> ## Documentation Index
> Fetch the complete documentation index at: https://www.usenotra.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Prompts
> Prompts are the questions Notra asks AI engines for you. Add them by hand, import a CSV or accept suggestions from Google Search Console.
Prompts are the questions Notra asks AI engines for you, and competitors are the brands it looks for in the answers next to yours. Both sit under the **Visibility** group of the GEO sidebar and belong to the active project.
The **Prompts** page lists every prompt that goes into a scan, and prompts come from two sources:
* **Custom** prompts you add by hand, import or accept from a suggestion, which Notra stores per project with an `enabled` flag.
* **Auto** prompts derived from the project's brand context. Notra reads the brand identity's company description and audience, extracts a product category and an audience phrase, and generates up to eight category questions from those that never name your brand. Notra regenerates auto prompts from the brand context on every read, so they always show as enabled.
Auto prompts have no `createdAt` in the API.
Write each prompt the way a person types into ChatGPT, with lowercase, one intent and no trailing question mark. Usually add a concrete detail about the person's situation (role, team size, budget, stack), and never use the brand's own category jargon. A custom prompt must be between 8 and 300 characters. There is no limit on how many custom prompts a project can track, but every prompt multiplies the answers a scan uses. During onboarding Notra drops suggested prompts that contain your company name or an alias, because a branded question tells you nothing about unaided recall.
## Adding and pausing prompts
Press **Add Prompt** (shortcut `P`) to open a dialog with a **Question** field and, for suggestions, a **Website** field. For each prompt, the prompts table shows a **Presence** badge, the **Engines** that mentioned you and the **Best** position. The badge reads **Search only** when only the grounded engines named you and **Invisible** when no engine did.
You can pause a custom prompt without deleting it, in which case the next scan skips it and the prompt keeps its history.
```bash Create theme={"system"}
curl -X POST https://api.usenotra.com/v1/projects/$PROJECT_ID/geo/prompts \
-H "Authorization: Bearer $NOTRA_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "prompt": "what tools should i use for automating changelogs" }'
```
```bash Pause theme={"system"}
curl -X PATCH https://api.usenotra.com/v1/projects/$PROJECT_ID/geo/prompts/$PROMPT_ID \
-H "Authorization: Bearer $NOTRA_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "enabled": false }'
```
```bash Delete theme={"system"}
curl -X DELETE https://api.usenotra.com/v1/projects/$PROJECT_ID/geo/prompts/$PROMPT_ID \
-H "Authorization: Bearer $NOTRA_API_KEY"
```
`GET /v1/projects/{projectId}/geo/prompts` returns both sources in one list, and each item has `id`, `prompt`, `enabled`, `source` (`custom` or `auto`) and `createdAt`.
## Bulk import
**Import CSV** on the Prompts page accepts a file with a `prompt` column and an optional `enabled` column. Before you confirm, the dialog summarises **Ready to import**, **Duplicates skipped** and **Rows with problems**. One import takes up to 500 rows and 1 MB, so run more imports to add more.
The API endpoint accepts either structured `rows` or raw `csv` text, and it skips prompts that already exist instead of duplicating them.
1 to 500 objects with `prompt` (8 to 300 characters, required) and `enabled` (boolean, optional).
Raw CSV text, up to 1 MiB, with a `prompt` column and an optional `enabled` column. The API uses it when you omit `rows`.
```bash Rows theme={"system"}
curl -X POST https://api.usenotra.com/v1/projects/$PROJECT_ID/geo/prompts/import \
-H "Authorization: Bearer $NOTRA_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"rows": [
{ "prompt": "what tools should i use for automating changelogs" },
{ "prompt": "best way to turn github activity into a blog post", "enabled": false }
]
}'
```
```bash CSV theme={"system"}
curl -X POST https://api.usenotra.com/v1/projects/$PROJECT_ID/geo/prompts/import \
-H "Authorization: Bearer $NOTRA_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "csv": "prompt,enabled\nwhat tools should i use for automating changelogs,true" }'
```
```json theme={"system"}
{
"imported": 2,
"updated": 0,
"skipped": 0,
"issues": [],
"organization": { "id": "org_123", "slug": "acme", "name": "Acme", "logo": null }
}
```
`issues` lists rejected CSV lines as `{ line, message }` and is always empty for `rows`. Import allows 10 requests per 10 minutes per organization.
## Search Console suggestions
The Prompts page also hosts the **Google Search Console** card: "We read the queries your site ranks for and suggest the AI prompts people ask. Suggestions refresh weekly." Press **Connect Search Console** and choose a **Property**. Notra then syncs the last 28 days of queries every Monday, and **Sync now** runs it immediately. Suggestions appear in a table with **Search queries**, **Impressions**, **Clicks** and **Best position**, and **Track** turns one into a custom prompt. Suggestions you have not accepted also feed the search gaps on [Content gaps](/docs/geo/content-gaps#search-gaps).
## Writing good prompts
Use lowercase, one intent and a concrete detail about the person (role, team size, stack), and skip your own category jargon.
A branded prompt tells you nothing about whether engines recommend you unprompted.
Mix problem questions ("how do i keep release notes up to date"), category questions ("best changelog tools") and comparison questions ("globex alternatives").
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.