Skip to main content
A prompt is one question in one language. It runs every day on every platform of the project, in one or more markets. The All Prompts page lists each prompt once. The plan counts a prompt once for each market it runs in, pooled across the organization’s projects. See Markets and languages and Plan limits.

The tiles

The tiles count the prompts of the open view. On Active and Needs attention, they count the prompts that run, and on Paused, the paused prompts. Suggested shows the figures of Active. After a reword, the tip of the Visibility tile says that answers to earlier wordings are not counted here and are in All Answers. The Visibility tab counts them.

The views

Show switches the list between Active, Needs attention, Suggested and Paused. Active lists the prompts that run. Needs attention lists the running prompts that meet one of its rules. Paused lists the prompts that are not asked any more, with what they measured. Suggested holds the suggestions. The head of the list counts the prompts in view. It also says how many branded prompts the Branded filter leaves out, and how many prompts name no monitored brand in the period. A search box finds a prompt by its words. Language narrows the list to one language, when the prompts are in more than one. The filter row above the page narrows the list by topic and by persona.

Topics

The list groups the prompts under their topics, in the project’s order, with No topic last. A topic’s heading row is tinted, and its prompts sit indented under it.
  • The heading states the topic’s number of prompts and their visibility together.
  • A press on the topic’s name folds its prompts away, and a second press shows them again.
  • The box on the heading selects every prompt of the topic.
  • The three dots on a heading hold Edit and Delete. Edit changes the name and the prompts in the topic. Delete asks where the topic’s prompts go: to another topic, or to none.
  • A topic with no prompt shows its heading on Active while no search, language or topic filter narrows the list, so its menu stays in reach.
  • New topic, under the last topic, creates a topic. Its dialog also selects the prompts that move into it.
  • Topics, beside Personas, lists every topic with the prompts it holds, whatever the filters. It holds New topic, and Edit and Delete for each topic.
  • When suggestions wait under a topic, a line under its prompts counts them. A press opens Suggested on that topic alone.

Personas

Personas, beside Add prompt, lists every persona of the project with its sentence and the prompts that carry it. New persona creates one: who asks, one line on how they ask, and an optional sentence. Edit and Delete work on one persona. Before an edit saves a new sentence, the dialog says how many prompts it rewords. Delete asks where the persona’s prompts go: to another persona, or to none.

Needs attention

Needs attention lists the running prompts under each rule they meet. A prompt that meets two rules shows under both. Each rule is stated in words under its heading.

The columns

The table reads the period that the range control picks. Each figure counts that prompt’s analyzed answers in all its markets. The prompt’s panel shows each market’s figures and latest answers. A press on a column head sorts the prompts by that column inside each topic. When the table is wider than the page, the box and the prompt stay at the left, and the other columns scroll. Visibility and Avg. Position always show. Columns adds SoV, Sentiment, Names a brand and Own site, and this browser keeps the choice. The panel of every prompt holds all six. Sentiment and Avg. Position show only when a prompt that the filters keep has a figure in them. Own site shows once one of those prompts has an answer, so a 0 there means that no answer cites the site. The CSV export always holds all three. “Names a brand” tells two low visibilities apart. When few answers name any brand, the question is one that nobody wins yet. When most answers name a brand but not this one, the competitors win it. The competitor filter does not change this page. Share of Voice here always counts every monitored brand.

The marks under a prompt

names nobody is stricter than “Names a brand”. It counts every company an answer names, monitored or not, and it reads the prompt’s own last collection days, whatever the filters say. A prompt that names nobody measures nothing about visibility. Reword it toward a question that asks for providers, or replace it with a suggestion below. The sidebar badge on All Prompts counts the running prompts that name nobody in every market they run in. The day a prompt starts to name nobody, an Alert mail says so.

The prompt’s panel

A press on a row opens the prompt’s panel at the right of the page. Results and Edit switch its two tabs. Escape, a press beside the panel, or the close button closes it. Results holds the prompt’s figures: Visibility, Avg. Position, SoV, Sentiment, Names a brand and Own site. A prompt in several markets shows one line for each market, with its own Visibility and Avg. Position. A press on a market line shows its latest answers. View sources opens the Sources tab, filtered to this prompt. The prompt filter stays set on the other pages until you clear it. This page still lists every prompt. Edit holds every change to the prompt, in three groups. Each group saves on its own, and its head says what a change costs.
  • Question and topic changes the wording, the language, the topic or the persona, in every market of the prompt. An edit that would give two prompts the same words in one market is refused. A new wording, a new language or a new persona sentence restarts the prompt’s trend, so its save reads Save new wording. The row starts empty and counts only the answers to the new wording. Until the first of them, the row says “New wording, no answer yet” and links to the full history. A paused row says that the new wording is not asked while paused. The earlier answers stay in All Answers, each marked as an earlier wording. A topic is a label, so a move keeps the history.
  • Markets sets the markets the prompt runs in. Each market takes a place on the plan, and the line beside the save says what the change does to the plan. See Markets and languages.
  • Status pauses, resumes or deletes the prompt. Pause stops asking it in every market and frees its places in the plan. The prompt moves to the Paused view. Resume asks it again from the next collection, in every market but a stopped one. It needs room on the plan for each market it starts. Resume is also the ▶ on a paused row. Delete works only for a prompt that never ran in any market. A prompt with history can only be paused, and so can a reworded prompt whose earlier wording ran. Undo adds a deleted prompt again, in all its markets, for a short time after the delete. A paused prompt returns paused.

Adding prompts

  • Add prompt takes the wording, one or more markets, the language, and optionally a topic and a persona. When a prompt with the same words and language already runs in one of those markets, nothing is added.
  • In Add prompt and in the panel, a wording that names this brand gets a note under the field before the save. This brand’s name or one of its other spellings counts. The note says that the prompt saves as a branded prompt. The default figures measure whether AI names this brand unprompted, so every rate leaves the prompt out. The note names the Branded filter, in the filter menu of the filter row, which counts it. The save still goes through.
The box at the start of a row selects the prompt. While prompts are selected, a bar above the list holds Markets…, Topic, Persona, Pause, Resume and Delete. Each one changes every selected prompt, or none. Delete deletes only the selected prompts that never ran, and keeps the others. The menu at the end of a row holds View sources.

The latest answers

The panel lists the latest answer from each platform in one market, a failure or “No AI Overview” included, with a link to the full history. Each latest answer shows its platform and its result: Mentioned, Absent, Technical failure or No AI Overview. A failure and “No AI Overview” count in no figure. An answer opens the answer view. A preview answer of a project’s first day carries Preview. See The first day.

The full history

See full history opens the All Answers page, filtered to this prompt. That list holds every answer to the prompt, the answers to its earlier wordings included.

Suggestions

Questions worth adding, in the Suggested view, proposes new prompts. The line under a topic opens the view on that topic alone, and Show every suggestion shows them all again. Each suggestion comes from one thing that this project’s own answers showed, and it shows that thing as its evidence.
  • Suggest reads the answers of the last 7 days, without branded prompts, and writes the questions. It takes up to 3 signals of each kind of evidence per press. Plan limits states how many presses an organization has per day. The panel says when they are spent.
  • Track this adds the suggestion as a prompt. A suggestion that replaces a quiet prompt also offers Track this and pause the old one. It stops the old prompt in the suggestion’s market, which frees its place in the plan.
  • × dismisses a suggestion for good. After 3 dismissals, a signal gets no new questions until its evidence changes.
A suggestion carries a topic and a persona. A replacement keeps those of the prompt it replaces. A suggestion never creates a topic. Before you accept anything, the panel says how the batch splits between the personas.

Where it appears