How to measure AI shopping performance: GA4, Search Console, Merchant Center and recommendation visibility

On 13 May 2026, Google Analytics added an AI Assistant channel to its Default Channel Group. When the referrer matches a recognised AI assistant, GA4 can assign:

• medium: ai-assistant;
• channel: AI Assistant;
• campaign: (ai-assistant).

That is a real improvement over older advice telling retailers to build the whole channel definition themselves. It does not solve everything. Traffic can still lose referral information, AI influence can happen without a click, and Google's channel definitions keep visits from its own AI Overviews and AI Mode in Organic Search rather than in the new channel. So treating AI traffic as GA4's AI Assistant sessions alone is too narrow.

In June 2026, Google introduced dedicated generative-AI performance reports in Search Console for a subset of sites, and opened them to every site on 31 August 2026. The reports can show:

• impressions in generative AI features;
• pages shown;
• countries;
• devices;
• dates and trends.

That is useful for visibility in Google AI Mode and AI Overviews, but it still does not give you a cross-assistant recommendation score.

Google also announced AI performance insights for Merchant Center, including measures such as:

• share of voice on AI shopping journeys;
• shopping-funnel performance;
• product-term insights;
• product-attribute insights and completeness.

The rollout is market-limited: as of October 2026, Google's help page lists it as available for English queries in accounts in Australia, Canada, India, New Zealand and the US, so do not assume the report is in a UK account.

1. Recommendation visibility

Across a controlled set of relevant shopping prompts:

• did the product or brand appear?
• was it merely cited, or actually recommended?
• what reason did the assistant give?
• which alternatives won?

For ChatGPT, the ranking factors guide lists what OpenAI documents about the selection, so the reason an assistant gives can be read against it.

2. Platform visibility

Use Search Console, Merchant Center and whatever other reporting the platforms expose.

3. Traffic

Use GA4's AI Assistant channel, source and medium analysis, landing pages, and server or referrer evidence where relevant.

Sighted's pilot with Onboard Sportswear read across three of the four. The work consolidated duplicate URLs, submitted the site's first sitemap, addressed indexing gaps and enriched the products. On the first layer, AI citation coverage went from 0% to 33% of tested responses; on the second, impressions in Google's AI features rose 1,140%; on the third, organic clicks rose 59%.

4. Commercial outcome

Measure orders, revenue, conversion and, where defensible, agentic or AI-sourced and AI-influenced sales.

The goal is to build enough evidence to answer one question, without forcing every sale into an “AI” bucket: when you improve product information for a defined set of shopping needs, do recommendation visibility and commercial outcomes improve against a baseline or control?

The Bean Coffee Company pilot gives the fourth layer its shape. The pilot products were enriched, and their revenue rose 31% against 10% across the rest of the catalogue, with search clicks going from 121 to 366. A rise read against the rest of the same store answers the question above; a rise on its own does not.

Track:

• AI Assistant sessions and revenue in GA4;
• Google generative-AI impressions where available;
• Merchant Center AI visibility and share of voice where available;
• prompt-set visibility and recommendation rate;
• product-level enrichment coverage;
• click-through where measurable;
• conversion and revenue for treated products;
• comparison or control performance;
• the dates of product-data changes.

1. Define a controlled set of relevant shopping prompts and record, for each, whether the product appeared, whether it was cited or recommended, the reason given and which alternatives won.
2. Read GA4's AI Assistant channel for traffic, with source and medium, landing pages and referrer evidence, and remember that AI Overviews and AI Mode visits stay in Organic Search.
3. Open Search Console's generative-AI performance reports for impressions in Google's AI features.
4. Check whether Merchant Center's AI performance insights are available in your account; as of October 2026 they are not listed for the UK.
5. Measure orders, revenue and conversion for the treated products against a comparison or control set.
6. Keep the dashboard above with the dates of every product-data change, so each change can be read against the signals.

The free AI visibility audit is the baseline for the first layer's inputs: what each of up to 50 product pages states, and what it leaves out. On paid plans Sighted then asks the assistants a fixed set of shopper questions on a schedule and records whether your store is named, so a change can be read against the traffic and sales signals you already trust. The change it measures is its own enrichment of the catalogue, the technology behind SO-ME.