For UK ecommerce businesses, AI search is no longer just a branding experiment. Customers are asking ChatGPT, Gemini, Microsoft Copilot, Perplexity and similar tools for product recommendations, comparisons and buying advice, then clicking through to stores that appear in the answers.
The measurement challenge is that a sale influenced by AI is not always a sale attributed to AI. A customer may research in a chatbot, later search your brand on Google, return through a saved tab or type your URL directly. In analytics, that order may be reported as organic search or direct traffic, even though the buying journey began in an AI conversation.
That does not make AI traffic impossible to measure. It means ecommerce teams need a disciplined reporting model: capture identifiable AI referrals, keep organic search distinct, protect the integrity of direct traffic, and clearly label what remains unknown. Done properly, this gives a small business a practical view of sales coming from AI tools without making claims that the data cannot support.
Why AI attribution needs a different approach
Traditional channel reporting assumes that the source which sends the visitor to your website is reasonably visible. A paid ad can carry click identifiers, an email can contain UTM parameters, and a search engine generally appears as an organic source. AI-assisted journeys are more complicated because the recommendation, evaluation and final click can occur at different times and in different apps.
There is still real value in measuring the visits you can identify. Shopify’s analysis of Q1 2026 storefront data found that tracked referral sessions from AI chatbots grew more than eightfold year on year. It also found that AI-referred sessions often landed on product pages and, in its dataset, converted more strongly than organic-search sessions. That makes AI-referral reporting commercially useful, particularly for retailers with considered purchases, technical products or niche ranges. Read Shopify’s AI search analysis.
However, do not turn that signal into a sweeping conclusion that every sale with an AI-influenced customer journey belongs in an “AI” channel. Google Analytics now identifies recognised chatbot referrals as an AI Assistants channel, while Google’s own AI Overviews and AI Mode are classified under Organic Search. In other words, two customers who both used AI may legitimately appear in two different reporting buckets. Google’s default channel-group documentation makes this distinction explicit.
The answer is not to force every AI-related visit into one number. It is to report three core channels separately: identifiable AI referrals, organic search and direct or unknown traffic. Add an “AI-influenced” indicator only as a supplementary management measure, not as a replacement for source-based attribution.
Start with a clear channel taxonomy
Before changing tags or dashboards, agree on plain-English definitions. Your marketing manager, agency, finance lead and web developer should use the same language. This prevents a common mistake: reporting a rising direct-traffic figure as proof that chatbot discovery is working.
1. AI referrals: traffic you can identify
This channel contains sessions where the landing visit has an identifiable AI-assistant source, referral or campaign tag. It includes clicks that arrive from services such as ChatGPT, Gemini, Copilot, Perplexity, Claude or other AI tools when the referral details survive the journey into your analytics platform.
For example, OpenAI says ChatGPT adds utm_source=chatgpt.com to referral URLs from ChatGPT search, which gives publishers a useful and relatively clean tracking signal. OpenAI’s publisher guidance confirms that this parameter can be used in analytics tools to analyse inbound traffic. Your reports should preserve that source rather than allowing it to disappear inside a broad referral bucket.
AI referrals are the channel to use for statements such as: “Orders directly attributed to chatbot clicks,” “Revenue from recognised AI assistant referrals,” and “Conversion rate of visitors who arrived from an AI tool.” It is a measured, auditable channel.
2. Organic search: search-engine visits, including some AI-led discovery
Keep Organic Search separate. It should include traditional unpaid search clicks from Google, Bing and other search engines, plus AI-generated Google search experiences that your analytics platform classifies as organic. Google documents that clicks from AI Overviews and AI Mode fall within Organic Search rather than AI Assistants. See the channel definitions here.
This may feel unsatisfying when your team knows a buyer saw an AI Overview. But it is more honest than creating a false split that your data cannot reliably support. In monthly reporting, describe this group as “Organic Search, including search-engine AI experiences where source data does not distinguish them.”
3. Direct and unknown: preserve the uncertainty
Direct traffic is not a synonym for loyal customers, nor is it a hidden AI channel. In GA4, direct traffic means Analytics has no clear referral source. Missing UTM tags, privacy controls, browser or app behaviour, redirects, offline documents and ad blockers can all contribute to direct classification. Google’s explanation of direct traffic is worth sharing with everyone who reads the dashboard.
Report this group as Direct / unknown in management commentary. That wording makes the limitation visible. Do not reassign a percentage of direct revenue to AI just because chatbot traffic is increasing. You may develop an internal estimate later, but it must remain clearly separate from attributed revenue.
Set up the measurement foundation
Reliable reporting starts with basic ecommerce tracking. Confirm that your analytics property records product views, add-to-basket actions, checkout starts and purchases, and that purchase events include order value, currency, transaction ID and the products sold. Reconcile total online revenue in your analytics platform against Shopify or your ecommerce platform each month. The totals will not always match exactly because of consent choices, payment flows and timing, but large or persistent gaps need investigation.
In GA4, use the Traffic acquisition report for session-level channel performance. It is designed to show where new and returning sessions came from, and includes dimensions such as Session source, Session medium, Session source / medium and Session default channel group. Google’s GA4 Traffic acquisition guide explains the available dimensions and metrics.
Set purchases as a key event and include total revenue, transactions, conversion rate and average order value in your working report. Also add engagement rate, product-view rate, add-to-basket rate and checkout-start rate. These supporting measures show whether a weak sales figure reflects poor traffic quality, an unsuitable landing page, a stock problem or checkout friction.
Create an AI-referral source list
Do not rely on one channel label alone. Export several weeks of GA4 session-source data and search for known AI services, unusual referrals and campaign values. Start an internal source register with four fields: source value seen in analytics, service name, reporting channel and date checked.
For instance, values clearly associated with ChatGPT can be mapped to ChatGPT within AI referrals. Do the same for identifiable Gemini, Copilot and Perplexity sources. Keep the original source value in the data, even when your dashboard rolls it into a parent AI-referrals line. This lets you answer the next question quickly: which assistant is sending traffic, revenue and high-value orders?
Review the list monthly. AI services, browser integrations and analytics channel definitions change quickly. A source that arrives as a referral today may be recognised automatically as an AI assistant later. Source governance matters more than a long, untested list of presumed domains.
Use UTMs where you control the link
You cannot add UTM parameters to links an AI tool chooses to cite organically. You can, however, tag links that your business deliberately distributes. This includes a product link given to an affiliate, a creator campaign, a PR partner, a webinar slide, a downloadable buying guide, an email sequence or a paid placement within an AI-related product.
Use a consistent convention, such as utm_source=partnername, utm_medium=referral and utm_campaign=spring_product_launch. If you run a specific collaboration with an AI platform or a chatbot-based assistant, use a transparent medium such as ai-assistant and a campaign name that describes the activity. Never use “AI” as a catch-all tag on links that were actually shared through email, paid social or influencer activity; that would overwrite useful source data.
Check that UTMs persist through consent screens, redirects, subdomains, checkout and payment-provider returns. A test order from each major landing-page route is more valuable than assuming the implementation works.
Build a simple monthly reporting model
A small business does not need an elaborate attribution warehouse to begin. Create one monthly acquisition report using session-level source data and a separate order-level reconciliation from your ecommerce platform. Keep the report stable for at least a quarter so trends become meaningful.
The three core channel lines
- AI referrals: Sessions, orders and revenue from recognised AI assistants, identifiable chatbot referrers and correctly tagged AI-assistant campaigns. Break this down by ChatGPT, Gemini, Copilot, Perplexity and “other identified AI” beneath the total.
- Organic search: Unpaid search-engine traffic. Include a note that Google AI Overviews and AI Mode are treated as organic by GA4, so this line may contain AI-assisted search discovery that cannot be isolated reliably.
- Direct / unknown: Sessions and orders with no clear source. Do not claim these are AI sales. Track the trend and investigate material changes, especially after website, consent-management or campaign changes.
For each line, report sessions, percentage of total sessions, orders, conversion rate, revenue, percentage of total revenue, average order value and top landing pages. Add the previous month and the same month last year where the business has sufficient history. Percentages are essential because an increase in AI-referral revenue may simply reflect overall sales growth.
Then add a short commentary block called Attribution notes. It should explain any tracking changes, new sources added to the AI register, major direct-traffic movements and whether organic growth was led by branded or non-branded search. This makes the numbers useful for decision-making rather than merely decorative.
Add an AI-influenced view, but keep it separate
Senior teams will reasonably ask, “What is AI doing beyond the clicks we can see?” Answer with a second, clearly labelled view rather than changing the core channel totals.
Your AI-influenced evidence can include customer survey responses at checkout, post-purchase surveys asking “How did you first hear about us?”, branded-search growth following a period of AI visibility, increases in direct traffic alongside no other major activity, and qualitative checks of whether your products are being recommended for relevant prompts. None of these proves order-level causation on its own. Together, they help build an informed picture.
Use wording such as: “£X was directly attributed to identifiable AI referrals. Additional AI influence is likely but cannot be quantified reliably from web analytics alone.” That is a far stronger statement than presenting an inflated “AI revenue” number.
Find the pages and products AI traffic actually buys
Channel revenue is only the beginning. Segment AI-referral sessions by landing page, device, new versus returning customer, product category and order value. If ChatGPT visitors predominantly land on a specific product page and convert well, the commercial action may be to improve stock availability, delivery information, comparison content and on-page answers for that product—not simply to chase more traffic.
Compare the AI-referrals segment with Organic Search on the same landing pages where possible. This reduces the risk of attributing better conversion entirely to the channel when the real explanation is that AI visitors are landing on your strongest product page. Shopify’s research suggests AI-referred visitors often arrive directly at product-detail pages, which is another reason page-level analysis matters. Shopify’s findings on AI-referred product-page sessions.
For example, a British outdoor-equipment retailer might find that Copilot and Perplexity referrals produce few sessions but a high rate of visits to a waterproof-jacket page. Organic search might deliver far more sessions to category pages but a lower conversion rate. The sensible conclusion is not that AI has replaced SEO. It is that the channels perform different jobs: organic search brings broader discovery, while identifiable AI referrals may bring shoppers who have already narrowed their choice.
Common mistakes to avoid
- Combining AI referrals and organic search: This hides the operational difference between chatbot clicks and search-engine traffic, and makes performance harder to interpret.
- Calling all direct sales “dark AI”: Direct is inherently uncertain. Treat it as unknown unless you have additional evidence.
- Using last-click revenue as the whole story: It is useful for accountability, but it misses research that happened before the visit. Pair it with surveys and landing-page analysis.
- Changing source rules without documenting them: A sudden improvement may be a classification change, not a commercial gain. Record rule changes and report them.
- Reporting only visits: AI traffic can be small but commercially meaningful. Orders, revenue, average order value and product mix should lead the discussion.
Conclusion: measure what you know, label what you do not
AI search and chatbots are creating another route into ecommerce, but they do not remove the need for sound analytics. The practical approach for UK small businesses is to capture identifiable AI referrals, retain Organic Search as its own channel, treat Direct as direct or unknown, and use surveys and trend analysis to understand wider AI influence.
Start this month: audit your traffic sources, verify purchase tracking, create an AI-referral register and build the three-line report. Within a few reporting cycles, you will have a defensible view of where AI is already generating measurable orders—and the evidence needed to decide where to invest next.
