Sample workspace · yoursite.com · fictional
What a prompt was worth, and who came to read the page
Visibility is only interesting if it ends in money, and only honest if the arithmetic is shown. Both tables below carry the part they cannot explain rather than folding it into the part they can.
Revenue per prompt
The five questions that paid
AI-assisted revenue
$78,400
USD · 90-day window. Every session whose first touch was an assistant.
Resolved to a prompt
$63,050
80.4% of the above. The rows below.
Unattributed
$15,350
AI-assisted, but the landing prompt could not be identified. Shown, never spread.
Attributed sessions
1,162
54 converted — 4.6% of them.
| Prompt | Sessions | Conversions | Rate | Revenue |
|---|---|---|---|---|
| best project cost tracking software for agencies | 412 | 19 | 4.6% | $22,800 |
| project cost tracking tool that integrates with Xero | 268 | 14 | 5.2% | $16,100 |
| alternatives to Quantly for cost tracking | 197 | 9 | 4.6% | $10,350 |
| is yoursite.com worth it for a small team | 154 | 8 | 5.2% | $9,200 |
| cheapest way to track billable hours against a fixed-fee budget | 131 | 4 | 3.1% | $4,600 |
| Resolved to a prompt | 1,162 | 54 | 4.6% | $63,050 |
| AI-assisted, prompt unknown | — | — | — | $15,350 |
| AI-assisted revenue | — | — | — | $78,400 |
Per-prompt revenue is the part of AI-assisted revenue that resolves to a single tracked prompt. The remainder is AI-assisted sessions whose landing prompt could not be identified — it is shown rather than spread across the rows.
An AI-assisted session is one whose first touch was an assistant referral. Where the landing page maps to a tracked prompt, the revenue is attributed to that prompt; where it does not, it stays in the remainder.
Agent traffic
Which bots read the site, and why they came
Agent requests
38,930
90-day window, from the edge log.
Verified by reverse DNS
91.4%
The remainder claimed a known user agent that its IP could not back up, and is excluded from the split below.
Largest class · Live agent
47.3%
Fetching a page to answer a question right now
| Bot class | What it is doing | Share | Share | Requests |
|---|---|---|---|---|
| Live agentChatGPT-User, Perplexity-User, Claude-User | Fetching a page to answer a question right now | 47.3% | 18,420 | |
| Search indexGooglebot, Bingbot, PerplexityBot | Building the index an assistant searches | 28.4% | 11,060 | |
| TrainingGPTBot, ClaudeBot, Google-Extended | Collecting pages for model training | 18.8% | 7,310 | |
| Shopping agentOAI-SearchBot, ShoppingBot | Comparing price and availability on behalf of a buyer | 5.5% | 2,140 |
Classes are assigned from the verified user agent, not from a string match — an unverified agent is counted in the total and left out of the split rather than guessed at.
A crawler collecting training data and an agent fetching a page to answer a question right now are not the same visitor. Splitting them is the difference between a log line and a decision about robots.txt.