Episode 4
AI search ROI template
How to Measure AI Search ROI and Citation Share
Share of Model in the episode is brand mention rate here: the share of tracked prompts whose answers name you. Citation share compares those mentions with the competitors on the same rows. Query coverage is the share of prompts that cite a URL you own.
0 / 12 checked. Saved in this browser.
1
Citation presence
Start with 10 to 15 commercial prompts. Add more once the log is a habit.
2
Citation quality
3
Citation impact
Three layers
| Layer | Question | Metrics |
|---|---|---|
| 1. Citation presence | Were you named? | Brand mention rate, citation share, query coverage |
| 2. Citation quality | How were you described? | Accuracy, sentiment, position, differentiators |
| 3. Citation impact | What reached pipeline? | AI referral traffic, AI impressions, self-reported attribution |
Cadence
| When | Look at | Write down |
|---|---|---|
| Weekly | 10 to 15 top commercial prompts | Wrong answers and missing blocks |
| Monthly | The logs and Search Console | Mention rate, AI impressions, AI conversions |
| Quarterly | Named competitors | Citation share, content gaps, pipeline |
Download prompt log
Row 2 of the file is an example. Delete it before you log real prompts.
| Column | What to enter |
|---|---|
| Date | YYYY-MM-DD |
| Engine | Which answer engine produced this response |
| Prompt | The question, identical across engines |
| Brand mentioned | yes or no. Brand mention rate is the share of yes. |
| Owned URL cited | yes or no. Query coverage is the share of yes. |
| Competitors mentioned | Names, separated by semicolons |
| Accuracy | accurate, outdated, wrong, or absent |
| Sentiment | positive, neutral, negative, or absent |
| Position | leader, primary option, alternative, or absent |
| Differentiators | Value props the answer actually stated |
| Cited URL | URL the answer pointed at, if any |
| Citation type | owned, earned, or none |
| Notes | What you will change |
Download weekly log
Row 2 of the file is an example. Delete it before you log real prompts.
| Column | What to enter |
|---|---|
| Week starting | Monday, YYYY-MM-DD |
| GA4 AI sessions | Sessions in the AI referral channel |
| GA4 AI key events | Conversions from that channel |
| GSC AI impressions | Generative AI report impressions |
| Self-reported AI | Form answers that name an AI tool |
| Branded search impressions | Search Console queries that are your brand |
| Notes | What changed that week |
GA4 session-source regex
Custom channel group. ChatGPT also adds utm_source=chatgpt.com on many outbound links. The playbook lists the hosts.
.*(chatgpt\.com|perplexity\.ai|claude\.ai|copilot\.microsoft\.com|gemini\.google\.com).*Questions
- What is Share of Model on this sheet?
- The percentage of your target prompts where the answer names your brand. The playbook calls that brand mention rate.
- What goes in the weekly log?
- AI referral sessions and conversions, Search Console AI impressions, branded search impressions, and how many people named an AI tool when asked how they heard about you.