---
title: "AI Shopping Agents and Ecommerce Visibility"
description: "How AI shopping agents discover and recommend products, what signals they use, and how to measure AI-driven ecommerce referrals."
tldr: "AI shopping agents (ChatGPT, Perplexity, Microsoft Copilot, Google Gemini) route product discovery for a measurable and growing share of ecommerce transactions. In its Q2 2026 results Shopify reported AI-driven traffic and orders to its stores both tripled year on year, down sharply from the eightfold and thirteenfold growth it reported a quarter earlier. The signals AI agents use to select products mirror traditional ecommerce SEO: complete product descriptions, accurate structured data, strong reviews, and return and shipping policies. Visual search adds a parallel channel, with Google Lens handling more than 25 billion visual searches a month."
publishDate: 2026-08-22
updatedDate: 2026-09-18
author: "Liam Hayward"
---
AI shopping agents are AI assistants, such as ChatGPT, Perplexity, Microsoft Copilot and Google's Gemini and AI Mode, that recommend products inside an answer and send the shopper on to buy. They draw on the same data as traditional ecommerce SEO, product structured data, feeds, reviews and store policies, but a recommendation can drive a purchase without the user ever browsing the merchant's site in the conventional sense. Understanding which platforms are involved, what signals they use, and how to track their contribution is now part of ecommerce SEO.

## Which AI platforms surface product recommendations?

**Google AI Mode and AI Overviews** surface product recommendations drawn from the Shopping Graph, fed by Merchant Center data and product page structured data. For queries with product intent, Google's AI surfaces draw on Merchant Center listings directly, not solely on organic rankings. Shopping eligibility and catalogue completeness matter as much as organic search optimisation here.

**ChatGPT Shopping** is available via ChatGPT's web browsing features and its dedicated shopping interface. Shopify's Agentic Storefronts connect Shopify merchant catalogues into ChatGPT, Google AI Mode and Gemini, Microsoft Copilot, and Meta, allowing product discovery within the conversation interface. The channel is active by default for eligible stores rather than something a merchant enrols in. For ChatGPT specifically that is all it is: Shopify describes it as "a discovery-focused referrer platform", where the purchase completes in the merchant's own checkout rather than in the conversation, so a ChatGPT product placement is a referral to optimise for, not a transaction to instrument.[^7]

**Perplexity Shopping** surfaces product recommendations within answers to product-intent queries, drawing on retailer data through direct partnerships and real-time web retrieval.

**Microsoft Copilot** integrates product search into its conversational interface via Bing's product index.

In its Q1 2026 results, Shopify reported AI-driven traffic up roughly eightfold year on year and orders from AI-powered search up around thirteenfold over the same period.[^1] A quarter later the picture had changed. On the Q2 2026 earnings call, Shopify president Harley Finkelstein said AI-driven traffic and orders to Shopify stores had both tripled year on year, so the gap between the two rates had closed and both had decelerated sharply.[^5]

Finkelstein put that alongside a second point from the same call: traditional search sessions are up 1.3 times over two years and still hold roughly a third of all storefront sessions, which he framed as AI complementing search rather than replacing it. He also twice noted that agentic volume remains small against a very large GMV. Absolute volumes rose; it is the year-on-year multiple, measured off a much larger base, that fell.[^5]

Finkelstein gave three further figures for the quarter: new buyer orders arriving at nearly twice the rate of other channels, 75% of AI-attributed orders coming from outside Shopify's top 100 categories, and AI searches served by Shopify's Catalog converting at twice the rate of those relying on scraped data.[^5] The last of those is a direct argument for supplying structured catalogue data rather than leaving agents to scrape a storefront. The second suggests the channel favours specialised and long-tail merchants over the largest categories.

## What signals do AI shopping agents use to select products?

The two platforms that publish a product-data contract, OpenAI's Agentic Commerce feed and Google's Merchant Center, ask for the same things traditional ecommerce SEO already needs:[^14][^13]

**Complete product descriptions** that answer buyer questions. Descriptions written to anticipate what a purchaser needs to know (what it is, who it is for, key specifications, what it is not) are more likely to match the natural language queries AI systems receive. Thin manufacturer-supplied copy that answers none of these questions is not a strong retrieval signal. OpenAI's own feed guidance asks merchants to "use concise, factual copy that helps users understand products", and accepts plain text or bullet-style text.[^12]

**Accurate, consistent structured data:** Product, Offer, MerchantReturnPolicy, OfferShippingDetails, and AggregateRating aligned with live pricing and availability. OpenAI's feed spec asks for current price and availability on every row and for review counts and ratings that "describe the same review population"; inconsistencies between structured data, feed data and the visible page are what those rules exist to catch.[^14]

**Review volume and recency:** OpenAI's feed contract carries `review_count` and `star_rating` per product, with the instruction to omit a rating where there are no reviews, so a product without reviews is submitted without one.[^14] Whether a rating changes which product is recommended is not documented.

**Return and shipping policy completeness:** the same feed carries `accepts_returns`, a return window in days, a policy URL and shipping charges, with omitted shipping meaning "unknown, not free shipping".[^14] A product with no policy data is submitted as unknown on every one of those fields.

**Catalogue taxonomy accuracy:** correctly categorised products match more query patterns than products assigned to vague or inaccurate categories. OpenAI lists category taxonomy among the optional feed fields that "can improve answer quality", alongside HTML and Markdown descriptions and seller links.[^12]

Adobe put a number on how much of that content is actually machine-readable. Its AI Content Visibility Checker scores a page for the share of its content an LLM can read: across US retail, homepages averaged 75% and category pages 74%, but individual product pages came in at 66%, the worst-scoring template of the set. Adobe's reading is that "retailers have thousands of SKUs, and our data shows that much of the content is currently invisible to LLMs".[^9] The product template is both the weakest and the one an agent is most likely to land on.

## How does visual search fit in?

Visual search is a parallel product discovery channel. Google Lens handles more than 25 billion visual searches a month, of which Google puts 20% at commercial intent, figures given by Dan Taylor, Google's VP of global advertising, at a press briefing in March 2026.[^3] Both are Google's own internal numbers rather than independently audited ones. A user photographing a product in a shop, a friend's home, or in a magazine can use Lens to find it for purchase.

For ecommerce visibility in visual search:

- High-quality product images with clean, uncluttered backgrounds perform better in visual matching.
- Multiple image angles increase the chance that the product is matchable from different visual perspectives.
- Images submitted to Google Merchant Center with accurate category data improve Shopping Graph matching for visual queries.
- Alt text on product images contributes to the textual indexing of visual content, connecting visual searches to relevant product pages.

Visual search and AI shopping are converging: Google's AI Mode surfaces visually matched products alongside conversational recommendations for product-intent queries.

## Do AI-referred visitors convert better?

Yes, on every panel that has measured it, and by less in ecommerce than the cross-industry headlines suggest. The widely quoted four-to-five-times multiple is a cross-industry average, with ecommerce at the bottom of its range at roughly 1.3 times.[^2] The study behind that low end, Visibility Labs' analysis of 94 seven- and eight-figure stores across 2025, put ChatGPT referrals at 1.81% conversion against 1.39% for non-branded organic, a 31% lift, with non-branded organic still 70 times larger by sessions.[^11] Adobe Analytics' US retail series has AI-referred traffic converting 60% higher than non-AI traffic in July 2026, the eleventh consecutive month it outperformed after starting 38% worse in March 2025.[^8][^9] Both are panel measurements by companies selling into the market they measure, and the Visibility Labs window closes before the August 2026 change in how ChatGPT selects sources. The likely explanation is selection: an agent surfaces products that match a stated intent, so its visitors arrive closer to a decision than a category browser does.

## How do you measure AI shopping referrals?

Measuring AI-driven ecommerce traffic requires combining several sources:

**GA4 channel groups:** Google Analytics 4 added 'AI Assistant' as a default channel group in May 2026, automatically categorising AI-referred traffic. Google's channel definition names ChatGPT, Gemini, Deepseek, Copilot and Grok, and states that the channel excludes Google's own AI Overviews and AI Mode, which continue to land in Organic Search.[^6] Perplexity was confirmed via a Google Analytics LinkedIn post in June 2026 but has not been added to the docs.[^4] The exclusion matters for an ecommerce store: the two Google surfaces at the top of this page are the ones GA4's AI channel will not show you. Check your own reports rather than assuming: if a platform you care about is still landing in Referral, add a custom channel definition (for example, `session_source contains perplexity`) so it appears alongside the other AI channels.

**Merchant Center AI performance insights:** this is the first-party report for the surfaces GA4 excludes. Google describes it as "a transparent view of how your brand is discovered across Google's generative AI ecosystem, specifically revealing your performance for conversational queries with shopping intent on AI Mode and AI Overviews", including a share-of-voice metric measuring your AI impressions against competitors' for related queries. It has since left its US pilot: Google now says the report is "currently available for English-language queries for Merchant Center accounts in Australia, Canada, India, New Zealand, and the United States".[^10] Where you have access, start here rather than with GA4.

**Google Merchant Center performance reports:** Merchant Center reports clicks from Shopping surfaces separately from GA4 organic data. As AI Mode and Shopping AI Overviews become more prominent, Shopping referral data becomes increasingly relevant alongside standard organic metrics.

**Shopify's agentic attribution reporting:** Shopify shipped cross-channel attribution for agentic selling into the admin in Q2 2026, letting merchants manage AI channels and track their performance in one place.[^5]

**Branded search volume as a proxy:** AI citations compound into branded search. A rising trend in branded impressions in Google Search Console, correlated with AI visibility tool citation data, is a proxy signal for AI-driven brand discovery that has not yet converted to a directly attributable session.

[^1]: [AI drove orders at Shopify up 13 times in Q1 — PYMNTS](https://www.pymnts.com/earnings/2026/ai-drove-orders-shopify-up-13-times-q1/)
[^2]: [Why AI search traffic converts at 4-5×: what the data actually shows — Pixis](https://www.pixis.ai/blog/why-ai-search-traffic-converts-at-4-5x-what-the-data-actually-shows/)
[^3]: [Google VP: "AI is changing search... over 25 billion searches per month with Google Lens" — The Asia Business Daily](https://www.asiae.co.kr/en/article/2026031214442621001)
[^4]: [Google Analytics AI Assistant traffic measurement roll-out — Google Analytics LinkedIn](https://www.linkedin.com/posts/were-rolling-out-brand-new-automated-%F0%9D%90%80-ugcPost-7470508409834942464-SyJ8/)
[^5]: [Shopify (SHOP) Q2 2026 earnings call transcript — The Motley Fool](https://www.fool.com/earnings/call-transcripts/2026/08/12/shopify-shop-q2-2026-earnings-call-transcript/)
[^6]: [\[GA4\] Default channel group — Google Analytics Help](https://support.google.com/analytics/answer/9756891)
[^7]: [Agentic storefronts — Shopify Help Center](https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts)
[^8]: [Adobe: AI-referral traffic spending, converting more than others — Digital Commerce 360](https://www.digitalcommerce360.com/2026/08/19/adobe-ai-referral-traffic-data-july-2026/)
[^9]: [Adobe report: U.S. retailers see surge in AI traffic, but many websites are not entirely readable by machines — VMblog](https://vmblog.com/news/adobe-report-u-s-retailers-see-surge-in-ai-traffic-but-many-websites-are-not-entirely-readable-by-machines/)
[^10]: [AI performance insights — Google Merchant Center Help](https://support.google.com/merchants/answer/17200695)
[^11]: [ChatGPT traffic converts 31% better than non-branded organic search (94 eCommerce sites analyzed) — Visibility Labs](https://visibilitylabs.com/blog/chatgpt-vs-organic-search-conversion-rates/)
[^12]: [Best practices — OpenAI Developers, Agentic Commerce Protocol](https://developers.openai.com/commerce/guides/best-practices)

[^13]: [Optimizing your website for generative AI features on Google Search — Google Search Central](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide), the local business and ecommerce section: Merchant Center feeds and Business Profiles "can help your products and services to be visible in both AI responses and other Google Search results".
[^14]: [Product feed spec — OpenAI Developers, Agentic Commerce Protocol](https://developers.openai.com/commerce/specs/feed). The review, returns and shipping attribute tables, as of 18 September 2026.