In short
- Traditional SEO metrics (rankings, impressions) don't work for AI-search
- Focus on: mention rate, citation rate, share of voice, correctness
- Manual prompt testing remains essential
- Automation is limited but growing
- Link AI metrics to business outcomes
Your Google Analytics dashboard tells you nothing about your AI visibility. Rankings in Search Console? Irrelevant for ChatGPT. We need new metrics.
Why traditional metrics don't work
Rankings
AI doesn't give "position 1". There's no ranked list. You're cited or not.
Impressions
AI tools don't show impressions to publishers. You don't know how often your content is "seen".
CTR
Many AI answers don't lead to clicks. The value is in the mention, not the click.
Traffic
AI referral traffic is limited measurable and often low, even with good visibility.
The metrics that do work
1. Mention Rate
What it is: Percentage of relevant prompts where your brand is mentioned.
How to measure:
- Define your prompt test set (20-50 prompts)
- Test monthly in ChatGPT, Perplexity, Gemini
- Document per prompt whether you're mentioned
Formula: (Prompts with mention / Total prompts) × 100
Example: 12 of 40 prompts = 30% mention rate
Benchmark:
- 0-10%: Urgent improvement needed
- 10-30%: Room for growth
- 30-50%: Good position
- 50%+: Strong
2. Citation Rate
What it is: Percentage of mentions where you're also cited as source (with link).
How to measure: In Perplexity you can see this directly. In other tools: check if your website is explicitly mentioned/linked.
Formula: (Mentions with citation / Total mentions) × 100
Example: 8 of 12 mentions have citation = 67% citation rate
Why important: Citations deliver direct traffic and are a stronger trust signal.
3. Share of Voice
What it is: How often you vs competitors are mentioned in your category.
How to measure: Test comparative prompts ("best X agencies", "top Y services") and count who's mentioned.
Formula: (Your mentions / Total mentions in category) × 100
Example: In 10 category prompts you're mentioned 4x, competitor A 6x, competitor B 3x. Your share: 31%.
4. Correctness Score
What it is: How correct is the information AI gives about you?
How to measure: Rate each mention on a scale of 1-5:
- 1 = Completely incorrect
- 2 = Mostly incorrect
- 3 = Partially correct
- 4 = Mostly correct
- 5 = Completely correct
Why important: Incorrect mentions can damage your brand.
5. Sentiment
What it is: Is the mention positive, neutral or negative?
How to measure: Rate each mention:
- +1 = Positive
- 0 = Neutral
- -1 = Negative
Average sentiment score indicates how AI positions your brand.
6. AI Referral Traffic (limited)
What it is: Traffic from AI tools to your website.
How to measure in GA4:
- Acquisition → Traffic source
- Filter on: chat.openai.com, perplexity.ai, etc.
Note: This is limited measurable. Many AI users don't click through.
Reporting template
Monthly AI-search scorecard
| Metric | Last month | This month | Δ |
|---|---|---|---|
| Mention rate | 25% | 30% | +5% |
| Citation rate | 50% | 67% | +17% |
| Share of Voice | 28% | 31% | +3% |
| Correctness | 4.1 | 4.3 | +0.2 |
| Sentiment | 0.4 | 0.5 | +0.1 |
Trend visualization
Plot these metrics monthly in a chart. Rising lines = progress.
Tools and automation
What you can automate now
- Basic prompt scheduling (with scripts)
- Screenshot captures
- Spreadsheet templates
What still needs to be manual
- Assessment of answers
- Correctness scoring
- Context analysis
Emerging tools
Tools are emerging that automate AI tracking. The market is still young. Manual validation remains important.
What changed in 2026
AI tracking tools have matured
Where you did everything manually in 2024, specialized tools now automatically track mention rate and share of voice across ChatGPT, Perplexity and Gemini. Manual sampling remains necessary as validation, but the baseline can be automated.
AI referral traffic is becoming more measurable
More and more AI tools send a recognizable referrer. In GA4 you can bundle traffic from chatgpt.com, perplexity.ai and copilot.microsoft.com with a custom channel group. Volumes are still modest, but the trend is rising.
Correctness becomes a boardroom metric
As more decision-makers trust AI answers, "what does AI say about us?" becomes a reputation issue. Incorrect claims about pricing, services or positions require active monitoring, not an annual check.
Your first measurement in 30 minutes
- Write 20 prompts your ideal customer would ask (e.g. "best marketing agency for B2B SaaS in Belgium").
- Test them in ChatGPT, Perplexity and Gemini. Note per prompt: mentioned yes/no, cited yes/no, correct yes/no.
- Calculate your baseline. Mention rate = mentioned / 20. Citation rate = cited / mentioned.
- Record in a spreadsheet with date, tool and prompt. This is your baseline measurement.
- Repeat monthly with exactly the same prompts. Only then do you see real progress.
Link to business outcomes
Ultimately you need to link AI metrics to what matters:
Leads
Add "How did you find us?" to forms with option "AI tool (ChatGPT, etc.)"
Brand awareness
Periodic surveys: "Have you seen our brand in AI answers?"
Sales conversations
Track if prospects mention AI tools in their research process.
FAQ
Can I outsource this?
Yes, agencies like MatthCon offer AI-search tracking as a service.
How much time does it take myself?
2-4 hours per month for a good baseline.
How do I compare with competitors?
Share of Voice metric. Test the same prompts for all players.
Which metric should I start with?
Mention rate. It's the simplest to measure and the most honest baseline: are you mentioned at all when someone asks a relevant question? Only once that's above 30% does citation rate become the more interesting lever.
How often should I measure?
Monthly is enough for most B2B organizations. AI models don't refresh their sources daily, so weekly measuring mostly produces noise. Pick a fixed moment, use exactly the same prompts and document consistently.
Need help? MatthCon sets up complete AI-search tracking. Schedule a call.