Managing Prompts: A Complete Guide

Prompts are the core unit of measurement in GrackerAI. Each prompt is a real question your buyers might ask an AI search engine — "What's the best EDR for healthcare?" or "Acme Auth vs Auth0 for B2B SaaS?" — and GrackerAI tracks how AI engines respond to it over time.

The Prompts page (Dashboard → Monitoring → Prompts) is where you build, organize, and analyze your prompt library. This guide covers how to add prompts, import and export them in bulk, understand the classification system, and write prompts that actually surface the visibility gaps you need to close.

Examples throughout use B2B SaaS and cybersecurity scenarios.

What the Prompts Page Shows You

When you open the Prompts page, you'll see two summary cards at the top and a sortable, filterable table below.

Types Distribution

This card breaks your prompt library into three buckets:

• Organic — Category, problem-aware, and educational prompts that don't name your brand or a competitor. Example: "What is the best vulnerability scanner for cloud-native environments?"
• Brand Specific — Prompts that name your own brand directly. Example: "Acme Auth pricing and reviews"
• Competitor Comparison — Prompts that name one or more competitors, with or without you. Example: "Splunk vs Sentinel for SOC teams"

A healthy library skews heavily toward Organic prompts (this is where most AI search demand actually sits), with a smaller dedicated set of Brand Specific and Competitor Comparison prompts for monitoring and switcher intent.

Intents Distribution

This card breaks your library by buyer intent:

• Informational — The buyer is learning. "How does SAML SSO work?" "What is XDR?"
• Commercial — The buyer is evaluating tools. "Best CRM for mid-market SaaS." "Top EDR platforms for healthcare."
• Branded — The buyer already knows a specific brand and is researching it. "Is Datadog worth the cost?" "CrowdStrike Falcon reviews."

A balanced library has a mix of all three. Pure Informational prompts catch buyers early in the journey but convert slowly. Pure Commercial prompts are higher-intent but more competitive. Branded prompts measure your reputation and your ability to capture switcher intent.

The Prompt Table

Each row in the table is one prompt. The columns:

The Brands column is the highest-signal piece of the table. If a prompt shows your icon, you're being cited. If it shows competitor icons but not yours, you have a gap to close. If it shows a "+3" or "+5" badge, multiple brands are showing up — typical for category and listicle prompts.

Adding Prompts

There are three ways to add prompts. Each one is suited to a different stage of building out your library.

1. Add Prompt (Manual)

Click + Add Prompt to create a single prompt. Use this when you have a specific question in mind — usually triggered by a sales conversation, a competitor's new launch, or a gap you spotted in the Brands column.

When adding a prompt, you'll set:

• Prompt text — The full natural-language question.
• Type — Organic, Brand Specific, or Competitor Comparison.
• Intent — Informational, Commercial, or Branded.
• Language — Defaults to English (US). Change for non-English markets.
• Group — Optional grouping label (more on groups below).

Adding prompts one at a time is fine for the first 10–20. After that, bulk import is faster.

2. Bulk Import (CSV)

Use the dropdown next to + Add Prompt to bulk-import prompts from a CSV file. This is the fastest way to seed a library when you're starting from scratch or migrating from a spreadsheet of prompts you've already drafted.

A typical import file looks like this:

prompt,type,intent,language,group
"Best EDR tools for healthcare SaaS",Organic,Commercial,English,EDR-Healthcare
"CrowdStrike alternatives for mid-market",Competitor Comparison,Commercial,English,Comp-CrowdStrike
"How to detect ransomware before it spreads",Organic,Informational,English,Prob-Ransomware
"Acme Security vs SentinelOne",Competitor Comparison,Branded,English,Comp-SentinelOne
"Acme Security pricing and plans",Brand Specific,Branded,English,Brand-Pricing

A few rules that prevent most import errors:

• Wrap prompt text in quotes if it contains commas or special characters.
• Use exact category names — Organic, Brand Specific, Competitor Comparison for Type; Informational, Commercial, Branded for Intent. Typos cause rows to import as "uncategorized."
• Group names should follow your monitor naming convention so the two systems stay aligned. If you have a monitor called Comp-Splunk-Alternatives, name the prompt group Comp-Splunk-Alternatives too.
• One prompt per row. Don't combine prompts with semicolons or pipes.

If a row fails to import, GrackerAI will flag it with the reason — usually a typo in Type or Intent, or a missing required field. Fix and re-upload only the failed rows.

3. Prompt Generator (AI-Assisted)

Find this in the left sidebar under Actions → Prompt Generator. The Prompt Generator uses AI to suggest prompts based on your domain, ICP, and competitor set. It's the right tool to use when:

• You're starting a new monitoring effort and don't have a baseline list yet.
• You've identified a gap (a competitor showing up where you don't) and want suggested prompts to expand coverage in that area.
• You're entering a new vertical or geography and need to seed prompts you haven't thought of yet.

The Prompt Generator is a starting point, not a finished product. Always review generated prompts before activating them — the model sometimes produces prompts that are too generic ("best software 2026") or too specific to be useful. Edit, delete, or accept each one before they enter your library.

Exporting Prompts

The Export button in the top right of the prompts table downloads your full prompt library as a CSV. The export includes every column visible in the table — visibility score, prompt text, status, type, intent, brand mentions, and response count.

Common reasons to export:

• Quarterly reporting to leadership or board — pull the export, calculate average visibility, and chart trends.
• Sharing with agencies or contractors who handle your content but don't need full GrackerAI access.
• Backup before bulk operations — export, then bulk-edit or bulk-delete with confidence that you can restore.
• Migrating prompts between accounts — useful for agencies managing multiple clients or for moving from a sandbox account to production.

If you only want a subset (e.g., just competitor prompts for a specific competitor), filter the table first using the filters above the table. The Export will only include rows currently visible.

Filtering, Searching, and Sorting

The Prompts page is built for libraries of hundreds of prompts. As your library grows, the filter controls become essential.

Filters

Above the table:

• All types — Filter by Organic, Brand Specific, or Competitor Comparison.
• All languages — Filter by language. Useful if you've set up monitors across multiple geographies.
• All groups — Filter by the group label you assigned. This is the most powerful filter once you've adopted a consistent naming convention.
• Per page — Default is 20. Bump it to 50 or 100 when reviewing the full library.

Search

The Filter by prompt search box does a text match against the prompt itself. Useful when you know roughly what you typed but don't remember the exact group. Search "ransomware" to find every ransomware-related prompt across all groups.

Sorting

Click a column header to sort:

• Visibility ↓ — Sort highest-visibility first. Use this to find your wins.
• Visibility ↑ — Sort lowest-visibility first. Use this to find your gaps. This is the most useful sort. Your worst-performing prompts are usually where the highest-leverage content opportunities sit.
• Prompt ↕ — Alphabetical. Useful when scanning for duplicates.
• Responses ↕ — Sort by how many AI responses GrackerAI has captured. High response count = AI engines actively answer this query, so visibility here matters more.

Bulk Actions

Each row has a checkbox on the left. Select multiple rows (or use the header checkbox to select all visible rows) to perform bulk actions:

• Bulk pause / activate — Useful when temporarily disabling a category of prompts (e.g., pausing all monitors during a holiday freeze).
• Bulk delete — Use carefully. Export first.
• Bulk re-tag — Change Type, Intent, Language, or Group on multiple prompts at once. Useful when you realize half your "Informational" prompts are actually "Commercial" intent.

Bulk operations only apply to the rows currently selected — they don't apply globally. If you want to delete all prompts in a group, filter to that group first, then select all visible.

Groups: The Most Underused Feature

The Group field is optional, but it's the single biggest organizational lever in the Prompts page. Groups let you slice your library by theme — by competitor, by use case, by buyer persona, by vertical — without creating separate accounts or monitors.

A useful grouping convention mirrors the monitor naming convention from our Setting Up AI Monitors guide:

Example for a B2B SaaS observability vendor:

• Brand-Reviews, Brand-Pricing, Brand-Compliance
• Comp-Datadog, Comp-NewRelic, Comp-Grafana
• Cat-Observability-Best-Of, Cat-APM-Tools
• Vert-Fintech, Vert-Ecommerce
• Prob-Latency-Debugging, Prob-Cost-Optimization

Example for a cybersecurity EDR vendor:

• Brand-Reviews-CISO, Brand-FedRAMP, Brand-Pricing
• Comp-CrowdStrike, Comp-SentinelOne, Comp-Defender
• Cat-EDR-Best-Of, Cat-XDR-Best-Of
• Vert-Healthcare, Vert-FinServ, Vert-Critical-Infra
• Prob-Ransomware, Prob-Insider-Threat, Prob-Audit-Readiness

Once groups are set, every report and filter becomes 5x more useful. You can answer "what's our visibility on CrowdStrike-comparison prompts?" in two clicks.

Connecting Prompts to Monitors and Responses

Prompts don't exist in isolation. They're linked to two other parts of GrackerAI:

• Monitors (Monitoring → Monitors) — Each prompt belongs to one or more monitors. The monitor controls when the prompt runs (daily, weekly, monthly), which AI engines it queries, and what persona/location it simulates. The prompt itself is the question; the monitor is the experimental setup.
• Responses (Monitoring → Responses) — Every time a prompt runs, the AI-generated response is captured. Click into any prompt to see the actual response text from each AI engine, which sources were cited, and which brands were mentioned.

The typical workflow:

1. Add prompts to your library — manually, via bulk import, or via the Prompt Generator.
2. Group prompts logically.
3. Create monitors that target specific groups (e.g., a monitor that runs all Comp-CrowdStrike prompts daily, on ChatGPT, Perplexity, Claude, and Gemini).
4. Review responses to understand why you're winning or losing each prompt — what content is being cited, what your competitors are saying, and where the content gaps are.
5. Use Recommendations (Actions → Recommendations) to see GrackerAI's suggestions for closing gaps.

Writing Effective Prompts

The single highest-leverage skill in GrackerAI is writing prompts that actually mirror what your buyers ask. A few rules:

Write Full Natural-Language Questions

AI search users type the way they speak. They include context, qualifiers, and constraints.

The strong versions are how real prospects phrase questions. The weak versions are Google search queries — AI engines may answer them, but the answers are generic and don't surface the prompts where your differentiation actually matters.

Include Buyer Context

For every prompt, ask: who would type this and what context would they include? Real context includes:

• Company size ("for a 200-person company," "for an early-stage startup," "for enterprise teams")
• Industry or vertical ("for healthcare," "for fintech," "for SaaS companies")
• Tech stack ("on AWS," "with Kubernetes," "Microsoft 365 environments")
• Compliance requirements ("SOC 2 ready," "HIPAA compliant," "FedRAMP authorized")
• Budget tier ("under $1000/month," "for cost-conscious teams," "enterprise budget")
• Use case ("for ransomware prevention," "for SOC analysts," "for product-led growth")

A prompt without context is a category prompt. A prompt with context is a buyer prompt — and buyer prompts are where pipeline lives.

Mix Intent Types Within a Group

A monitor or group with only Commercial prompts gives you a narrow view. A balanced group has:

• 1–2 Informational prompts ("How to detect ransomware," "What is EDR")
• 2–3 Commercial prompts ("Best EDR for healthcare," "Top XDR platforms")
• 1–2 Branded prompts ("CrowdStrike vs Acme Security," "Acme Security reviews")

This spread lets you see how your visibility evolves across the buyer journey — not just at the bottom.

Don't Forget the Negative Space

The most valuable prompts are often the ones you're not yet ranking on. After your first two weeks of data, sort by Visibility ↑ to see the prompts where competitors are showing up but you're not. These are the prompts to add content for, not the ones where you already win.

Cybersecurity-Specific Prompt Tips

If you're in cybersecurity, your prompt library should include a few categories that don't apply to general B2B SaaS:

• CVE-specific prompts — "Best tools to mitigate CVE-2026-XXXX" or "Response strategy for [active CVE]." These have short shelf lives but high urgency. Add them when a major CVE drops; retire them when the news cycle ends.
• Breach-event prompts — "Tools that would have prevented [breach name]." Capture switcher intent during industry incidents.
• Compliance-framework prompts — "Best DLP tools for HIPAA," "Tools for PCI DSS continuous monitoring," "FedRAMP-authorized SIEM platforms." These are evergreen and high-intent.
• Threat-pattern prompts — "How to detect lateral movement," "Best tools for insider threat monitoring," "Ransomware kill chain detection."

These prompt types over-index on Perplexity and Claude (where security practitioners do most of their research), so make sure those engines are selected on the monitors that run them.

Maintaining Your Prompt Library

A library that worked six months ago probably doesn't work today. AI engines change, competitors emerge, and buyer language shifts. Treat your prompt library as a living document.

Monthly Review (15 minutes)

• Sort by Visibility ↑ and look at your bottom 10 prompts. Do they need new content, or are they misaligned with what buyers actually ask?
• Check the Brands column for new competitor icons. New entrants in your category will appear here before they appear in your sales pipeline.
• Look at Responses count. Prompts with very few responses may not be common enough — consider replacing them with higher-volume questions.

Quarterly Review (1 hour)

• Export the full library and compare against last quarter's export. What's improved? What's regressed?
• Audit your Type and Intent classifications. Misclassified prompts skew your distribution charts.
• Retire dead prompts — anything with 0% visibility AND minimal competitor presence over a quarter is a prompt no one is asking. Replace with new hypotheses.
• Add prompts for any new competitors, new product features, or new verticals you've entered.

After Major Events

Spin up new prompts (and retire old ones) after:

• A competitor's funding round, acquisition, or product launch
• A major industry breach or vulnerability disclosure (cybersecurity)
• A significant pricing or packaging change on your side
• Entering a new geography or vertical
• Launching a new product line or major feature

Common Mistakes to Avoid

Treating prompts as keywords. "EDR pricing" is a search phrase. "How much does CrowdStrike Falcon cost for a 500-person company?" is a prompt. Always write in full questions.

Skipping the Group field. Without groups, a 200-prompt library becomes unfilterable. Adopt a naming convention from day one.

Importing too many prompts at once. Bulk-importing 500 prompts on day one means you can't tell which ones are working. Start with 50–100, get a clean two-week baseline, then expand.

Ignoring the Type and Intent classifications. These power the distribution charts and the filter controls. A library where everything is misclassified as "Organic / Informational" is a library where you can't tell what's working.

Only tracking Branded and Competitor prompts. This is the most common mistake. Branded prompts measure reputation, not demand. Most actual demand sits in Organic Commercial prompts ("best [category] for [use case]"). Skew your library accordingly.

Adding prompts and forgetting them. Without monthly and quarterly reviews, your library decays. Visibility scores are only useful if you act on them.

Using the Prompt Generator output without editing. AI-suggested prompts are starting points, not finished products. Always review and edit before activating.

Need Help?

If you'd like a custom prompt library tailored to your industry, ICP, and competitive landscape, book a call with our team (opens in a new tab) or reach out via the in-app chat. We work with B2B SaaS, cybersecurity, and fintech teams to design prompt libraries that map directly to revenue motions.

For more guides, visit gracker.ai/docs (opens in a new tab).