• There is no AI equivalent of a fixed ranking position. Measure frequencies and trends.
• Track mentions and citations separately. They are different outcomes.
• Build a prompt set that reflects real customer questions across the buying journey.
• Combine three data sources: prompt tracking, analytics referrals and server logs.
• Report against competitors. Share of voice is more meaningful than a raw count.

In traditional SEO you can say "we rank third for this keyword". AI search does not work that way. Responses vary by run, by user and by platform, there is no public data on what people ask, and a brand can be influential in an answer without receiving a single click. Measurement is still possible. It just looks more like market research than rank tracking.

The core metrics

• Brand presence (mention rate): the share of responses in which your brand is named.
• Citation rate: the share of responses that link to your domain as a source.
• Share of voice: your mentions or citations as a proportion of all tracked brands.
• Position: where you appear in a list or how early in the answer.
• Sentiment: whether the description is positive, neutral or negative.
• Accuracy: whether what is said is correct.
• Cited sources: which domains and pages are used for your topics, including third parties.
• AI referral traffic and conversions.
• AI crawler activity.

Keep mentions and citations apart. Mixing them produces misleading numbers.

Building a prompt set

• Cover the journey: problem awareness, category research, comparison and brand-specific questions.
• Write prompts the way people speak to assistants, with context and constraints.
• Draw on sales calls, support tickets, on-site search, "People also ask" and community threads.
• Tag prompts by topic, funnel stage and product so you can segment results.
• Split branded from non-branded prompts.
• Keep the set stable so trends are comparable.

No tool can tell you the true volume of a prompt, as the platforms do not publish it. Treat your set as a representative sample.

Handling variability

• Run each prompt multiple times and on a regular schedule.
• Report percentages across many runs, not single outputs.
• Track each platform separately.
• Test in the right country and language.
• Look for sustained movement over weeks. Daily noise means little.

Data source 1: prompt tracking tools

A growing group of platforms automates this: running prompts across assistants, parsing mentions and citations, and charting trends against competitors. Many SEO suites have added similar modules. When choosing, check which AI platforms are covered, whether data comes from the real consumer interface or an API (results differ), how often prompts run, and whether you can export raw responses.

Data source 2: analytics

In GA4, AI traffic mostly arrives as referrals from domains such as chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai. ChatGPT also appends a utm_source parameter to many links.

• Create a custom channel group for AI platforms using a regular expression on the source.
• Report sessions, engagement and conversions for that channel.
• Look at landing pages to see which content AI sends people to.

Limits: many AI visits lose their referrer and appear as direct, particularly from apps. Visits from Google AI Overviews and AI Mode are recorded as Google organic. So analytics undercounts.

Data source 3: search engine tools

• Google Search Console includes AI Overview and AI Mode impressions and clicks within Web search totals. Check current documentation for dedicated filters.
• Bing Webmaster Tools has introduced AI performance reporting showing how often your pages are cited in Copilot and related experiences.

Data source 4: server logs

Logs show AI crawlers and user-triggered fetchers requesting your pages. User-triggered hits (ChatGPT-User, Claude-User, Perplexity-User) are a useful proxy for real-time demand, since each one means a page was pulled into someone's conversation. Track volumes by bot, top requested pages and status codes.

Connecting it to the business

• Compare conversion rates of AI referral traffic with other channels.
• Add "How did you hear about us?" to forms, with an AI assistant option.
• Watch branded search and direct traffic as indicators of exposure without clicks.
• Relate changes in visibility to the work you did, so you learn what moves the numbers.

A simple reporting framework

• Visibility: mention rate, citation rate and share of voice by platform and topic.
• Perception: sentiment and accuracy.
• Sources: top cited domains, and gaps where competitors appear and you do not.
• Traffic: AI referral sessions and conversions.
• Access: AI crawler activity and errors.

Common mistakes

• Drawing conclusions from a handful of manual prompts.
• Tracking only branded prompts, where you will naturally look strong.
• Treating synthetic prompt volumes from tools as real demand.
• Judging GEO on referral traffic alone.
• Ignoring accuracy. Being mentioned with wrong information is a problem, not a win.
• Comparing figures across tools that use different methods.

Add your title here

This is a paragraph. Writing in paragraphs lets visitors find what they are looking for quickly and easily. Make sure the title suits the content of this text.