change the score without changing your brand’s presence.

Once the audit identifies the pages to improve, use generative engine optimisation for store-wide changes or the AI Overviews guide for Google citation work.