Stop Guessing, Start Knowing Your True AI Brand Mention Rate
A Statistically Rigorous Way to Measure How Often ChatGPT, Gemini and other LLM providers Recommend Your Brand (Based on Anthropic's Research)
What This Calculator Does:
- • Tells you EXACTLY how many prompts to test to achieve your desired confidence level
- • Accounts for LLM "wobble" (the fact that AI gives different answers to the same prompt)
- • Handles clustered prompts (related questions that aren't truly independent)
- • Saves you from wasting resources on too many tests OR making decisions on insufficient data
How to Use This Calculator
Step 1: Define Your Precision Goals
- Precision: How accurate do you need to be? (±2% means if your true rate is 40%, you'll measure between 38-42%)
- Confidence: How sure do you want to be? (95% = industry standard)
Step 2: Set Your Testing Parameters
- Resampling (K): How many times you'll test each prompt (recommended: 3-10)
Step 3: Get Your Magic Number
The calculator will tell you:
- • Exactly how many unique prompts you need
- • Total prompts required to test (prompts × resampling)
- • Estimated testing cost
- • Time to complete
Calculator Settings
80%85%90%95%99%
0.1%1%2.5%4%5%
Percentage points of acceptable error margin
110203050
Number of times to regenerate each question's answer
Formulas
Sample Size Calculation
n = z² × σ² / ME²
n = 1.96² × 0.0367 / 0.0200² = 353
Total Platform Calls
Total = n × K
Total = 353 × 3 = 1,059
Required Sample Size
Unique Prompts Needed
353
prompts
Total LLM Platform Calls
1,059
prompts per platform
With 95% confidence and ±2% margin of error, you need 353 independent prompts with 3 resamples each (1,059 total prompts across all platforms).
Test Across All Major LLM Platforms
To achieve 95% confidence with ±2% accuracy, test 353 prompts × 3 resamples = 1,059 total prompts on each platform below, as each AI has different training data and recommendation patterns.
ChatGPT
Gemini
Claude
PerplexityGrok