Key takeaways
- ChatGPT's shopping feature only triggers on roughly 9% of prompts, so raw traffic numbers hide what's really happening -- you need to track trigger rate, not just impressions
- Profound's research across 260 million prompts found that Walmart and Target dominate buy-link clicks, while Amazon barely appears -- brand visibility in AI shopping is not correlated with market share
- Most e-commerce analytics tools weren't built for AI-mediated commerce, which means standard dashboards miss AI-referred sessions, citation patterns, and carousel appearances entirely
- The 7 metrics below give you a complete picture: from whether AI models even know your products exist, to whether that visibility converts into actual revenue
AI retail traffic grew 393% year-over-year in Q1 2026, according to Elogic Commerce's research. ChatGPT alone has 900 million weekly users. And conversion rates from AI-referred shopping sessions are running about 42% higher than traditional search.
Those numbers sound like a gold rush. But here's the uncomfortable truth: most e-commerce brands have no idea whether they're part of it.
The problem isn't effort. It's that the metrics most teams are watching -- Google Analytics sessions, ROAS, keyword rankings -- weren't designed to capture what happens when an AI model recommends your product instead of a search engine listing it. The measurement gap is real, and it's growing.
This guide covers the seven metrics that actually matter for ChatGPT Shopping in 2026, why each one is harder to measure than it looks, and what to do about it.
1. Shopping trigger rate
The first thing most brands get wrong is assuming ChatGPT Shopping is everywhere. It isn't.
Profound's analysis of 2 million tracked prompts found that the shopping feature -- the product carousel with images, prices, reviews, and buy links -- only activates on about 9% of queries. That means 91% of prompts, even ones that seem shopping-adjacent, never show a product card.
Trigger rate is the percentage of your tracked prompts where ChatGPT actually surfaces a shopping carousel. It tells you whether the queries you care about are even in the game.
Why does this matter? Because if you're optimizing for prompts that rarely trigger shopping, you're spending effort on the wrong targets. The prompts that do trigger tend to be specific, transactional, and category-driven -- "best running shoes under $150" fires the carousel; "how do I start running" almost never does.
What to track: For each prompt you care about, log whether the shopping feature appeared. Over time, you'll build a map of high-trigger vs. low-trigger query types in your category. Focus your optimization on the high-trigger ones.
2. Brand mention rate in AI responses
Before you can win a buy-click, you need to exist in the response.
Brand mention rate measures how often your brand name appears anywhere in ChatGPT's answer to a relevant shopping prompt -- in the product carousel, in the explanatory text, in a recommendation, or even as a passing reference. It's the AI equivalent of organic visibility.
This is where most brands discover a humbling reality. Profound's research found that Amazon -- the dominant force in e-commerce -- barely shows up in ChatGPT Shopping results. Walmart and Target, by contrast, appear consistently. AI visibility doesn't follow the same rules as market share or Google rankings.
The implication: your brand's presence in AI responses depends on what AI models have learned from the content they've been trained on and crawled. If your product pages, reviews, and editorial coverage are thin, vague, or poorly structured, AI models simply won't have enough signal to cite you.
Tracking this manually across thousands of prompts is impractical. Tools like Promptwatch are built specifically for this -- they monitor how often your brand appears across ChatGPT, Perplexity, Claude, and other AI models, broken down by prompt and product category.
3. Share of carousel slots
Getting mentioned is one thing. Getting a buy-link slot is another.
When ChatGPT Shopping does trigger, the carousel typically shows between 3 and 8 products. Share of carousel slots measures how often your products appear in those slots, and where -- position 1 vs. position 6 matters, just like it does in traditional search.
Profound's data on 22.5 million individual buy offers found that carousel position is not random. Certain retailers consistently dominate the top slots. The factors that seem to influence position include review volume and recency, price competitiveness, product description quality, and how well your content matches the specific phrasing of the prompt.
This metric is harder to track because it requires actually running prompts and recording the carousel output -- not just checking whether your brand was mentioned. You need to know: when the carousel appears, are you in it? And if so, where?
Some teams are building this tracking manually using scripts that query ChatGPT and parse the output. Others are using platforms that do it at scale. Either way, if you're not measuring carousel position, you're missing the most direct signal of AI shopping performance.
4. AI-referred traffic and its conversion behavior
This is where the measurement gap bites hardest.
When a user clicks a buy link from a ChatGPT Shopping carousel, the traffic that lands on your site often shows up in analytics as "direct" or gets misattributed to other sources. Standard UTM parameters don't exist in AI-generated responses. The referrer string from ChatGPT is inconsistent. And many analytics setups aren't configured to catch it.
The result: brands are receiving AI-referred traffic and have no idea. They're also missing the conversion data that would tell them whether AI-referred visitors behave differently from search-referred ones -- and the early evidence suggests they do. Elogic's research puts AI-referred conversion rates about 42% higher than traditional search, which makes sense: someone who asked an AI for a product recommendation and clicked through is further along in their decision than someone who typed a keyword into Google.
To capture this properly, you need one of three approaches:
- A JavaScript snippet that detects the referrer and tags sessions correctly
- Google Search Console integration (limited but useful for AI Overview traffic)
- Server log analysis, which catches traffic that client-side scripts miss
Promptwatch supports all three methods. Most standard analytics platforms don't handle this natively yet, so it's worth auditing your setup before you assume your AI traffic data is accurate.
5. Prompt coverage and gap rate
Most e-commerce brands have a mental model of "the keywords we rank for." The AI equivalent is prompt coverage -- the set of shopping queries where your brand appears in AI responses.
Prompt coverage measures how many of the relevant prompts in your category you're visible for. Gap rate is the inverse: the percentage of relevant prompts where competitors appear but you don't.
This is one of the most actionable metrics on this list because it directly points to content you should create. If ChatGPT recommends three competitors when someone asks "best eco-friendly yoga mats under $80" but not you, that's a specific, fixable gap. The AI model isn't citing you because it doesn't have sufficient content to draw on -- your site probably doesn't have a page that directly addresses that query.
Profound's research across 13,000 product categories gives a sense of how wide these gaps can be. Even well-known brands are invisible for large portions of the relevant query space. The brands that are winning have systematically identified these gaps and created content that answers the specific questions AI models are trying to answer.
Tools that do answer gap analysis -- showing you exactly which prompts competitors rank for that you don't -- are worth the investment here. It turns a vague "we need more AI visibility" goal into a specific content backlog.
6. Citation source quality and distribution
When ChatGPT recommends your product, it's drawing on sources. Those sources might be your own product pages, but they might also be third-party reviews, Reddit threads, YouTube videos, or editorial content on other sites.
Citation source quality measures which external sources are driving AI recommendations in your category, and whether your brand is well-represented in those sources. Distribution measures how spread out those citations are -- a brand that only appears in one source is more fragile than one that appears across many.
This matters because your AI visibility isn't entirely under your control. If the top Reddit thread in your category recommends a competitor, that thread is probably influencing ChatGPT's recommendations. If a YouTube review from 18 months ago describes your product inaccurately, AI models may still be drawing on it.
The practical implication: AI visibility strategy isn't just about your own website. It includes monitoring what third-party sources are saying, identifying which forums and review sites carry the most weight in your category, and making sure your brand is represented accurately in those places.
Most brands aren't doing this at all. They're optimizing their product pages while ignoring the Reddit discussions and YouTube reviews that AI models are actually citing.
7. Visibility-to-revenue attribution
The final metric is the hardest and the most important: connecting AI visibility to actual revenue.
Metrics 1 through 6 tell you how visible you are and where the gaps are. Metric 7 tells you whether any of it is working. Without it, you're optimizing for visibility scores without knowing if visibility translates to sales.
The challenge is that the attribution chain is long and leaky. A user asks ChatGPT about running shoes. Your brand appears in the response. They don't click immediately -- they close the tab, think about it, and come back two days later via a direct search. Your analytics credits the direct search. The AI recommendation that started the journey is invisible.
This is why last-click attribution is particularly broken for AI-influenced commerce. The Coupler.io ecommerce analytics report for 2026 specifically calls out AI agents as a source of behavioral data distortion -- AI-assisted sessions look different from traditional sessions, and standard models misread them.
Fixing this requires moving toward first-touch or multi-touch attribution models that can capture the AI touchpoint, even when the conversion happens later through a different channel. Server log analysis helps because it captures the initial AI-referred visit even if the user doesn't convert immediately.
The brands that will win in AI-mediated commerce aren't just the ones with the best visibility -- they're the ones who can prove the ROI of that visibility and reinvest accordingly.
Putting it together: a measurement framework
Here's a simple way to think about these seven metrics as a system:
| Metric | What it measures | Why most brands miss it |
|---|---|---|
| Shopping trigger rate | Whether your target prompts activate the carousel | Teams track impressions, not trigger events |
| Brand mention rate | Presence in AI responses | No standard tool tracks this by default |
| Share of carousel slots | Position and frequency in buy-link carousels | Requires active prompt monitoring |
| AI-referred traffic | Sessions and conversions from AI sources | Misattributed as direct or organic |
| Prompt coverage & gap rate | Which queries you're visible for vs. competitors | Most brands don't know their prompt universe |
| Citation source quality | Which external sources drive AI recommendations | Brands focus on their own site, ignore third-party sources |
| Visibility-to-revenue attribution | Whether AI visibility drives actual sales | Last-click models hide the AI touchpoint |
The good news is that you don't need to track all seven perfectly from day one. Start with brand mention rate and AI-referred traffic -- they're the most immediately actionable and give you a baseline. Then layer in prompt coverage and gap rate once you have a sense of where you stand.
Tools worth knowing about
A few platforms are worth looking at depending on where you are in this journey.
For AI visibility monitoring and prompt tracking, Profound has published some of the most rigorous research on ChatGPT Shopping behavior and their platform reflects that depth.
For brands that want to go beyond monitoring and actually fix the gaps -- creating content that gets cited, tracking which pages AI models crawl, and attributing revenue back to AI visibility -- Promptwatch covers the full loop.
For e-commerce intelligence more broadly, Triple Whale handles multi-touch attribution and is building AI channel tracking into its data model.
For specialized AI shopping optimization (particularly ChatGPT's Rufus and shopping carousels), Azoma is worth evaluating if you're a larger retailer.
What to do this week
If you're starting from zero, here's a concrete first step: run 20 of your most important shopping prompts through ChatGPT and record three things for each -- whether the shopping carousel appeared, whether your brand was mentioned anywhere in the response, and which competitors appeared. That's your baseline.
It will probably be uncomfortable. Most brands discover they're invisible for a significant portion of the queries that matter to them. But that discomfort is useful -- it tells you exactly where to focus.
The brands that figure out AI shopping measurement now will have a meaningful head start. The window where this is a competitive advantage rather than table stakes is probably 12 to 18 months. After that, everyone will be doing it.