GEOflux docs
Prompts

Reading prompt performance

The full analytics for one prompt, and when to use them

The Overview tells you how you do across all your prompts. When a topic underperforms, or you have just shipped content aimed at one question, you need to zoom into a single prompt: which engines name you, who you lose it to, which sites the answers rely on, and what the answers actually say.

Open any prompt from the Prompts page, or from a link anywhere it appears, to reach its detail page. It uses the same building blocks as the Overview, scoped to that one prompt and your selected period.


What you see

From top to bottom:

  • Header: the prompt, the engines that answered it, and its topic, tags, Buyer Journey step and Branded vs Non-branded value. You can change all of them right here. A footer shows when tracking started and how many answers the prompt has in the period.
  • Summary strip: Visibility, Share of voice, Sentiment (0 to 10) and Mentions for your brand, and your rank among your brand and tracked competitors on this prompt.
  • Visibility over time: the share of this prompt's answers that name each brand, Over time or By brand.
  • Share of Voice: a Composition donut or Over time, with the vs. all brands / vs. competitors toggle. With no competitors tracked it stays on all brands.
  • Sentiment: your brand's tone Over time, or every brand's score By brand, with the score zones marked.
  • Rankings: your brand and your competitors, ranked by visibility by default. Click a column header to sort by share of voice, sentiment or mentions instead, and click a brand to focus the page on it.
  • Competitor breakdown: how each engine treats you against each competitor, so an engine where you are missing stands out.
  • All Brands: every brand that came up in this prompt's answers, with the date each was first seen. Click one to open its profile.
  • By persona: visibility, sentiment and share of voice for each persona attached to the prompt, against the base answer. Without personas, the card links you to create one.
  • Where the answers come from: the domains cited when engines answer this prompt. These are your citation targets for this exact question; each links to its source page.
  • Answers: the answers themselves, 12 per page, as a grid or a table. Click one to read it in full.

Filters

The filter bar offers Period, Brand (your brand by default), Engine and, when your project has personas, Persona. Pick one persona at a time: with two or more selected, the page shows all answers. Topic, tag, Buyer Journey and Branded vs Non-branded filters are not offered, since the page already covers one prompt. See Filters and time periods.

On the Free plan, prompts run every 3 days, so over-time charts plot only the days a run happened, and change badges read too few runs until the period holds at least 3 runs.


Warnings you may see

  • "This prompt may not be right for tracking brand visibility": no answer in the selected period named any brand, and the period covers at least 3 days of the prompt's life. The banner states how many days it looked at. A prompt like this cannot report visibility, share of voice or sentiment; rewrite it as a recommendation or comparison question. See Prompt warnings.
  • "No answer to this prompt mentions [your brand] or a competitor you track": over the last 7 days the answers named brands, but never yours or a tracked competitor. If those brands are competitors, track them; if not, the question is off-topic for what you sell, so rewrite or archive it. This needs at least 3 answers, a prompt at least 3 days old and at least one tracked competitor. When both warnings apply, only the first shows.
  • "No answers captured for this prompt in this period": pick a longer period, or wait for the prompt's first run. The header stays, so you can still edit the prompt.

Above each over-time chart, a notice says when today's answers are still coming in: today's point is incomplete and may change. It follows the project's run as a whole, so it can show even when this prompt's own answers are in. See When data updates.


When to use it

  • Find what drags a topic down. On the Overview's Topics tab, expand the weak topic in the Topic Leaderboard to see its prompts, then open the ones where competitors appear and you do not. Those are the most valuable prompts in your set: gaps you can work on.
  • Plan content for one question. The Answers show what the engines say and in what shape (lists, comparisons, step-by-steps); Where the answers come from shows which pages they trust. Together they tell you what to write and where to get mentioned.
  • Watch one question recover. After a content change or a PR push, pin a chart note on that day on the Overview, then watch this prompt's visibility here over the following weeks. Expect weeks, not days.
  • Check an audience. Compare the base answer with each persona's in By persona before deciding a persona is worth its credits.

To search or export answers across all your prompts, use the Answers page.

On this page