CiteRank AI
Methodology

How we score AI visibility.

Open methodology. Every score CiteRank AI ships is reproducible, with the prompt graph, replay count and confidence interval surfaced in-product.

  1. Step 01

    Prompt graph generation

    We build a category-specific prompt graph from your industry, sub-vertical, geography and competitor set. Each prompt is tagged by intent — discovery, comparison, evaluation, conversion — so the scoring weights match buying-stage value.

  2. Step 02

    Multi-LLM execution (10–30 replays per prompt)

    Every prompt is replayed 10–30 times per engine to control for model temperature and answer drift. We run against 8 engines: ChatGPT, Gemini, Claude, Perplexity, Google AI Overviews, Microsoft Copilot, xAI Grok, and Meta AI.

  3. Step 03

    Citation detection & entity matching

    Responses are parsed for explicit citations (URLs, source cards), inline brand mentions, and entity-level references. We deduplicate near-mentions and resolve corporate aliases so 'CiteRank', 'CiteRank AI' and the .in domain all match a single entity.

  4. Step 04

    Composite visibility score

    Citation share, recommendation order, sentiment and answer volatility are combined into a single 0–100 visibility score per engine, then weighted by intent value across the prompt graph. We report the score with a 95% confidence band so single-run noise doesn't drive decisions.

  5. Step 05

    Competitor benchmarking & gap-fix briefs

    Every brand score has matched competitor scores. Where you lose, we ship a gap-fix brief — the specific prompts you don't appear on, the sources cited instead, and the entity, schema or content changes that would close the gap.

  6. Step 06

    Re-runs & attribution

    Weekly automated re-runs (daily on Growth+) track movement. Optional GA4, Search Console and HubSpot connections tie AI visibility shifts to downstream sessions, conversions and pipeline.

What we don't do
  • · No single-shot prompts. Every score is averaged across 10–30 replays per engine.
  • · No black-box scoring. The composite score formula and weights are published in-product.
  • · No fabricated samples. Marketing dashboards labelled "Illustrative — sample data" use synthetic numbers; customer reports use only that customer's real prompt runs.