How do you measure traffic and sales from AI assistants on a fashion website?

To measure traffic and sales from AI assistants, create a dedicated channel in your analytics for referrals from assistants such as ChatGPT, Perplexity, Gemini and Copilot, connect it to orders and revenue, and add the citation and search data the platforms themselves publish. That captures the visible part. The rest, shoppers who read an answer and later arrive directly or in store, has to be estimated with surveys and trend analysis.

Why is AI assistant traffic hard to measure?

There are three reasons. First, many assistant answers satisfy the question without any click. Second, some clicks arrive without a clear referrer, for example from apps, so they land in direct traffic. Third, the largest AI surface for most fashion sites, Google's AI Overviews and AI Mode, is reported inside normal Google search: Google states that this traffic is included in the overall Search traffic in Search Console's Performance report under the Web search type.

The volume is growing from a low base. Adobe reported that traffic from generative AI tools to US retail sites rose 693.4% year on year in November and December 2025, based on Adobe Analytics data covering over 1 trillion visits, while noting that the base of users remains modest. For most fashion brands this is still a small channel in sessions, which is exactly why it gets lost in default reports.

How do you identify AI referrals in Google Analytics 4?

Start with what the platforms tell you. OpenAI states that "ChatGPT automatically includes the UTM parameter utm_source=chatgpt.com in referral URLs", so ChatGPT search clicks appear with that source. Other assistants usually appear as referral sources with their own domains.

In Google Analytics 4, create a custom channel group with a channel for AI assistants. Google's documentation describes this use case directly and confirms the main mechanics:

• Custom channel groups are created under Admin, Data display, Channel groups, and need Editor access.
• Standard properties can have two custom channel groups in addition to the default; 360 properties can have five.
• Rules can use regular expressions, so one channel can match a list of assistant domains.
• Traffic is assigned to the first channel whose definition it matches, so the AI channel must sit above the generic Referral channel.
• Custom channel groups can be applied to reports retroactively.

Maintain the list of domains as a governed asset. Assistants change domains and launch new apps, and a stale regex silently moves traffic back into Referral.

Read also

How does AI personalisation work in fashion e-commerce?

Which metrics should a fashion brand report?

Report returns alongside revenue. In fashion a channel that converts well but returns heavily is not necessarily profitable, and AI-referred shoppers may behave differently once they have compared options in an assistant.

What can platform reports tell you about citations?

Microsoft launched an AI Performance report in Bing Webmaster Tools in February 2026, in preview. It shows how often a site is cited as a source in AI-generated answers across Microsoft Copilot, Bing's AI summaries and select partner integrations, which pages are cited, and sample grounding queries. It does not show clicks. Use it as a leading indicator alongside your own prompt audits, not as a traffic report.

How do you capture influence that never clicks?

1. Add an option such as "AI assistant (for example ChatGPT, Gemini)" to the "How did you hear about us?" question in checkout or post-purchase surveys.
2. Track branded search impressions in Search Console and compare them with periods of improved AI visibility.
3. Ask store staff to log when customers mention an assistant recommendation, for a defined test period.
4. Run controlled content changes on a subset of product pages and compare citation and referral trends against a control group.

None of these methods is precise on its own. Together they give management a reasonable view of whether AI visibility is moving the business, without pretending to an attribution accuracy the data cannot support.

Product-level views are often more useful than totals. Look at which landing pages receive AI referrals: if they are mainly policy, sizing or care pages, assistants are using the site to answer questions, which is a sign of trust but also a sign that product pages may be missing the same facts.

Read also

Do AI stylists sell? The metrics to judge a conversational shopping assistant

How should the setup be governed?

Build one dashboard that places these views side by side for management: citations and share of answers as leading indicators, AI referral sessions and revenue as the visible outcome, and survey responses as a check on the halo effect. Keep the definitions on the dashboard itself, so readers can see what each number includes and excludes.

Assign one owner for the channel definition and domain list, document the rules, and annotate reports when definitions change. Check server logs periodically for assistant user agents fetching pages, because a rise in fetches without a rise in referrals is itself a signal that answers are using your content without sending visits. Review the framework every quarter, as assistants are adding shopping, checkout and reporting features at a fast pace.