I had a conversation last month with a SaaS founder in Bengaluru who was convinced their AI SEO was working — because organic traffic was up 20%. When I asked how they were tracking their LLM citations specifically, they looked at me blankly. "That's in analytics, right?"
It isn't. And that gap — between what you can see in GA4 and what's actually happening in AI search — is where most brands are flying blind. Let me fix that.
Why standard analytics miss AI search traffic
When someone discovers your brand through a ChatGPT or Perplexity answer, they rarely click a link in that response. Instead, they open a new browser tab and Google you — or type your URL directly. That journey shows up in GA4 as direct traffic or branded search, not as a referral from an AI engine.
This is the core measurement problem with LLM search. The attribution chain is broken. You need a different measurement framework entirely — one that works upstream of the click.
KPI 1: Citation Rate
Definition: The percentage of your target queries where your brand or website is cited in at least one AI engine's response.
This is the single most important metric. If you're not cited, nothing else matters. How to measure it:
- Define your top 20–30 target queries (these are the questions your ideal customers ask when researching solutions like yours)
- Run each query through ChatGPT, Perplexity, Claude, and Gemini once a week
- Record whether your brand is mentioned and whether your URL is cited as a source
- Track this as a % over time: (queries where cited ÷ total queries) × 100
A well-optimised brand in a competitive niche should target 40–60% citation rate on their core queries. Starting from scratch, 10–15% in the first 90 days is a realistic early win.
KPI 2: Citation Position
Definition: When your brand is cited, where does it appear in the AI response — primary source, secondary source, or passing mention?
Being the first source cited in a Perplexity answer is worth dramatically more than being mentioned in passing at the end of a response. Develop a simple scoring system:
- Score 3 — Primary source, cited in the direct answer paragraph
- Score 2 — Secondary source, cited in supporting detail
- Score 1 — Passing mention without source citation
Track your average citation score across your query set. This tells you whether you're gaining depth of authority, not just breadth of mentions.
KPI 3: Branded Search Uplift
Definition: Month-over-month growth in branded Google searches (your company/product name), used as a proxy for LLM-driven awareness.
This is measurable via Google Search Console. Filter for queries containing your brand name and track impression and click trends. When AI citations increase, branded search reliably follows within 2–4 weeks. I've seen this pattern consistently across client accounts.
KPI 4: LLM Referral Traffic
Definition: Direct, trackable traffic from AI surfaces that do generate clicks.
Some AI engines do send measurable traffic:
- Perplexity — shows up as referral from perplexity.ai in GA4
- ChatGPT browsing mode — shows up as referral from chat.openai.com
- Claude.ai — shows up as referral from claude.ai
- Google AI Overviews — typically attributed to organic Google traffic (harder to isolate)
Create a custom channel grouping in GA4 called "AI Search" that aggregates these referral sources. Track sessions, engagement rate, and goal completions from this channel.
KPI 5: Share of Voice in AI Answers
Definition: Your citation rate relative to your top competitors across the same query set.
This is the competitive intelligence layer. If you're cited in 30% of your target queries but your main competitor is cited in 70%, you have an AI visibility gap that directly translates to lost pipeline. Run competitor queries alongside your own and build a share-of-voice scorecard quarterly.
Want a custom AI visibility scorecard for your brand?
I'll run your top 20 queries across 4 AI engines and show you exactly where you stand versus competitors — free in the audit call.
Building your measurement stack
You don't need expensive tools to start. Here's a lean measurement setup that works:
- Google Sheets — manual citation tracking spreadsheet (query × AI engine × score)
- Google Search Console — branded search uplift tracking
- GA4 custom channel grouping — LLM referral traffic
- Perplexity Pro (optional) — faster manual query testing
Once you have 4–8 weeks of baseline data, you can start making decisions about where to invest your GEO effort. Measurement first. Strategy second. Always.