Managing Prompts

Overview

Prompts are the questions OneGlance asks AI providers on your behalf. Effective prompt management directly impacts the quality and relevance of your brand visibility data.

Changes to prompts take effect on the next scheduled run or when you manually trigger a run. Editing prompts does not re-analyze existing responses.

Accessing the Prompts Page

Navigate to Prompts from the sidebar
Select your workspace from the dropdown (if you have multiple)
The page displays all prompts with their performance metrics

Prompts Page Layout

The prompts interface shows:

// From apps/web/src/app/(auth)/prompts/page.tsx:755
<Table>
  <TableHeader>
    <TableRow>
      <TableHead>Prompt</TableHead>
      <TableHead>GEO Score</TableHead>
      <TableHead>Sentiment</TableHead>
      <TableHead>Visibility</TableHead>
      <TableHead>Position</TableHead>
    </TableRow>
  </TableHeader>
  <TableBody>
    {sortedPromptsWithMetrics.map(({ prompt, metrics }) => (
      <TableRow onClick={() => setOpenPrompt(prompt)}>
        <TableCell>{prompt.prompt}</TableCell>
        <TableCell>{metrics?.geoScore}</TableCell>
        <TableCell><SentimentMetricCell sentiment={metrics.sentiment} /></TableCell>
        <TableCell>{metrics?.visibility}%</TableCell>
        <TableCell><PositionMetricCell position={metrics.position} /></TableCell>
      </TableRow>
    ))}
  </TableBody>
</Table>

Click any prompt row to view individual AI provider responses.

Adding New Prompts

Click Add Prompt

From the Prompts page, click + Add Prompt in the top toolbar.

Enter Your Prompt

Type or paste your prompt in the dialog textarea.

Prompts are case-sensitive and should be phrased exactly as a user would ask an AI assistant.

Good examples:

What are the best CRM tools for sales teams under 50 people?
Which marketing automation platform has the best email deliverability?
Compare Salesforce, HubSpot, and Pipedrive for mid-market B2B companies

Avoid:

CRM tools (too vague)
[Brand Name] vs competitors (AI providers won't understand the placeholder)
best crm????? (poor grammar/formatting)

Save Changes

Click Add to stage the prompt, then click Save Changes to persist all prompt modifications.

You can add multiple prompts before saving. All changes batch together in a single save operation.

Editing Existing Prompts

Select the Prompt

Click the checkbox next to the prompt you want to edit.

Click Edit

The Edit button appears when exactly one prompt is selected.

// From apps/web/src/app/(auth)/prompts/page.tsx:530
<Button
  variant="outline"
  size="sm"
  disabled={selectedRows.size !== 1}
  onClick={() => {
    const idx = Array.from(selectedRows)[0];
    setEditIndex(idx);
    setEditPromptValue(promptData[idx]?.prompt ?? '');
    setDialogOpen(true);
  }}
>
  <Pencil size={16} />
  Edit
</Button>

Modify and Save

Update the prompt text in the dialog, then:

Click Update to stage the change
Click Save Changes to persist

Editing a prompt does not re-analyze past responses. To see updated metrics, run a new analysis after saving.

Deleting Prompts

Select Prompts

Check one or more prompts to delete.

Click Delete

The Delete (N) button shows the count of selected prompts.

Confirm and Save

Prompts are removed from the table. Click Save Changes to finalize.

Deleting a prompt removes it from future runs but preserves historical response data in the database.

Viewing Prompt Responses

Click any prompt row to open a detailed view of AI provider responses. For each AI provider response, you’ll see:

Metrics
Response Text
Sources

GEO Score: Overall visibility score (0-100)
Sentiment: How positively/negatively the brand was mentioned
Visibility: Prominence percentage
Position: Rank in recommendation list (if applicable)

Example from apps/web/src/app/(auth)/prompts/page.tsx:992:

{record.is_analysed && record.brand_analysis && (
  <div>
    <span>GEO Score: {record.brand_analysis.geoScore.overall}</span>
    <SentimentMetricCell sentiment={record.brand_analysis.sentiment.score} />
    <span>Visibility: {record.brand_analysis.presence.visibility}%</span>
    <PositionMetricCell position={record.brand_analysis.position.rankPosition} />
  </div>
)}

The full AI-generated response with markdown formatting.Click View full response to expand lengthy responses:

// From apps/web/src/app/(auth)/prompts/page.tsx:1077
<div
  className={isExpanded ? "" : "line-clamp-3"}
  dangerouslySetInnerHTML={{ __html: formatMarkdown(record.response) }}
/>

Citation links the AI provider referenced:

URL: Source link
Title: Page/article title
Cited Text: Excerpt quoted by the AI

// From apps/web/src/app/(auth)/prompts/page.tsx:1094
<SourcesHoverLinks items={record.sources} />

Filtering Responses

Use the filters in the modal to narrow results:

Model Filter: View responses from a specific AI provider (ChatGPT, Claude, etc.)
Time Range: Filter by when the prompt was run (7d, 14d, 30d, all)

Prompt Strategy Best Practices

1. Use Buyer-Intent Language

Frame prompts as real questions buyers ask during research.

Example: Software Buyer Journey

Awareness Stage:

“What tools help with sales pipeline management?”
“How do modern CRMs compare to spreadsheets?”

Consideration Stage:

“Compare Salesforce vs HubSpot for mid-market teams”
“Which CRM has the best Slack integration?”

Decision Stage:

“Is [Competitor] worth the price for a 10-person sales team?”
“What are the main differences between [Brand A] and [Brand B] pricing?“

2. Cover Multiple Use Cases

Different buyer segments may trigger different AI responses.

✅ Good coverage:
- "Best CRM for startups under 10 employees"
- "Best CRM for enterprise teams with complex workflows"
- "Best CRM for real estate agencies"
- "Best CRM for B2B SaaS companies"

❌ Poor coverage:
- "Best CRM" (only one generic prompt)

3. Test Competitive Positioning

Include prompts that mention competitors directly.

Comparative prompts:
- "Compare [Your Brand] vs [Top Competitor] for [use case]"
- "Why choose [Competitor A] over [Competitor B]?"
- "[Competitor] alternatives for [specific need]"

Alternative prompts:
- "Alternatives to [Top Competitor]"
- "Tools like [Category Leader] but cheaper"
- "Open-source alternatives to [Commercial Product]"

If your brand appears when users ask about competitors, that’s strong AI visibility. Track these prompts closely.

4. Balance Specificity

Mix broad and narrow prompts.

Prompt Type	Example	Purpose
Broad	”What are the best project management tools?”	Measures category-level visibility
Feature-Specific	”Which project management tool has the best Gantt charts?”	Tests feature differentiation
Niche	”Project management tools for remote creative agencies”	Tracks audience-specific positioning
Branded	”Is Asana good for marketing teams?”	Measures direct brand awareness

5. Refresh Prompts Quarterly

AI training data evolves. Update prompts to reflect:

New product launches (yours and competitors)
Emerging buyer questions (check support/sales logs)
Category trends (“AI-powered CRM” vs “CRM with AI”)

Interpreting Prompt Performance

Sorting and Filtering

The prompts table supports sorting by:

GEO Score
Sentiment
Visibility
Position

Use case: Identify which prompts generate the strongest brand visibility.

High GEO (70+): Your brand is well-positioned for this query type
Medium GEO (40-69): Brand appears but not prominently
Low GEO (<40): Brand rarely mentioned or poorly positioned

Action: Invest in content/SEO for low-performing prompt topics.

Status Indicators

Prompts may show different status messages:

No responses yet

The prompt exists but hasn’t been run. This appears when:

You just added the prompt
No scheduled runs have executed since adding it

Solution: Manually trigger a run or wait for the next scheduled execution.

Analysis in progress...

AI providers returned responses, but analysis hasn’t completed.From apps/web/src/app/(auth)/prompts/page.tsx:863:

{reason === 'unanalyzed' ? "Analysis in progress..." : ...}

Typical duration: 1-3 minutes per response.Solution: Refresh the page or wait. Analysis runs automatically after responses are collected.

Brand not mentioned in this prompt

AI providers responded, but your brand didn’t appear in any answers.

{reason === 'brand-not-mentioned' ? "Brand not mentioned in this prompt" : ...}

Common causes:

Prompt topic unrelated to your category
Competitors dominating AI training data
Brand name/domain mismatch (AI uses different name)

Solution: Review the responses to understand why. Consider:

Rephrasing the prompt
Creating more category-specific content
Checking if AI providers recognize your brand name

Exporting Prompt Data

Export prompt performance for external analysis.

Apply Filters (Optional)

Use model and time filters to narrow the export scope.

Choose Export Format

Click the Export dropdown and select:

JSON: Structured data with full nested objects
CSV: Flat data for spreadsheet analysis

Review Export Contents

Both formats include:

// From apps/web/src/app/(auth)/prompts/page.tsx:656
const exportData = {
  generatedAt: new Date().toISOString(),
  workspaceId,
  report: {
    title: 'Prompt Performance Export',
    version: '2.0',
    filters: { modelFilter, timeFilter, sortBy, sortDirection },
  },
  overview: {
    totalPrompts: sortedPromptsWithMetrics.length,
    analyzedPrompts: analyzedRows.length,
    unanalyzedPrompts: sortedPromptsWithMetrics.length - analyzedRows.length,
  },
  detailedData: {
    rows: promptRows.map(({ prompt, metrics, modelProvider }) => ({
      promptId: prompt.id,
      prompt: prompt.prompt,
      modelProvider,
      geoScore: metrics?.geoScore,
      sentiment: metrics?.sentiment,
      visibility: metrics?.visibility,
      position: metrics?.position,
      responses: [...] // Full response objects
    }))
  }
};

Troubleshooting

Save Changes button stays disabled

Cause: No modifications detected or you’re currently editing a prompt.Solution:

Ensure you’ve made actual changes (text modifications, additions, deletions)
Complete or cancel any open edit dialog

Metrics don't update after editing prompt

Expected behavior: Editing a prompt doesn’t retroactively change analysis.Solution: After saving edits, trigger a new run:

Go to Schedule page
Click Save Schedule (triggers immediate run)
Wait for completion, then check updated metrics

Some AI providers show 'no data' for specific prompts

Possible causes:

Provider failed for that specific prompt
Provider disabled in Settings
Rate limiting or API errors

Solution: Check the prompt response detail modal:

If provider is listed but shows error → temporary API issue, retry run
If provider missing → verify it’s enabled in Settings

Can't delete last prompt

Requirement: Workspaces must have at least one prompt.Solution: Add a new prompt before deleting the last one, or accept that you need at least one active prompt.

Next Steps

Scheduling

Automate prompt runs to track visibility trends over time

Interpreting Metrics

Understand GEO scores, sentiment, and competitive positioning in depth

Team Collaboration

Share prompt strategy and results with your team

API Reference

Programmatically manage prompts via the OneGlance API

Setup Guide - Creating your first workspace and prompts
Analysis Concepts - How OneGlance analyzes AI responses
Interpreting Metrics - Understanding aggregate metrics

Get Started

Core Concepts

Guides

Overview

Accessing the Prompts Page

Prompts Page Layout

Adding New Prompts

Editing Existing Prompts

Deleting Prompts

Viewing Prompt Responses

Filtering Responses

Prompt Strategy Best Practices

1. Use Buyer-Intent Language

2. Cover Multiple Use Cases

3. Test Competitive Positioning

4. Balance Specificity

5. Refresh Prompts Quarterly

Interpreting Prompt Performance

Sorting and Filtering

Status Indicators

Exporting Prompt Data

Troubleshooting

Next Steps

Scheduling

Interpreting Metrics

Team Collaboration

API Reference

Build docs developers (and LLMs) love

Get Started

Core Concepts

Guides

​Overview

​Accessing the Prompts Page

​Prompts Page Layout

​Adding New Prompts

​Editing Existing Prompts

​Deleting Prompts

​Viewing Prompt Responses

​Response Details Modal

​Filtering Responses

​Prompt Strategy Best Practices

​1. Use Buyer-Intent Language

​2. Cover Multiple Use Cases

​3. Test Competitive Positioning

​4. Balance Specificity

​5. Refresh Prompts Quarterly

​Interpreting Prompt Performance

​Sorting and Filtering

​Status Indicators

​Exporting Prompt Data

​Troubleshooting

​Next Steps

Scheduling

Interpreting Metrics

Team Collaboration

API Reference

​Related Concepts

Build docs developers (and LLMs) love

Overview

Accessing the Prompts Page

Prompts Page Layout

Adding New Prompts

Editing Existing Prompts

Deleting Prompts

Viewing Prompt Responses

Response Details Modal

Filtering Responses

Prompt Strategy Best Practices

1. Use Buyer-Intent Language

2. Cover Multiple Use Cases

3. Test Competitive Positioning

4. Balance Specificity

5. Refresh Prompts Quarterly

Interpreting Prompt Performance

Sorting and Filtering

Status Indicators

Exporting Prompt Data

Troubleshooting

Next Steps

Related Concepts