Quick answer: Measure AI search visibility with five metrics: citation rate (how often AI answers link to you), mention rate (how often they name you), AI share of voice (your share against competitors), accuracy and sentiment (whether they describe you correctly), and AI referral sessions (visits from AI platforms in GA4). Collect the first four from a fixed monthly prompt set and the fifth from analytics.
Your Search Console looks stable. Your rankings look fine. Yet leads say they “asked ChatGPT” and got a competitor. That gap exists because classic SEO reports don’t measure what AI answers say. This post gives you a measurement framework you can run with a spreadsheet, plus a free template.
Why rankings don’t show AI visibility
Rankings measure where your page appears in a list. AI answers don’t show a list — they show one synthesised response with a few sources, and often answer without any click at all. A page can rank #3 and never be cited, or rank #12 and be quoted in every answer. You need to measure the answer, not just the results page. (More on that shift in What Is Zero-Click Search.)
The five metrics
| Metric | Definition | How to collect | Cadence |
|---|---|---|---|
| Citation rate | % of tracked prompts where an AI answer links to your domain | Prompt set, logged manually or with a tracker | Monthly |
| Mention rate | % of tracked prompts where your brand is named, linked or not | Same prompt set | Monthly |
| AI share of voice | Your mentions ÷ total mentions of you and named competitors | Same prompt set, competitor columns | Monthly |
| Accuracy and sentiment | Whether answers describe your products, prices and positioning correctly, and in what tone | Read and score each answer (correct / partly / wrong; positive / neutral / negative) | Monthly |
| AI referral sessions | Sessions from chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and similar | GA4 custom channel group | Weekly |
Track citations and mentions separately. An AI answer can recommend your brand without linking to you — that’s still visibility, and it often precedes citations.
Step 1: build your prompt set
Your prompt set is the AI-era keyword list. Make it reflect what buyers actually ask, not what you’d like to rank for.
- Start with 30–50 prompts. Pull them from sales calls, support tickets, People Also Ask, and your top Search Console queries rewritten as full questions.
- Split by funnel stage: awareness (“what is…”), consideration (“best… for…”, “X vs Y”) and decision (“is [brand] good for…”).
- Mix brand and non-brand. Non-brand prompts show discovery; brand prompts show accuracy.
- Write them like real users. Include location (“in Mumbai”), budget (“under ₹50,000”) and context, because AI answers change with detail.
- Freeze the set for a quarter so month-on-month numbers are comparable.
Step 2: run it across engines
Run every prompt in ChatGPT (search on), Google AI Mode, Google AI Overviews (where they trigger), Perplexity and Microsoft Copilot. Use a logged-out or clean session where possible, and set the location to India.
Answers vary between runs, so run each prompt at least twice and record both. Log: engine, date, cited (Y/N), cited URL, mentioned (Y/N), competitors named, and an accuracy score.
Step 3: set up GA4 for AI referrals
- In GA4 go to Admin → Data display → Channel groups and create a new channel group.
- Add a channel called AI Assistants, matching session source by regex:
chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai. - Place it above Referral so these sessions aren’t counted as generic referrals.
- ChatGPT also adds utm_source=chatgpt.com to many cited links, which makes those visits easy to spot.
- Build an Exploration showing landing pages × AI source, to see which pages earn AI traffic.
Step 4: choose your tooling
- Manual (free): the spreadsheet template below. Fine for up to about 50 prompts monthly.
- Paid trackers: several platforms now automate prompt runs, citation logging and share of voice across engines. They save time at scale; check which engines and countries each covers, and whether India-specific locations are supported.
Start manual. You’ll learn what matters before you pay for automation.
Step 5: report it in one page
Leadership needs four numbers and one story each month: citation rate, AI share of voice vs your top three competitors, AI referral sessions (with conversions), and the biggest accuracy issue found. Add one action per number.
Free template
Download the AI Visibility Prompt Tracker (.xlsx). It has columns for prompt, funnel stage, engine, run date, cited, cited URL, mentioned, competitors, accuracy and notes, with share-of-voice formulas built in. [download button — file to be uploaded]
FAQ
What is AI share of voice? The share of brand mentions in AI answers that belong to you, compared with named competitors, across a fixed set of prompts.
Can Google Search Console show AI Overview clicks? Clicks from AI Overviews and AI Mode are counted within Search Console’s standard web performance data, but not broken out separately. That’s why a prompt set is still needed.
How often should I track AI citations? Monthly is enough for most brands. Weekly makes sense for fast-moving categories or during a launch.
What is a good citation rate? There’s no universal benchmark yet. Track your own trend and compare your share of voice against direct competitors on the same prompts.
What’s the difference between a citation and a mention? A citation is a link to your site in the answer. A mention is your brand named in the text, with or without a link.
Start with a baseline: run the 10-check Universal Search audit, then set up your prompt tracker.