# Prompt Explorer

Discover, draft, review, and create tracked prompts from your own questions, search data, discussions, or guided templates.

Canonical: https://docs.equalai.io/ai-visibility/prompt-explorer
Markdown: https://docs.equalai.io/ai-visibility/prompt-explorer.md

Choose **AI Visibility → Prompt Explorer**. The page heading is **Prompt Research**. This workflow creates prompts in the current project; generating drafts is not the same as executing those prompts against models.

## Start a workflow

1. Confirm the current project and choose **Get started** on a method card. A disconnected Google Search Console card can show **Connect & start** instead.
2. In **Target market**, select a country and language. These are single selections despite the plural labels **Target countries** and **Target languages**. Changing country resets the suggested language, so check both before continuing.
3. Review the project-derived brand, domain, topic, and industry context where offered, then choose **Continue**.
4. In **Configure source**, supply the method-specific inputs below and choose **Generate drafts**.
5. In **Review & create**, review and edit the proposed wording, select the prompts to keep, and remove unwanted drafts.
6. Check topic assignments, then choose **Create N prompts**. EqualAI returns to the tracked prompts page.

![Initial Add Manually Target market step with country and language selectors, prefilled EarlySEO brand context, and Continue marked](https://docs.equalai.io/images/guides/prompt-market.webp)

*EarlySEO example at the initial Target market step. **1.** Target countries and languages. **2.** Prefilled brand context. **3.** Continue to configure the source. Each market field accepts one selection; review the example defaults for your own project before continuing. No generated prompts are shown.*

Creating prompts requires project editing permission, an active usable subscription, and available workspace-wide prompt capacity. Draft estimates are not guarantees of the number saved: blank entries and duplicates can be removed, generation can return fewer suggestions, and creation can fail if an applicable limit is reached.

## Choose a source

| Method | Inputs and behavior |
| --- | --- |
| Add Manually | Enter a **Prompt**. The screen also offers an intent choice; newly created prompts from this workflow are currently saved with Research intent rather than preserving that choice. |
| Add in Bulk | Paste prompts one per line. **Questions only** keeps lines ending in `?`; **One per line** accepts the cleaned lines. Leading list markers and empty lines are removed. |
| Google Search Console | Connect an authorized Google account, select available query rows, and choose **Prompt style**: Question, Comparison, or Buyer. Queries become draft questions; this does not import historical AI responses. |
| People Also Ask (PAA) | Supply a **Seed query** and **Depth**: Standard or Deep. The workflow requests live search-question data and can use related keyword data and AI-assisted rewriting. Standard targets up to 8 questions; Deep targets up to 16. Results are not guaranteed to be verbatim PAA entries. |
| Reddit Insights | Search discussions, optionally narrow **Subreddit scope**, and select source discussions. Choose Pain points, Recommendations, or Objections. Drafts are constructed from the selected discussions; they are not a comprehensive analysis of Reddit sentiment. |
| Magic Prompts | Use brand/category context, **Audience**, and **Funnel stage**: Awareness, Evaluation, Purchase, or Retention. Current suggestions are generated from buyer-question templates, not a fresh model analysis of your website. |
| Topic Prompts | Enter **Topic or keyword** and select angles such as pricing, comparison, implementation, or alternatives. The method builds template-based variations. |
| Reputational Prompts | Choose dimensions such as Trust, Pricing, Customer support, Security, Accuracy, or Alternatives. The method creates questions to track; it does not itself measure reputation. |
| Translate Prompts | Paste one prompt per line, select the target country/language, and choose **Localized** or **Literal translation**. This method uses AI generation, so review the meaning, brands, and language before saving. |

The **Generate prompt ideas** group combines guided templates with AI translation. Treat all output as drafts to review, not verified evidence about your market.

![Prompt Research method cards for Google Search Console, People Also Ask, Reddit, and AI-assisted prompt sources](https://docs.equalai.io/images/guides/prompt-explorer.webp)

***1.** Search-data sources: GSC, PAA and Reddit. **2.** AI-assisted prompt sources. These cards are entry points, not generated results; the method table above distinguishes live source data, templates, and translation.*

## Work with source data

For Search Console, see [Google Search Console setup](https://docs.equalai.io/integrations/google-search-console). The query table includes clicks, impressions, and position. Its selected-query average position is an arithmetic average, not an impression-weighted aggregate. The connection's site determines the data; picking a target market does not add a country filter to that Search Console query request.

Reddit search offers sort choices Relevance, Hot, Top, New, and Most comments, plus time ranges from Past day through All time. Choose discussions that actually support the questions you want to track. Searching and selecting do not post anything to Reddit.

PAA, Reddit search, and translation depend on their relevant service availability. If the screen reports an unavailable service, retry later or use manual entry; do not treat an error as an empty research result.

## Review topics and duplicates

Drafts can be grouped into topics automatically. Read the topic shown with each prompt; grouping failure is not fatal and prompts can fall back to the workflow's topic. Small batches may remain under that shared topic. The Create action waits while grouping is in progress.

Edit wording before saving and use **Remove prompt** for unwanted drafts. Duplicate normalization during creation ignores case, repeated whitespace, and trailing question/exclamation/full-stop punctuation within the submitted batch. An exact-text match already in the current project reuses that prompt rather than creating another. Broader normalization does not apply to every existing prompt, and reusing one does not update its saved topic or locale.

After creation, use **Tracked prompts** to review the resulting inventory and continue the tracking workflow. Selecting a target locale records the prompt's locale; it does not prove that every template-based draft has been translated into that language.

## Resolve a blocked step

- **Generate drafts** unavailable: complete the required source input or select at least one query/discussion/dimension as appropriate.
- No Search Console queries: verify the connected site's access and available data; reconnect as a workspace administrator if required.
- Empty or unsuitable PAA suggestions: try a clearer seed, verify country/language, or enter reviewed prompts manually.
- Translation differs from the source meaning: edit or remove the draft before creating it.
- Creation error: inspect the message and check the tracked prompts list before retrying a large batch, so you do not unintentionally add duplicates.