Skip to content

Prompt explorer

Prompt explorer is a preview available only to ReachLLM platform admins. Authorized admins can open it in the Track section of the sidebar and choose a brand. The page remains at /prompt-research; direct links and agent workspace actions use the same admin and brand-access checks. Customer and team-admin roles do not grant preview access.

Other users can choose Generate in Your Prompts for independently reviewed additional questions. Normal onboarding and prompt generation keep the same quality checks.

Start with a product or service, then open its buyer needs, topics and clusters to see relevant questions. For example, a brand with several products can have separate branches for each product instead of one long question list. Branches use the business information and question labels available; not every branch has every level.

Use the breadcrumbs above the list to return to a broader topic or the whole brand. Search questions searches within your selected branch. Filter by Suggested, In tracking or Archived, or keep All questions selected. Browsing, searching and filtering saved questions do not start new research.

Counts describe saved questions, not searches submitted by buyers or measured market demand. Older questions without supported product or topic labels appear under Unassigned questions. Their reviewed topics remain available inside that branch, without inventing a product association. Existing tags and prompt wording are preserved. A missing label alone does not establish an uncovered buyer need.

Select Find questions to prepare up to 25 reviewed suggestions. Open a supported product or topic first to focus the run there, or start from the brand to research more broadly. The system saves approved questions progressively in batches of up to five. A run can return fewer questions rather than padding the set.

Research runs in the cloud, so you can navigate away or close the page. Return to see saved questions and progress. Temporary failures receive bounded automatic retries; the status shows when this happens. Earlier saved questions remain available during another run and after a stopped run. A quality rejection is not bypassed by accepting unreviewed questions.

New questions are reviewed against the verified business and its buyers. Review considers business fit, buyer relevance, natural wording, supported constraints, sufficient context and distinct meaning. Short requests, comparisons and non-English questions can be valid. Questions do not need an invented persona or a long scenario to qualify.

Open Why this question? for its rationale, available source passages and demand limitations. Generated wording stays labeled Suggested, including when real search or customer evidence informed it. If an exact source passage is unavailable, the suggestion does not claim one.

Choose Track on a question, or select several and choose Track selected. This appends eligible questions to Your Prompts using your normal tracked-prompt allowance. It preserves existing prompts, history and archive choices, and does not start a visibility run.

The More options menu keeps supporting work out of the main question list:

  • Add customer questions saves questions from sales, support or interviews, one per line, with an optional source label. Retain any context a follow-up needs. These are supplied observations, not automatically verified native AI conversations. A rewritten question needs its own visibility check.
  • Sources shows supporting records and connection status.
  • Review tracked questions diagnoses wording and context without rewriting, removing or archiving tracked prompts. Low or unchanged visibility alone is not a reason to remove an important buyer question. Keep a stable comparison set when measuring progress.
  • Research focus lets you choose priorities and Save focus for a whole-brand run. Finding questions inside a supported product or topic uses that opened scope instead.
  • How it works explains the method and links to citation discovery.
  • Archived questions opens the archived filter. Use Restore to bring back a suggestion. Archiving preserves its evidence, and researching again does not automatically discard earlier suggestions.

The Sources panel distinguishes the available signals:

  • Brand information provides business context. It does not measure buyer demand.
  • Google Search Console can contribute actual query/page observations for the brand's connected property. Clicks and impressions retain their Google Search units and date range. They are not AI conversation counts. The workspace uses the canonical brand owner's connection and can ask you to check it in Traffic if access needs attention.
  • Customer questions preserve the language you supplied and its source label.
  • Observed AI conversations remain unavailable until an appropriately licensed feed is connected. The presence of a prompt map does not imply access to private ChatGPT conversations.

Search Console returns top available rows and may omit queries. Empty results mean that no usable rows were returned for that period, not that the market has no demand. An unavailable connection does not erase earlier source evidence. New runs can use newer Search Console observations even after the current source library fills. Original customer questions and the source details saved with earlier suggestions remain available.

Exact-prompt demand can be unknown even when the question fits your business. A generated prompt does not inherit the search counts of a related keyword or topic. The explorer is an organized set of questions to evaluate, not a complete record of what people ask AI assistants.

New-brand setup and free reports keep their existing complete 50-prompt requirement. Their generation uses the same business and quality checks. If preparation or review cannot finish, setup remains recoverable instead of being padded with generic questions.

Generate and Generate more open Prompt explorer for authorized platform admins. For other users, the confirmed generation runs in the cloud and saves reviewed questions in Your Prompts > Saved. Generate to Saved in the Saved tab uses this workflow for authorized users, including admins. Existing prompts remain. Review the questions, then explicitly choose which to test once or start tracking. Generation does not start paid visibility checks.

Research and adding suggestions use no execution credits. They do not run paid AI visibility checks. Use Your Prompts when you are ready to run tracking, or Discover prompts to look for questions where a fresh answer mentions your brand or cites your website. Those checks use the normal enabled-platform and execution-credit rules, with the applicable cost shown before a manual run.

Prompt explorer proposes relevant questions. Discovery saves verified appearance opportunities. A website citation means the saved answer linked that page; it does not prove that real buyers asked the exact generated wording or that the page caused a recommendation.

Research has bounded operational limits. If another run is active or the workspace reaches its current research limit, read the saved suggestions and try again when the limit clears.

During the admin preview, authorized platform admins can ask the agent to explore a product or topic, find questions, save a research focus, review the current set or add specific saved suggestions to tracking. Other users can use Your Prompts or the agent to save ideas and generate reviewed questions into the Saved library, without admin preview access. The agent uses the same workspace, permissions and quota checks as the app. It reads saved progress rather than restarting an active run. Only persisted questions that passed review are ready to add, even if later batches are still running. It should explain where evidence is unavailable.