Brand Book and Scheduled Chats: how to set up weekly AI visibility reporting
A reporting setup that stays consistent: define what counts as a mention, freeze the panel, and automate the weekly summary. Includes a template and notes for agencies.
Quick verdict
Most AI visibility reports fail in one of two ways. The definitions drift, so last month's number does not compare with this month's. Or the report depends on one person who has to remember to build it. Both are fixable. Define what counts as a mention once, freeze the panel, and automate the summary. Promptwatch gives you three features for that: the Brand Book, scheduled reports and PDF exports, and Scheduled Chats.
Step 1: write the definitions down
Before you automate anything, decide:
| Question | Example decision |
|---|---|
| What is our brand name and aliases? | Company name, product name, common misspellings |
| Does a product name alone count? | Yes, if it clearly refers to us |
| Does a passing "also consider" count? | Yes, but record position |
| Do we count mentions in negative context? | Yes, and track sentiment |
| Which competitors are in scope? | A fixed list, reviewed quarterly |
In Promptwatch, the Brand Book's analysis instructions decide what counts as a mention, and brand aliases catch alternate names. Writing the rules there means every response is judged by the same standard, and a change in definition becomes a visible edit instead of a silent drift.
Step 2: freeze the panel
Pick the prompts, engines, countries, and languages for the reporting period. Promptwatch runs prompt panels daily per model, country, and language. Do not add or remove prompts mid-month, or annotate the change.
Also fix the competitor set. Promptwatch auto-discovers competitors, so review the list when the quarter begins.
Step 3: choose the weekly metrics
Keep it short. A useful weekly page has five items:
- Visibility Score and Share of Voice, with a comparison to the prior period.
- Position in Answer for priority prompts.
- Sentiment, with the weekly AI summary.
- The biggest gain and the biggest loss in citations.
- One recommended action.
The A/B period comparison lines up any two date ranges, and the heatmap compares any two dimensions, such as brands by model or topics by country.
Step 4: add the traffic line
A visibility report with no traffic line invites the question "so what?" Add AI-referred sessions and conversions from visitor analytics. If you have crawler logs, add one sentence on bot errors. Professional and above include them.
Step 5: automate the delivery
Promptwatch provides several delivery paths.
Scheduled PDF reports. PDF reports of any dashboard, with scheduling. White-label branding on PDF exports and schedules is listed for agencies.
Scheduled Chats. Agent Chat is a conversational analyst over live data that can run multi-step workflows, render charts, and generate PDF reports. Scheduled Chats deliver results in-app or in Slack. Agent Skills, such as a 7-day GEO briefing or a competitor gap analysis, provide a ready-made prompt for the recurring task.
Weekly digest. The Action Board sends a weekly digest email, and the Discovery agent suggests up to five evidence-backed actions per weekly run.
MCP. The hosted MCP server lets other agents pull the data. TreeHouse, a Promptwatch customer, reported automating weekly reporting with the MCP.
Step 6: add a review habit
Automation does not replace reading. Set a 20-minute weekly slot. Read the summary, check one anomaly, and confirm the recommended action. If a number looks wrong, check the Brand Book rules first, then the prompt panel, then the engine.
A template
| Section | Content |
|---|---|
| Headline | One sentence: up, down, or flat, and why |
| Scores | Visibility, Share of Voice, Position, Sentiment, with prior period |
| Movers | Top gain, top loss, with the likely cause |
| Traffic | AI-referred sessions and conversions |
| Technical | Any crawler errors or access issues |
| Next | One action, one owner, one date |
Agency notes
Agencies add two needs: separate client workspaces and presentable output. Promptwatch agency plans include unlimited projects, unlimited pitch projects, custom reports, and the Looker Studio connector. White-label workspace and client portal are listed for agencies, though we will not attach a plan or price to them. Slack briefings are not available on white-label workspaces. Check which options you need before choosing an agency tier.
Plan notes
Custom reports with Data Studio start on Professional at $245 a month and are on all agency plans. Essential at $95 includes dashboards and PDF exports. Usage alerts at 85% of a quota help avoid a panel stopping mid-month. Agency plans start at $199 with unlimited projects. The 7-day trial is on monthly billing only.
Other tools in the ranking offer reporting too. Check each entry for scheduling and white-label details, as they vary, for instance in AthenaHQ's entry.
Who should pick which
| You are... | Recommendation |
|---|---|
| A solo marketer | Brand Book plus one scheduled PDF |
| An in-house team | Scheduled Chats with a Slack delivery |
| An agency | Agency plans, custom reports, Looker Studio connector |
| Wanting hands-off summaries | Agent Skills and Scheduled Chats |
| Needing a board-ready deck | PDF exports, edited by a human |
Define it, freeze it, schedule it, read it. Promptwatch holds all four steps in one place, which is why it is the tool we would set up first.