Skip to main content

​
What a monitor groups

A monitor is a panel of prompts that runs on a repeating schedule against the models you pick. Every prompt inside it shares the monitor’s settings:
  • Tracking language: the language the prompts are asked in.
  • Target country, plus an optional state and city: where the model is told the user is.
  • Persona: an optional description of who is asking, so answers reflect that audience.
  • Models: which AI models answer the prompts.
One context per monitor is a deliberate design choice: it keeps your time series clean, because a visibility change inside a monitor can’t be explained away by a mix of languages or markets. That is why most teams run one monitor per market, named like “US - English” or “DE - German”. See what is a monitor and three monitors to set up for how to split them. What a monitor does not define is the prompt type (Organic, Brand Specific, Competitor Comparison) or the intent. Those are properties of each prompt, so a single monitor can hold a mix. If you want a monitor per type, that is a naming convention on your side, not a setting. See working with prompts and organizing your setup.

​
How many prompts a monitor can hold

There is no per-monitor cap. The limit is the number of active prompts your plan allows, shared across all projects in your organization. The number per plan is on the pricing page. The wizard checks this limit when it suggests starting prompts, and again whenever you add prompts later. See response limits for how prompts translate into responses.

​
How to create a monitor

Project viewers cannot create monitors; you need an editor role.

​
Open the Monitors page

Click Prompts in the left rail, then Monitors. Hover the tile for a moment to get the flyout, or use the list on the left once you are inside the section. The page is titled Monitors. The Promptwatch left rail with the Prompts tile hovered and its Monitoring flyout open, listing Prompts, Monitors, Query Fanouts and the other monitoring pages. Shortcut: press cmd+K (Ctrl+K on Windows) and choose Create Monitor to skip straight to the wizard.

​
Click New Monitor

Click New Monitor at the top right. If the project has no monitors yet, the empty state shows “Create your first monitor” with a Create Monitor button instead. The Promptwatch Monitors page with the New Monitor button at the top right and a list of monitor cards.

​
Step 1 of 2: Create Monitor

The Create Monitor dialog is the first of two steps (the progress bar shows both). Fill in the context the prompts will run in:
  • Monitor Name: required. Pick a name that says what the panel covers, like “US - English”.
  • Tracking Language: the language your target audience uses with the model. Prefilled from the project language.
  • Target Country: the country users are expected to be in. Prefilled from the project country. Choosing a country other than the default needs the country-level tracking feature, which is locked on Explore and available from Essential up.
  • State and City: two separate optional fields. State only works when the target country is the United States (the placeholder reads “US only” otherwise). Both are plan-locked below Business. Changing the country clears both fields.
  • Persona (Optional): pick None, one of your project’s personas, or Create new to add one without leaving the wizard. When a persona is chosen, a Persona stacking switch appears: it wraps each tracked prompt in the persona’s prompt at run time, so the model answers as if that person asked. The persona needs a usable stackable prompt for this; otherwise the wizard shows an error. See personas.
  • Models: the models that will answer. The list shows ChatGPT, AI Overview, AI Mode, Copilot, Gemini and Perplexity, with ChatGPT, Perplexity and AI Overview ticked by default. Click Show all models to reveal the rest. A model reads “Not available in this region” when the chosen country does not support it, and models outside your plan are disabled with an “Upgrade” pill (Explore gets ChatGPT only). At least one model is required. See choosing your models.
The Promptwatch Create Monitor dialog, step 1 of 2, with monitor name, tracking language, target country, state, city, persona and model fields. Click Next.

​
Step 2 of 2: Starting prompts

Promptwatch generates up to 10 suggested prompts from the monitor name, the location and the persona, so the panel starts with something to track. If you have fewer than 10 active prompts left on your plan, a banner tells you how many it will suggest instead. Each suggestion is a row with the prompt text, its type and intent, and Topics and Tags fields. From here you can:
  • Edit any row, or remove it with the trash icon (the last remaining row can’t be removed).
  • Click Add prompt to add one of your own.
  • Click Topic all prompts or Tag all prompts to label the whole batch at once.
The Promptwatch Starting prompts step listing suggested prompts with type and intent, with Back, Create without prompts and Create monitor buttons. Then choose:
  • Create monitor to save the monitor with the prompts shown. Empty rows are skipped. The button stays disabled when no row has text or the batch exceeds your remaining capacity.
  • Create without prompts to save an empty monitor and add prompts later.
  • Back to change the settings from step 1.
If generation failed, a Retry button appears. If you have no prompt capacity left, an alert titled “Prompt limit reached” asks you to create the monitor now and add prompts once there is room. Closing the dialog at this step asks “Discard starting prompts?” before throwing the suggestions away.

​
After you click Create monitor

You land on the new monitor’s Overview page and see a “Monitor created successfully” toast. The monitor is active immediately and its starting prompts are queued, so the first responses arrive after the first run. Along the top of the monitor you’ll find the Overview, Prompts, Responses, Sentiment, Citations, Offsite Mentions and Heatmap tabs, plus a Settings button for changing the monitor later (see managing monitors). To grow the panel from here, see adding prompts.