Monthly AI Visibility Report Template
A monthly AI visibility report has six sections, one per data source that actually exists: access, index coverage, search performance, assistant referrals, agent hits in your logs, and a prompt panel. This template gives the columns for each, how to collect them, and the one rule that keeps the report honest: compare full months to full months and report trends, not single answers.
By George Shvaya ·
1. Access
Cadence: Monthly, or after any CDN, firewall or robots.txt change
Run the AI crawler checker or the full audit on the homepage and the five pages that earn the most. Record the verdict per agent. Any non-200 to a search or user agent is a blocking item for the month.
| Agent | robots.txt verdict (allowed / blocked) | Live fetch status to homepage | Live fetch status to top 5 service pages | Change since last month |
|---|---|---|---|---|
| … | … | … | … | … |
Source: Webvello AI crawler checker or /api/aeo-audit; vendor IP lists for verification.
2. Index coverage
Cadence: Monthly
Search Console, Pages report, filtered to your sitemap. Google documents that a page must be indexed and snippet-eligible to appear in AI Overviews or AI Mode, so this ratio is the ceiling on everything below it.
| Submitted URLs | Indexed URLs | Indexed ÷ submitted | Discovered, not indexed | Crawled, not indexed | Change since last month |
|---|---|---|---|---|---|
| … | … | … | … | … | … |
Source: Google Search Console Pages report.
3. Search performance
Cadence: Monthly, full calendar months only
Search Console Performance report, web search type, which per Google includes AI Overviews and AI Mode traffic. Sum by calendar month and never compare a partial month to a full one.
| Month | Clicks | Impressions | Impression-weighted avg. position | Top 10 question-style queries by impressions | Change since last month |
|---|---|---|---|---|---|
| … | … | … | … | … | … |
Source: Google Search Console Performance report.
4. Assistant referrals
Cadence: Monthly
Analytics referral report or an exploration filtered to those hostnames. Volumes are small and referrers are sometimes stripped, so treat the number as a floor and watch the direction.
| Referrer host (chatgpt.com, perplexity.ai, copilot.microsoft.com, gemini.google.com, claude.ai) | Sessions | Landing pages | Conversions | Change since last month |
|---|---|---|---|---|
| … | … | … | … | … |
Source: Google Analytics 4 or equivalent.
5. Agent hits in server logs
Cadence: Monthly
Count requests by user-agent token and verify a sample against the published IP lists. User-initiated agents are the closest proxy for an assistant reading a page to answer someone. Training crawlers say nothing about visibility; keep them in a separate row.
| Agent (OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot, Perplexity-User, GPTBot, ClaudeBot) | Requests | Distinct URLs fetched | Top 10 URLs | Change since last month |
|---|---|---|---|---|
| … | … | … | … | … |
Source: Server or CDN access logs.
6. Prompt panel
Cadence: Monthly, same prompts, fresh sessions
Twenty to forty real customer questions asked in each assistant in a logged-out or fresh session. Record the answer the same way every month. Answers vary between runs, so report the trend across months, never a single result.
| Prompt | Assistant | Named? (yes/no) | Cited with link? (yes/no) | Competitors named | Date | Change since last month |
|---|---|---|---|---|---|---|
| … | … | … | … | … | … | … |
Source: Manual; ChatGPT, Perplexity, Claude, Google AI Mode.
Want the baseline run for you?
The written audit fills sections one, two and six for your site and tells you which fixes change next month’s numbers.
How to read the report
- One section moving alone is noise. A prompt panel win with flat impressions and no agent hits is a lucky run. Three sections moving the same direction is a trend.
- Access failures override everything. If a search or user agent got a 403 this month, nothing below that row is interpretable until it is fixed.
- Partial months are not data. Close the month before you sum it. Google’s own reporting lags, so pull a few days after month end.
- Write the one fix. Every report ends with the single change that most likely moves next month’s numbers, and last month’s fix is checked against this month’s rows.
Common questions
How many prompts should the panel have?
Twenty to forty. Fewer than twenty and one noisy answer moves the number; more than forty and nobody runs it every month. Use the questions customers actually ask, in their words.
Why not a single AI visibility score?
Because the sources measure different things with different reliability. A composite hides which one moved. Report the six sections separately and let the trend in each speak.
What is a good month?
Access stays clean, indexed-versus-submitted rises or holds, impressions on question-style queries grow, user-initiated agent hits grow, and the prompt panel names you for more prompts than last month. Any one of those alone is noise; together they are a trend.
Can Webvello run this for me?
Yes. The written AI visibility audit establishes the baseline and the monthly program runs this report with the fixes it implies.