How to Use Promptwatch's Agent Chat Side Panel (Cmd+J) to Speed Up Weekly Reporting in 2026

Agent Chat now sits next to nearly every Promptwatch dashboard page and answers from the charts and filters on screen. Here's a practical weekly reporting workflow built around that side panel, with prompts to copy and pitfalls to avoid.

Key takeaways

  • Since November 6, 2026, Promptwatch's Agent Chat opens next to nearly every dashboard page and answers from the charts, tables and filters you are looking at.
  • Cmd+J is the shortcut for the page-aware panel in the content editor and side panel. I could not confirm it for every page in the public changelog, so test it in your own account before you put it in a team SOP.
  • A weekly report goes fastest when you set the date range and filters first, then ask one question per tab in a single running conversation.
  • Table questions are answered for the whole filtered list, not just the visible page of rows, so you don't need to paginate before asking.
  • Check numbers against the on-screen chart before pasting anything into a client deck. Treat the chat as a fast analyst, not an unreviewed source of truth.

Why a page-aware panel changes weekly reporting

Most weekly AI visibility reports are the same chore. Open the dashboard, screenshot a chart, flip to citations, screenshot again, open a spreadsheet, and then spend an hour writing the paragraph that explains what moved. The numbers take ten minutes. The explaining takes the rest.

A page-aware assistant attacks that second part. When the assistant can see what is on the screen, you can say "these prompts" or "this drop" and it knows what you mean. That is the whole idea behind the pattern, and it is the same problem developers describe when they build page-aware agents: the user says "cancel these" and the agent has to know what "these" refers to. In a reporting context, "these" is the filtered table in front of you.

Promptwatch built this into Agent Chat in two steps. On October 16, 2026, the AI chat panel that already lived in the content editor started opening next to monitors, covering the overview, prompts, responses and citations tabs. On November 6, 2026, it expanded to nearly every dashboard page. Both changes are documented in the changelog entries for the monitor side panel and Agent Chat across the dashboard.

Favicon of Promptwatch

Promptwatch

Track and improve your AI search visibility
View more
Screenshot of Promptwatch website

What the side panel actually does

It helps to be precise here, because "AI chat" can mean anything.

  • It answers from the page you are on, using the filters and date range you have set.
  • It stays open as you move between a monitor's tabs and detail views, so one conversation can run across the whole report.
  • Each message shows which tab it was sent from, and the chat tells you when you have moved tabs since your last message.
  • Each page has its own suggested questions, and the panel follows along when you switch tabs.
  • Detail sheets, such as a claim or a topic, open next to the chat instead of covering it.

That last point sounds minor. It isn't. Half the friction of reporting is losing your place when a drawer slides over the thing you were asking about.

About Cmd+J

Cmd+J is the shortcut Promptwatch uses to open the AI chat panel in the content editor, and it is the one to try on dashboard pages too. The changelog entries I reviewed don't spell it out for every page, so press it once on a monitor tab and on a Trends page before you write it into a runbook. If it does nothing on a given page, use the chat button in the panel area instead. Windows users should try Ctrl+J.

The pages worth covering in a weekly report

The November 6 rollout covers a lot of ground. You do not need all of it every week. Here is how the pages map to a typical reporting rhythm.

Report sectionWhere to open the panelWhat to ask
Visibility movementMonitor overview and prompts tabsWhich prompts lost the most visibility this month, and which competitors took their place?
SourcesMonitor citations tabWhich cited domains don't I appear on yet?
ReputationBrand Narrative, Sentiment, Fact CheckWhich claims about us changed this week, and which models repeat them?
CompetitorsCompetitors pageWhich brands show up in answers that I don't track yet?
Technical healthCrawler Logs, Site HealthWhich pages did AI bots fail to read, and why?
TrafficVisitor AnalyticsWhich pages drove the most AI-referred visits compared with last week?
ContentContent GapWhich uncovered prompts matter most?

The first two prompts in that table come straight from the changelog's own examples, so they are known to fit the way the panel reads a page.

A weekly reporting workflow, step by step

This is the routine I would run. It takes the panel from a gimmick to something you can do in about 20 minutes.

Step 1: Fix the date range and filters first

The panel answers from the filters you have set. If you ask about "this week" while the date range is set to 30 days, you will get a 30-day answer. Set the range to the last 7 days, then choose the monitor, model, or topic you are reporting on. Use the A/B comparison on charts if you want week-over-week context on screen.

If you report per market, remember that a monitor's language and location are fixed once it is created, so each market is its own monitor. Run this routine once per monitor.

Step 2: Open the panel on the overview tab and get the headline

Start broad. Something like:

Summarize what changed in visibility, share of voice and average position 
for this monitor over the selected period. Name the biggest mover in each metric.

Then check the answer against the charts. Take thirty seconds for this. If the headline doesn't match the chart, you want to know before step 3, not after the client call.

Step 3: Move to the prompts tab and ask about losses

The chat knows you moved tabs, and because table questions are answered for the whole filtered list, you can ask about everything that matches rather than the 25 rows on screen.

Which prompts lost the most visibility this period, and which competitors 
took their place? List the top five with the before and after numbers.

This is the paragraph that normally eats an hour. You get a draft in seconds, and your job becomes verifying and trimming it.

Step 4: Go to citations and find the pitch list

Which cited domains don't I appear on yet? Rank them by how often they are cited.

The changelog calls these the sites to pitch or publish on. For a weekly report, this becomes the "next actions" section almost by itself.

Step 5: Check the reputation and technical tabs

On Brand Narrative, ask which claims shifted and which models repeat them. Fact Check is available on Business, Agency Scale and Enterprise plans, and it lets you mark a claim accurate or inaccurate and find the page AI is learning the wrong thing from.

On Crawler Logs and Visitor Analytics, a useful question is why no data is coming in, if a chart looks empty. The agent checks your setup status for you. API key values are never shared with it.

Step 6: Close with the Content Gap page

This is where reporting turns into action. On Content Gap, ask which uncovered prompts matter most, then have the agent draft content for the top few. If you are on a plan with the Content Agent, those drafts can go straight to your CMS, but the report itself only needs the list and a sentence on why.

Step 7: Ask for the written summary

Because the conversation ran across tabs, the last message can pull it together:

Write a short weekly report from everything we covered in this conversation: 
headline movement, the three biggest prompt losses, top citation opportunities, 
any reputation or crawl issues, and the one thing to do next week. 
Keep it under 250 words, plain language, no hype.

You can also ask Agent Chat to build a PDF report or a custom dashboard. On Promptwatch, scheduled chats can be delivered in-app or in Slack, which is the closest thing to a hands-off weekly briefing.

Prompts worth saving

I keep a short list of these. Adjust the wording to your own brand and markets.

  • Competitor check on the Competitors page: "Which brands show up in answers that I don't track yet?" The agent asks before it adds anything, which is how it should behave.
  • Sentiment: "Which topics have the weakest sentiment this period, and what claims drive it?"
  • Share of voice: "Where did we lose share of voice to a single competitor, and on which models?"
  • Crawl failures: "Which pages returned errors to AI crawlers this week, and what does the log say about why?"
  • Traffic: "Which pages sent AI-referred visits this week, and which of them are also cited in the monitor?"

The last one is a good example of the cross-page question that used to require two tabs and a spreadsheet.

Pitfalls I would watch for

The panel is fast, which is exactly why a few sloppy habits can cost you.

New projects have nothing to rank yet

During the first site audit, the agent says there is nothing to rank yet instead of reading the sample cards. That is the right behavior, but it can look like the feature is broken. Wait for the first real audit run before asking about site health.

Tab context is easy to lose

The chat flags when you have moved tabs since your last message. Read that flag. If you ask "why did this drop?" after switching from prompts to citations, you are asking about a different table.

Filters are part of the question

Every answer inherits your date range, model filter and topic filter. When two team members get different answers to the same question, check their filters before you blame the tool.

Permissions matter

Project viewers can read the knowledge base but cannot change it. If a teammate says the agent refused to update a source, that is probably why.

Always verify before it goes to a client

I treat the chat's numbers like a junior analyst's first draft. Spot-check two or three figures against the chart on screen. It takes a minute and protects you from the one wrong number that someone remembers.

How it compares with the generic agent approach

A lot of the weekly reporting content around right now is about general-purpose assistants, for example ChatGPT agent mode or workspace agents that pull Friday metrics, draw charts and draft a narrative. Those work, but you have to connect data sources and describe your data every time. A walkthrough such as this ChatGPT Agent Mode tutorial also notes that agent tasks run for several minutes each with the agent working alone, which is fine for a research report and awkward for a quick Monday check.

ApproachSetup effortKnows your AI visibility dataInteraction style
Promptwatch side panelNone beyond your monitorsYes, reads the page you are onConversational, per page
General agent mode (ChatGPT etc.)Connect sources, write long promptsOnly what you supplyLong autonomous runs
Spreadsheet plus manual writingExport every weekOnly what you exportManual

If you want the same data inside other assistants, Promptwatch also has a hosted MCP server with read and write tools for Claude, ChatGPT, Cursor and other clients, documented at promptwatch.com/integrations/mcp. That is useful when your report lives in a document or a Slack thread rather than the dashboard.

Making the routine stick

  1. Pick one day and time. Monday morning works because Promptwatch's weekly digest email lands then and the Action Board refreshes weekly.
  2. Keep one saved prompt list per monitor, in a shared doc.
  3. Run steps 1 to 7 in a single conversation so the final summary has everything.
  4. Paste the final summary into your report template and edit it by hand.
  5. Log the next actions on the Action Board so the same issue does not appear three weeks in a row without an owner.

If you want the numbers to be solid, read how the key metrics are defined first. The Academy pages on Visibility Score and Share of Voice explain what the panel is actually reading when it summarizes movement.

Plans and getting started

Agent Chat is included on every Promptwatch plan, with Agent Credits ranging from 500 to 6,500 per month depending on the plan. I did not find public detail on per-page limits or report export from the chat, so check promptwatch.com/pricing for current plan details. Plans start at $95 per month with a 7-day free trial on monthly billing, and there is a Black Friday offer of 30% off a plan that ends November 30 at 23:59 CET. If you want a baseline before you commit, the free AI visibility report needs no login.

For more options when comparing platforms that offer reporting and assistants, the GEO software directory at bestgeosoftware.com is a reasonable place to browse.

The short version

Set your filters, open the panel, ask one question per tab, and finish with a request for the written summary. The tool removes the hour of explaining. The checking is still yours.

Share:

AI Search Visibility Tools

© 2026 AI Search Visibility Tools · The best AI search visibility tools compared · RSS

AI Search Visibility Tools is an affiliate review site. When you click links to vendors or buy through links on our site, we may earn an affiliate commission at no extra cost to you.

AI Search Visibility Tools is a review website based on user reviews on Reddit and G2, and on publicly available information. We keep everything as up to date as possible, but pricing and features can change. Always confirm the details with the vendor before purchasing.