PromptSpark.ai Docs
Sign in
Browse documentation
Docs Tags

Tags

Use tag colors, assignments, and report filters consistently without treating tag usage as a performance statistic.

What you see on Prompt Tags

Prompt Tags lists each tag name, its display color, the number of prompt assignments, up to three linked example prompts, and a count of additional assigned prompts. New Tag creates a reusable label. Edit changes the name or color; Delete removes the tag according to the confirmation flow.

The Prompts number counts tag assignments to current prompt records. It is not the number of AI answers, mentions, citations, or runs. The Used by preview is intentionally limited to three prompts; “more prompts” is the remaining assignment count, not hidden performance data.

Tags versus Topics

Use Topic for one primary reporting folder and Tags for overlapping dimensions. A prompt might have the Topic “Running shoes” and Tags for “Awareness,” “UK,” and “Priority Q3.” Multiple tags can apply to the same prompt, so adding counts across tags can double-count prompts.

Color is a visual identifier only. It carries no severity, status, rank, or score unless your team deliberately establishes and documents such a convention.

Tags in reports and calculations

A tag filter keeps executions whose prompts carry the selected tag. Brand and source reports then calculate their metrics from that reduced execution set. If several tags are used in different views, compare their completed response denominators before comparing percentages.

Because one prompt can carry multiple tags, two tag-filtered reports can overlap. Their response counts and mentions must not be added as if the populations were disjoint. Use a governed export or inspect prompt assignments when you need mutually exclusive reporting groups.

Keep the vocabulary usable

Review the tag list as shared reporting metadata, not personal shorthand. Document the inclusion rule and a representative prompt for each important tag so another analyst can reproduce the same filtered population. This matters especially when several teams create prompts or when tag-filtered exports feed recurring stakeholder reporting.

  • Define what each tag means and who may create new ones.
  • Merge or retire near-duplicates such as different spelling or capitalization conventions.
  • Prefer a small stable vocabulary over a tag for every temporary observation.
  • Review tag assignments when prompt intent changes.
  • Include tag definitions when a filtered report is shared outside the operating team.
  • Audit unused tags separately from low-performing prompts; zero assignments describes organization, not visibility.