Skip to main content
How current is what you’re looking at? The short answer: your prompts run daily by default, responses arrive in batches through the day, and each response passes through an analysis step before it counts in your metrics. This page walks the pipeline so you can tell “still in flight” from “missing”.

​
How often prompts run

Every monitor runs its prompts on a schedule against its selected models. The default, and what every plan uses, is daily. Trial accounts also run daily. Organizations with the custom tracking frequency feature enabled can set an individual monitor to a slower cadence, which trades freshness for a lower response spend on prompts that don’t need daily tracking. That feature is off on every plan by default and is enabled per organization by sales, not by picking a plan. When it is on, the Tracking frequency select appears in step 1 of the create monitor wizard and in the monitor’s Settings (rail tile Prompts, then Monitors, open the monitor, then the Settings button). The options are Daily, Every 2 days, Weekly and Monthly, plus the organization default. A slower cadence set on a monitor is kept but ignored if the feature is later turned off. See creating a monitor and managing monitors. Slower cadences anchor to the day the monitor was created: an every-2-days monitor runs every second UTC day from its start, a weekly monitor on the same weekday it started, and a monthly one on the same day of month. A monitor started on the 31st runs on the last day of shorter months. Days between runs produce no new responses for that monitor, which is worth remembering before reading an empty day as a problem. Some events are fast-tracked. When you create a monitor, add or copy prompts, import prompts in bulk, reactivate a monitor or deactivated prompts, or change a monitor’s models, that run skips the cadence wait and jumps the bulk backlog, so a new setup typically starts producing data the same day. See what is a monitor.

​
What happens between a run and your charts

Each scheduled run goes through three stages:
  • Scheduling: the scheduler checks at the top of every hour, and on a monitor’s run day its prompt-model pairs are queued. Days follow UTC, so if you’re well east or west of UTC, “today” in the app starts and ends at a different hour than your local midnight, and a monitor that becomes eligible at UTC midnight can start up to an hour later unless it was fast-tracked.
  • Collection: prompts are asked on the actual AI platforms the way a real user would ask them, and this is dispatched in batches through the day rather than in one burst. Responses trickle in across hours. See how Promptwatch collects data.
  • Analysis: every completed response is analyzed to extract mentions, sentiment, and visibility. Analysis runs continuously, picking up new responses within minutes of completion, but a response only counts in mention and visibility metrics after this step finishes.
Charts read from analyzed data, so they update progressively as the day’s responses complete and clear analysis. There is no fixed “dashboard refresh time”; the data flows in as it’s ready. The app does cache a chart you have already loaded for up to about five minutes, so reload the page if you are waiting on the newest responses.

​
Why “today” can look empty

The most common freshness question is a today view that looks thin next to yesterday, and it’s almost always one of these:
  • The day isn’t done: with collection spread across the UTC day, a morning look catches a fraction of the day’s responses. Yesterday is complete; today is in progress. Compare complete days.
  • Analysis is trailing collection: charts only include analyzed responses, so mention and visibility numbers stay empty until that step finishes, even after collection has already completed.
  • It’s not a run day: a monitor on an every-2-days, weekly or monthly cadence produces nothing between runs, by design.
  • A model is mid-cycle: models within the same run finish at different times, so per-model views firm up at different hours. See no data for a model.
The brand visibility chart on your dashboard makes the first case visible: on the Visibility over time (Avg) tab, the last segment of each line is drawn dashed when it ends on today, so you can see at a glance which point is still filling in. The Promptwatch brand visibility chart on the dashboard with the last segment drawn dashed because today's responses are still arriving.

​
Why a monitor stops producing data

If a day still looks empty after the following day has started, that’s no longer freshness. Before you dig into individual prompts, rule out the reasons the scheduler skips a monitor entirely:
  • The response quota is used up: once your organization reaches its response limit for the period, or a project reaches its own cap, the scheduler skips the monitors that would exceed it. New responses resume when the period resets. See response limits.
  • The subscription is not entitled to collection: monitors run only while the organization is active, trialing, past due, or has never subscribed. See plans and billing.
  • The monitor or prompt is paused: only active monitors and active prompts are scheduled, and a pitch project’s monitors stop after its end date.
  • You are on the Explore plan: Explore runs only the ChatGPT Search model, so the other models show no data by design.
Once those are clear, work through why is my visibility zero.

​
How to read your dashboard with this in mind

Treat the last complete day as your freshest reliable data point, and today as a preview. For trend decisions, week-over-week comparisons of complete days beat reacting to a partial today. See reading your dashboard for the rest of the dashboard. One exception to everything above: the AI Insights and Topic Insights pages under Analytics are built from a daily rollup and their period ends yesterday. Search Insights follows the Google and Bing sync, and with Sync engine dates on it ends at Bing’s last complete weekly report. Each page shows its own “Last updated” time. See 360 Insights.