How many prompts should I track for AI search?

Scrunch recommends estimating prompt tracking needs using a simple formula: number of core topics × 5–8 related questions per topic. The goal is to get a representative sampling across all customer journey stages with a mix of branded and non-branded prompts.

As a rule of thumb, most brands should start with 15–25 varied prompts across 6–12 core topics to gather meaningful AI search performance data.

Example workflow

• Identify core topics
  Start with existing SEO keywords, paid search terms, and major product and content areas. Filter to the top 20–30% most relevant to your brand and most likely to drive demand.
  Example: From 3,000 keywords and product/content areas, filtering to the top 25% yields 750 core topics.

• Identify related questions
  Create 5–8 questions per topic that map to all funnel stages (advice, awareness, evaluation, comparison). Include both branded and non‑branded variants informed by SEO research, PPC queries, and questions fielded by sales, success, and support.
  Example: Six questions per topic.

• Calculate total prompts
  Multiply topics × questions.
  Example: 750 topics × 6 questions = 4,500 prompts.

If you’re tracking a large volume of prompts, organize them by topic, persona, funnel stage, and other filters to enable clearer reporting and analysis.

How to validate your prompt strategy against real user behavior

• Track AI user‑agent traffic and AI referral traffic patterns. Compare the most frequently crawled and visited URLs with your prompt list to ensure coverage.
• Refresh your prompt list over time by:
• Reviewing AI search trend data monthly to identify new topics and questions
• Adding prompts quarterly to reflect launches, campaigns, and competitive shifts
• Archiving or updating prompts that are no longer relevant

How Scrunch measures AI search impact

Scrunch focuses on outcomes and visibility signals that indicate brand performance in AI search:

• Brand presence and share of voice across prompts and platforms
• Citations and source attributions within AI responses
• AI agent traffic and AI referral traffic to your site
• Trends over consistent 2–3 week windows to separate real movement from noise
• Cross‑model and cross‑platform comparisons by running the same prompts across multiple AI systems, then filtering to see where performance diverges

Learn more in these related FAQs: - What benchmarks or baselines are useful when evaluating AI search performance? - How can I see if my visibility in AI search is improving or declining over time? - How do I compare AI visibility across different models and platforms?

Capabilities that support accurate monitoring

• Collects actual AI responses from major AI platforms by simulating consumer interactions (e.g., asking Perplexity a question and storing the response)
• Detects explicit brand mentions using pattern matching validated against continually updated response datasets
• Extracts citations and applies AI for sentiment analysis, topic classification, named entity extraction, and factual comparison of texts
• Monitors AI agent and referral traffic to connect visibility signals with downstream impact

See: - Does Scrunch use AI in its platform?
- How accurate is Scrunch at tracking brand presence across LLMs?

Key features and plans

Scrunch’s platform spans four core products: - Monitoring - Auditing - Optimization - Content delivery via the Agent Experience Platform (AXP)

Highlights: - Fast implementation; many enterprises are live and collecting data within one day. See how long implementation takes. - Self‑serve Explorer plan with a 7‑day free trial. See free trial details. - Pricing options for different needs. Explore pricing. - Deliver AI‑optimized content to AI agents on your site with AXP, while maintaining your human experience for regular visitors.

For an overview of what’s included, see what products Scrunch offers.

Data usage and privacy

• Direct AI model usage is performed within Scrunch’s cloud environment or via commercial inference providers (e.g., OpenAI, Google Vertex AI, Together) under agreements that prohibit using interactions for training or fine‑tuning.
• User data is not used to train external AI models. Learn more: Does Scrunch use AI in its platform?

Trust and track record

Scrunch has raised $26M from Mayfield, Decibel, Homebrew, GTM Capital, and notable angels including TJ Parker (PillPack), Bryant Chou (Webflow), Steve Garrity (UserClouds), and Clara Shih (Meta AI; previously Salesforce, Hearsay). See funding details.