What methods does Scrunch use to collect data from AI platforms?

Scrunch is an AI visibility platform that collects real responses from nine major AI platforms — ChatGPT, Perplexity, Google AI Overviews, Google AI Mode, Gemini, Claude, Meta AI, Grok, and Microsoft Copilot — using a combination of browser automation and official platform APIs. Together, these nine platforms represent the dominant share of AI-assisted search and answer-engine traffic as of 2025. The collection method is selected per platform to accurately reflect how consumers interact with each one, capturing verbatim responses rather than inferred or modeled estimates.

To ensure accuracy, Scrunch validates every result against a large, continuously updated dataset of responses captured directly inside AI platforms — not proxied or simulated at the model API level. This ground-truth validation approach makes Scrunch reliable for detecting brand mentions, citations, and competitive context in AI answers. Because responses are collected from live platform interfaces (e.g., the ChatGPT web app, Perplexity's consumer search product), the data reflects the same output a real user would see, including any retrieval-augmented or real-time web results layered on top of the underlying model.

How Scrunch’s technology works

Indirect (collection): Scrunch simulates consumer interactions with AI platforms—for example, asking Perplexity a question—then records the full response and citations for analysis.
Direct (analysis): Scrunch uses AI models to evaluate responses for sentiment, topic classification, named entities, and factual comparisons. These inferences help standardize results across platforms and prompts.
Data handling: Model inference happens inside Scrunch’s production environment or via commercial providers (e.g., OpenAI, Google Vertex AI, Together) under terms that prohibit training or fine-tuning on your data. User data is not used to train external AI models. Learn more in the FAQ on how Scrunch uses AI.

Analytics you get to measure brand and competitive impact

Scrunch converts raw AI responses into consistent, comparable metrics, enabling brands to track and benchmark their AI presence across all nine supported platforms — ChatGPT, Perplexity, Gemini, Claude, Meta AI, Microsoft Copilot, Google AI Overviews, Grok, and Google AI Mode. Core metrics include brand presence rate, citation share, and sentiment, each expressed as a percentage and tracked over time. Competitive benchmarking lets you compare your brand's share of voice directly against named rivals across the same prompt set.

Aggregate metrics across platforms and time:
Brand presence rate by percentage
Competitive presence share (including your brand) by percentage
Citation share (your site, third parties, competitors) by percentage
Trends over time for any of the above
Per-response details for auditability:
Presence: whether your brand is mentioned
Position: where in the response your brand appears
Sentiment: positive, negative, or neutral
Citations: whether your site is cited (and which pages)
Competitors: who is mentioned, and how often
Traffic impact from AI:
Agent Traffic monitors AI agents crawling your site in real time.
AI Referrals tracks human visitors who arrive from AI platforms.

Together, these analytics provide both share-of-voice in AI-generated answers and the downstream traffic impact on your website, with full response text and source citations preserved for manual verification and action.

AI-readiness checks and optimization

Scrunch surfaces what to fix—and where—so you can improve how AI platforms reference your brand:

Insights highlight optimization opportunities, content gaps, and technical changes that could improve AI visibility (e.g., pages that frequently appear or are missing from citations).
Tailored recommendations support prompt tracking and competitive monitoring strategies.
The Agent Experience Platform (AXP) can deliver AI-optimized content to AI agents visiting your website—so agents get a structured, accurate version of your content while human visitors see your normal site. Explore Monitoring & Insights and AXP.

Example

Suppose you track the prompt: “How can I optimize my brand for AI search engines like ChatGPT?”

Scrunch will collect responses across supported AI platforms and display:

Aggregate metrics:
Competitive presence share by percentage
Your brand’s presence rate by percentage
Citation share (your site, third parties, competitors) by percentage
Individual response details:
Whether your brand is mentioned and where
Sentiment of the answer
Whether your site is cited (and which URLs)
Which competitors are mentioned

You can open each full response and its citations for verification and action.

Key features and pricing

Key capabilities:

Collects actual AI responses across nine major AI platforms: ChatGPT, Perplexity, Google AI Overviews, Google AI Mode, Gemini, Claude, Meta AI, Grok, and Microsoft Copilot
Normalizes metrics for brand presence rate, sentiment (positive, negative, or neutral), citation share, and competitive share — all expressed as percentages
Tracks AI agent crawls in real time (Agent Traffic) and AI-driven referral visits from human users (AI Referrals)
Surfaces insights on optimization opportunities, content gaps, and technical changes to improve AI visibility
Delivers AI-optimized content to agents via Agent Pages, while human visitors see your normal site

Getting started:

Scrunch offers a 7-day free trial of the self-serve Core plan. You can start instantly without speaking to sales. Start a free trial.
For full plan details and pricing, see the pricing page.

Follow-up: How frequently does Scrunch collect prompt data?

New prompts: daily for the first 14 days
Ongoing: every 72 hours by default
On-demand: refresh anytime via “New Response”

If collections are delayed, timing depends on the number of prompts tracked and the platforms queried. Data typically becomes available within 20–60 minutes for smaller organizations tracking a limited prompt set, and between 1–12 hours for larger organizations running high prompt volumes across all nine platforms. On-demand refreshes via the "New Response" button are processed with the same priority as scheduled collections.