Technical Documentation

Engine documentation and statistical reference

Explore how AmpliRank samples probabilistic AI answers, computes 95% Wilson confidence intervals, and exposes workspace intelligence via REST APIs.

Getting Started

Quickstart & First Scan

Set up your brand workspace, configure buyer queries, and capture your first sample baseline in minutes.

1. Workspace & Brand Setup

AmpliRank models brand intelligence around multi-tenant workspaces. Within each workspace, you define your primary brand entity and your direct competitors. Tracked brands and competitors share the plan's brand allowance, and each entity requires an official brand name, canonical domain, and primary industry vertical.

2. Defining Buyer Question Sets

AI answer engines do not rank keywords; they answer natural language questions asked by buyers in research mode. Configure prompt templates across three core buyer intents: Category Exploration ('What are the top solutions for...?'), Head-to-Head Comparison ('Brand A vs Brand B'), and Technical Due Diligence ('Does Brand A support SSO and SOC2?'). Prompt sets are managed in the dashboard; the REST API runs prompts you have already created.

3. Executing Your First Scan

Select the providers configured for your workspace (up to your plan's provider limit) and trigger a scan. AmpliRank queries stateless provider sessions, parses the response text for brand mentions and sentiment, and extracts cited URLs and root domains found in the answer.

bashAmpliRank API v1

# Trigger a sampled scan through the REST API
curl -X POST https://amplirank.com/api/v1/polling \
  -H "Authorization: Bearer ampr_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "brandId": "YOUR_BRAND_UUID",
    "promptId": "YOUR_PROMPT_UUID",
    "providers": ["chatgpt", "claude"]
  }'

Methodology

Statistical Sampling & Confidence Intervals

Why single-query testing fails and how AmpliRank computes Wilson score intervals for defensible executive reporting.

The Non-Determinism Problem

Large language models operate probabilistically with non-zero temperature, continuous weight updates, and dynamic search retrieval augmented generation (RAG). Testing a single prompt once in a web browser produces anecdotal noise, not actionable marketing data.

Wilson Score Confidence Intervals

For every observed brand recommendation frequency, AmpliRank computes a 95% Wilson score confidence interval. If your brand is recommended in 14 out of 20 samples (70%), the 95% interval is [48.1%, 85.5%]. We report both the point estimate and the interval bounds so executive teams make decisions on statistical evidence rather than false certainty.

API samples are not a complete view of every answer to every user. Every metric represents one answer to one question from one provider at a specific point in time.

Recommended Sample Sizes

For weekly operational tracking, we recommend 10 samples per critical prompt per model (margin of error ~±15%). For monthly executive briefs or board reporting, we recommend 30 to 50 samples per prompt (margin of error ~±8%).

API Reference

Workspace REST API & Webhooks

Integrate AI brand intelligence directly into your data warehouse, BI dashboards, and internal workflows.

Authentication

All REST API endpoints require a workspace-scoped API key passed via the standard Authorization Bearer header. Keys use the ampr_live_ prefix, are stored as SHA-256 hashes, and are generated and revoked within Settings -> API Keys.

bashAmpliRank API v1

# List workspace brands
curl https://amplirank.com/api/v1/brands \
  -H "Authorization: Bearer ampr_live_..."

# Read the intelligence overview for one brand
curl https://amplirank.com/api/v1/brands/YOUR_BRAND_UUID/intelligence \
  -H "Authorization: Bearer ampr_live_..."

Endpoints

Every endpoint is scoped to the workspace that owns the API key, so a key can only read and write data in its own workspace.

httpAmpliRank API v1

GET    /api/v1/brands
POST   /api/v1/brands
GET    /api/v1/brands/{brandId}
PATCH  /api/v1/brands/{brandId}
DELETE /api/v1/brands/{brandId}
GET    /api/v1/brands/{brandId}/score
GET    /api/v1/brands/{brandId}/forecast
GET    /api/v1/brands/{brandId}/intelligence
GET    /api/v1/brands/{brandId}/diagnoses
POST   /api/v1/polling
POST   /api/v1/track
POST   /api/v1/webhooks

Rate Limits

Rate limiting is applied per workspace with a Cloudflare Durable Object. The currently limited endpoints are webhook dispatch (100 requests per minute) and attribution tracking (500 requests per minute). Limited responses include X-RateLimit-Limit, X-RateLimit-Remaining, and X-RateLimit-Reset headers, and return 429 with Retry-After when the limit is exceeded.

Webhook Delivery

Outbound webhooks are caller-triggered today: POST an event name (128 characters or fewer) with an optional brandId and data object to /api/v1/webhooks, and AmpliRank delivers it to enabled alert rules that match. Deliveries are signed with HMAC-SHA256 and carry X-Amplirank-Signature, X-Amplirank-Timestamp, X-Amplirank-Event, and X-Amplirank-Workspace headers. Scan-completion and citation events are not emitted automatically yet; webhook dispatch is included on Agency and Enterprise plans.

Agent Standards

llms.txt & Source Fact Publishing

Publish structured machine-readable facts and documentation to guide autonomous AI search agents and retrieval spiders.

What is llms.txt?

llms.txt is an emerging web standard that provides a curated, lightweight Markdown manifest of your company's core value proposition, product architecture, documentation, and pricing facts specifically formatted for LLM context windows and search crawlers.

Evidence Verification Before Publishing

AmpliRank enforces strict evidence linkage: you cannot publish fact claims to your live llms.txt feed unless backed by verifiable public documentation or official product specifications.

Ready to test against live AI answers?

Create a workspace to test prompt sets, inspect citations, and review statistical baselines.

Create workspace Read statistical methodology →