Turn a saved AI answer into a useful next step.

Write down what the answer actually shows.

Read the full saved response before creating a task. Check the question wording, provider, collection date, and supporting quote. A brand assessment summarizes an answer; the original response provides the context for interpreting it.

For example, a saved answer might recommend a competitor for agency reporting without mentioning your product. That is an observation in one sample. It does not establish that the provider always excludes your brand or that customers prefer the competitor.

Give the action a title describing work someone can perform: “Review our agency reporting description against the saved answer.” In the notes, separate the recorded finding from a hypothesis about what could explain it.

Keep the evidence selection attached.

Create a follow-up from a saved answer, a comparison, a change log entry, or a supported analysis screen when you want to preserve its context. Review Linked evidence in the action draft before saving. It shows the destination, date scope, and saved selection.

A link to a specific saved answer returns to that record with its current text and analysis. A saved comparison preserves the selected pair and its A/B order. A link to an analysis view opens current results using the saved filters. These links do not all represent a frozen snapshot.

Relative reporting periods are evaluated when the view opens. A last-30-days selection will move forward over time. Use a supported custom date range when a fixed period is essential, and keep the original observation in the action notes.

An action created from a general analysis screen preserves its filters, rather than turning every visible result into a permanent selection. Check the linked scope before sharing the task. If no evidence is attached, include enough context in the notes for another person to locate the finding.

Define the work and the later review.

Keep the task narrow enough to finish. In the example, first verify that the product supports the capability discussed in the answer. Then review whether the relevant page explains it accurately. A documented gap can justify a content update without promising that an AI provider will recommend the product.

Describe how you will review the work. You might check the published wording and revisit comparable saved samples for the same question and provider. Record missing answers and unassessed samples rather than treating them as negative brand results.

Creating an action, changing its status, or setting a due date does not submit provider questions or change the tracking schedule. Check the collection plan separately if the review needs new answers.

• Observation: what the saved evidence says.
• Work: what you will investigate or change.
• Review: which comparable evidence you will inspect afterward.

Choose a priority and a realistic review date.

Set a low, normal, or high priority based on the importance of the investigation. Priority helps organize the work; it is not a confidence score for the AI finding or a prediction of impact.

Due dates use UTC. The editor offers today, tomorrow, one week, or no due date, and you can choose another date. The action list identifies overdue active work using its recorded UTC date snapshot. Use Refresh actions to update that snapshot when needed.

Completed and archived actions are not overdue. Due-date filters such as due today and upcoming focus on unfinished, nonarchived work. Use the status and archive controls separately when you want to review completed records.

Use the list to find the next task.

Filter follow-up actions by title or notes, status, priority, due date, and archive state. The applied chips show the current selection and let you remove one setting at a time. Sorting can prioritize importance, the earliest due date, or the newest task.

The Open, In progress, Done, and Overdue counts cover nonarchived actions across the project, rather than only the current filtered page. Selecting a count opens its corresponding view. Read the matching-result count when you need the scope of the current list.

Expand Linked evidence to inspect the saved filters before opening the destination. Use Start work, Pause work, or Mark done to reflect the task’s progress. Marking a task done records that status; it does not verify an improvement in AI visibility.

Close the loop with a traceable result.

Update the notes with the work completed, what the later evidence shows, and what remains uncertain. If you changed a website page or the tracking setup, record that event in the change log with its actual date and reference.

Use Revision history to review earlier versions of an action. Archive records you want to remove from the active list, and use the archived view to find or restore them. A completed task can remain useful context for a later review.

Export all matching actions when you need a CSV of the selected project view. Check the filters before sharing it: an export of high-priority open tasks is not a complete record of all work. Keep observations and claims proportional to the saved evidence.