Response Data

When Asky monitors AI platforms, each query returns a response. Understanding these responses—their structure, citations, and mention patterns—is key to optimizing your AI search visibility.

How Asky Queries AI Platforms

Unlike most AI visibility platforms that query LLM APIs directly, Asky uses AI agents to simulate real users. This distinction is critical.

Why Not Use APIs?

Many competitors take a shortcut: they query ChatGPT, Claude, or Perplexity through their APIs. The problem? API responses are fundamentally different from what real users see.

Approach	What You Get	Problem
API Queries	Raw model output	No web search (or worse version), different citations, different responses
Asky’s Agents	Actual user experience	Real responses with live search results

API responses are flawed because:

No real-time search: APIs often don’t include the same web search capabilities that users get
Different system prompts: API calls use different instructions than the user interface
Missing citations: Perplexity’s API doesn’t return the same citations as the web interface
Outdated context: APIs may not have the same real-time retrieval as consumer apps
Model differences: The user interface uses a different (secret) system prompt, affecting the replies

If you’re optimizing based on API responses, you’re optimizing for data that doesn’t reflect what real users see. This leads to wasted effort and misleading metrics.

Asky’s Real User Simulation

Asky takes a different approach:

AI agents act as real users: Our agents interact with AI platforms exactly as a human would
Same interface, same results: Queries go through the same web interfaces users access
Real-time search included: When ChatGPT or Perplexity searches the web, we capture those results
Accurate citations: We capture the exact citations users see, not simplified API outputs
True representation: What Asky shows you is what your customers actually see

This means when you optimize based on Asky data, you’re optimizing for real user experiences, not artificial API outputs.

Response Structure

Each captured response contains several data points:

Response Text

The full text of the AI-generated answer. This is exactly what a user would see if they asked the same question.

Timestamp

When the response was captured. Track this to understand how responses evolve over time.

Citations

Citations are the sources AI platforms reference when generating responses. They’re crucial for understanding why certain brands appear in AI answers. See Citations below.

Why Citations Matter

Citations reveal:

What sources AI systems trust: If Forbes is cited frequently, getting Forbes coverage matters
Content that drives mentions: If a competitor’s blog post is cited, analyze (or have Asky analyze for you!) what makes it citable
Your citation presence: Are your pages being cited? If not, why not?
Domain authority signals: Which domains appear most frequently?

Brand Mentions

Asky analyzes each response to detect brand mentions and uses an AI agent to classify the sentiment of these mentions.

Mention Metrics

Metric	Description
Mentioned	Boolean - does your brand appear?
Mention Count	How many times your brand is mentioned
Sentiment	Is this a postive/netural/negative mention?

Sentiment Analysis

Asky classifies mention sentiment:

Sentiment	Indicators
Positive	Recommendations, praise, “best” associations
Neutral	Factual mentions, listings without opinion
Negative	Criticisms, warnings, “avoid” associations

Sentiment score is calculated using smoothed linear dilution.

Analyzing Response Trends

Individual responses are snapshots. Real value comes from tracking trends over time.

What to Track

Mention frequency: Are you mentioned more or less over time?
Sentiment shifts: Is sentiment improving or declining?
Citation patterns: Are new sources emerging?
Competitor changes: Are competitor mentions changing?

Response Changes

AI responses change because:

Model updates: AI providers update their models
Model variance: AI models are non-deterministic, so responses vary
Content changes: New content enters their search indexes
Real-time retrieval: Live web searches return different results
Query interpretation: Same query may be understood differently

Inform Strategy

Response data should inform:

Content topics: What questions should you answer?
Content format: What format gets cited (lists, guides, comparisons)?
Positioning: How are you described vs competitors?
Gaps: What mentions are missing that should be there?

Next Steps

Monitor Crawler Logs to ensure AI bots can access your content