GEO Reporting: How to Measure Your Store's Visibility in AI Search

Every agency now sells AI search optimisation. Very few of them can show you a report that proves it worked. If you run a Shopify store and someone has pitched you generative engine optimisation (GEO), the first question you should ask is how they plan to measure it. It is a fair question, and it deserves a straight answer rather than a slide about the future of search.

This article sets out what GEO reporting can genuinely measure today, what it cannot, and a cadence a real business can run without a data team. We covered the optimisation side in our guide to GEO for Shopify. This is the measurement side.

Start with honesty about the numbers

AI search visibility is real, but the traffic is small. Similarweb's 2026 cross-site analysis put average referral traffic from AI assistants at a fraction of one percent of total visits, against roughly a quarter of all traffic from organic search. Sector studies land a little higher, around 1% for some categories. Either way, this is not yet a channel that replaces Google for a UK ecommerce brand.

Two things make it worth measuring anyway. First, growth: the same datasets show AI referrals up close to tenfold over the previous 19 months. Second, quality. Multiple 2026 studies, Similarweb's included, found AI-referred visitors converting at roughly double the rate of organic search visitors. Someone who arrives after asking ChatGPT which trail running jacket suits Scottish weather has already done most of their research. They are further down the funnel, and that is exactly the kind of visitor a plateaued store needs more of.

Prompt-set tracking: the measurement that actually matters

The most useful thing you can measure is whether AI assistants cite you when a buyer asks a question in your category. Nobody outside OpenAI or Google can see the true volume of those questions, so you approximate it with a fixed prompt set.

Write down 20 to 40 buying questions that matter to your category. Not brand queries, buying queries: "best organic dog food subscription UK", "which linen bedding brands ship to Glasgow", "is [competitor] worth the money".Run each prompt on a schedule across the assistants that matter to your market. For most UK stores that is ChatGPT (which carries around three quarters of all AI referral traffic in 2026), Google AI Overviews and AI Mode, Perplexity and Copilot.Record three things per prompt: were you cited at all, what position you held in the answer, and which competitors appeared alongside you.Note the sources the assistant cited. If it is quoting a Reddit thread or a comparison blog rather than your own store, that tells you where the influence sits.

Do this monthly at minimum, and keep the prompt set stable. The value is in the trend, not any single run. Assistant answers vary between sessions, so a single check proves nothing. Ten checks across a quarter start to mean something.

Finding AI referrals in GA4 and Shopify analytics

The second layer is traffic that actually arrives. Both GA4 and Shopify's own reports will show it, but neither labels it well out of the box.

In GA4, the referrals from ChatGPT arrive with a source of chatgpt.com, and OpenAI often appends a utm_source parameter without a matching medium. GA4 then files those sessions under Unassigned, where nobody looks. The fix is a custom channel group: create an "AI Referrals" channel where session source matches a regular expression covering chatgpt.com, perplexity.ai, claude.ai, gemini.google.com and copilot.microsoft.com, add a second rule for the same sources with medium equal to (not set), and drag the channel above Referral so it wins the match. In Shopify analytics, the Sessions by referrer report will show chatgpt.com and perplexity.ai as plain referrers. It is cruder than GA4, but it is the number your finance team will trust because it lines up with orders.

Now the caveats, because they are large. Google AI Overviews and AI Mode clicks are not separated from ordinary Google organic clicks in either tool, and Google Search Console still lumps them together. A meaningful share of ChatGPT visits arrive without a referrer at all, especially from mobile apps, and show up as Direct. So whatever AI referral number you see is an undercount, and you cannot say by how much. Report it as a floor, not a total. If your wider analytics setup is shaky, fix that first: our guide to the Shopify metrics that actually matter covers the basics.

Crawler analysis: are you being read at all?

Before an assistant can cite you, its crawlers have to be able to reach you. OpenAI documents three separate agents: GPTBot for model training, OAI-SearchBot for indexing content that ChatGPT search links to, and ChatGPT-User for live page fetches when a user asks a question. Anthropic runs ClaudeBot, Perplexity runs PerplexityBot, and Google's Google-Extended token controls whether your content trains Gemini. One point catches people out: Google-Extended has no effect on AI Overviews. Those are built from the normal Search index, so if Googlebot can crawl and snippet you, you are eligible. You can check the current agent names in OpenAI's crawler documentation.

On a standard Shopify store you do not get raw server logs, so full crawler analysis is off the table. What you can do is check your robots.txt.liquid file to make sure nothing is blocking the search crawlers you want (blocking GPTBot for training while allowing OAI-SearchBot is a legitimate choice, and the two are independent), and confirm your product and collection pages return clean, fast HTML with structured data. Headless stores behind a CDN such as Cloudflare can go further and see bot hits by user agent. For everyone else, the practical proxy is simple: if you are showing up in the prompt-set tracking, you are being read.

AI visibility tools: what they do and do not tell you

There is now a crowded market of tools that automate prompt-set tracking. Profound sits at the enterprise end with pricing to match. Peec AI is the common starting point for agencies and mid-sized brands. Semrush sells an AI visibility toolkit as an add-on to its existing plans, and Ahrefs Brand Radar has a free tier and a large prompt index. Prices and features change monthly, so check before you commit.

Be clear on what they are doing. Every one of them runs prompts on your behalf and records what comes back. That is useful, and it saves hours, but it is still a sample of synthetic queries, not real buyer behaviour. None of them can see actual query volume inside ChatGPT. Some report a "share of voice" figure that looks precise and is really a percentage of their own prompt list. Treat the output as directional. A tool subscription is not a measurement strategy on its own.

AI visibility tools measure how often you appear in the questions they chose to ask. That is not the same as how often you appear in the questions your customers ask.

A GEO reporting cadence a real business can run

Here is what we run for clients on our Conversion Growth Retainer, stripped down so a store owner can do it in a couple of hours a month.

Baseline once. Run your full prompt set across the assistants that matter. Record citation rate, average position and the competitors cited. Set up the GA4 channel group and note the current monthly AI referral sessions, orders and revenue. Check robots.txt. Do this before any GEO work starts, or you will never know what changed.Monthly. Re-run the prompt set. Report citation rate against baseline, new prompts you now appear in, and prompts you have dropped from. Pull AI referral sessions, conversion rate and revenue from GA4 and Shopify side by side. Note which of your pages are being cited.Quarterly. Review the prompt set itself. Add questions that have emerged, retire ones nobody would ask. Compare AI referral conversion rate against your organic and paid channels. Decide whether the effort is earning its place.

What does movement mean? A rising citation rate on buying prompts, with the same prompt set, is the clearest signal that the content and structured data work is landing. AI referral sessions climbing with a conversion rate well above your site average means the channel is delivering the right kind of visitor, even at small volume. Citations rising while referrals stay flat usually means you are being mentioned without being linked, which points to a product page or comparison content problem, not a visibility problem. And if citations are flat after three months of work, say so and change approach.

Keep it in proportion

Attribution in AI search is immature. The volumes are small next to organic. The tools are samples dressed as census data. All of that is true, and none of it is a reason to ignore the channel, because the visitors it sends are unusually close to buying. It is a reason to report it honestly, as one line in a wider conversion picture, rather than as a headline number.

Which brings us back to the point we make to most stores that come to us. Most Shopify brands don't have a traffic problem, they have a conversion problem. AI search will send you a small number of well-qualified visitors. Whether they buy depends on the same product pages, trust signals and checkout that every other visitor sees. If those are not right, no amount of GEO reporting will fix the numbers. Our conversion rate optimisation work starts there.

If you would like a second opinion on how to baseline AI search visibility for your store, or on whether it deserves attention yet, get in touch. We are a senior team in Glasgow, we work with brands across the UK and beyond, and we will tell you plainly if the answer is "not yet".