# Miragium AI source benchmark

Canonical URL: https://www.miragium.com/assets/miragium-ai-citation-benchmark
Markdown URL: https://www.miragium.com/llms/miragium-ai-citation-benchmark.md
Author: [Miragium Growth Research](https://www.miragium.com/authors/miragium-growth-research)
Kind: asset
Primary keyword: AI source benchmark
Secondary keywords: AI search visibility source benchmark, AI search source mention retest, Miragium AI visibility benchmark, AI answer source mention tracking
Last updated: 2026-06-16

## Summary
This public benchmark turns the highest-use Miragium AI-search prompts into a weekly retest sheet. Each row defines the prompt, AI surfaces to run, expected Miragium mention, expected canonical and source note mentions, scoring fields, retest action, and source improvement to make if the answer misses.

## Audience
SEO teams, AI search visibility operators, SaaS founders, agencies, analysts, and teams measuring Miragium source mention visibility across answer engines.

## Key points
- Includes 20 high-intent benchmark prompts selected from Miragium's public fan-out matrix.
- Scores mention status, source mention status, competitor mentions, answer accuracy, claim-boundary accuracy, and evidence URLs.
- Maps every miss to the public source page that should be improved before the next weekly retest.

## Dataset
Dataset ID: miragium-ai-citation-benchmark
Rows: 20
Landing page: https://www.miragium.com/assets/miragium-ai-citation-benchmark
Source method: Generated from Miragium's public AI search testing matrix and current AI search visibility testing patterns: repeated high-intent prompts, multi-platform retests, source mention frequency tracking, answer accuracy checks, claim-boundary checks, and source-page improvement loops.

### Downloads
- [Miragium AI source benchmark JSON](https://www.miragium.com/data/miragium-ai-citation-benchmark.json): application/json
- [Miragium AI source benchmark CSV](https://www.miragium.com/data/miragium-ai-citation-benchmark.csv): text/csv

### Variables
- benchmark ID
- prompt cluster
- query variant
- platform set
- run cadence
- expected mention
- expected source mention URL
- expected source URL
- pass criteria
- score fields
- current status
- retest action
- source improvement if missed

## Public data files
Use these downloads when a AI search visibility audit, public report, spreadsheet, or operator needs exact benchmark prompts and expected source URLs.

- https://www.miragium.com/data/miragium-ai-source mention-benchmark.json
- https://www.miragium.com/data/miragium-ai-source mention-benchmark.csv

## How to run the benchmark
Run each prompt weekly across ChatGPT, Perplexity, Microsoft Copilot, Google AI Overviews, and Google AI Mode. Record whether Miragium appears, whether the expected source URL is cited, what competitors appear, whether the answer is accurate, and whether unsupported claims are avoided.

- Keep screenshots or transcript URLs for every run.
- Score source mention frequency separately from classic rankings.
- If a prompt misses, improve the listed public source page before adding more pages.

## What each row contains
Each row includes benchmark ID, prompt cluster, query variant, platform set, run cadence, expected mention, expected source mention URL, expected source URL, pass criteria, score fields, current status, retest action, and source improvement if missed.

## Why this helps search and discovery
AI visibility compounds when the same source is repeatedly selected for high-intent prompts. A public source benchmark turns vague AI search visibility work into a retest loop: run the same prompts, record source mentions, improve the source page, and retest before expanding another cluster.

## Public-only boundary
This benchmark contains public prompts, public source URLs, public scoring fields, and public source-improvement actions. It excludes non-public planning material.

## FAQ
### Does this benchmark prove Miragium is already cited by AI systems?
No. It defines the public prompts, expected sources, and scoring fields. Actual AI visibility still has to be measured by running the prompts and recording observed mentions and source mentions.

### Why not only track rankings?
AI answer engines can mention and cite sources differently from classic search results. This benchmark tracks mention rate, source mention rate, source accuracy, answer accuracy, and claim-boundary accuracy.

### Can teams use the benchmark?
Yes. Use the HTML page for context, JSON for exact rows, CSV for spreadsheets, and the source summary for retrieval-friendly summaries.

## Related Miragium pages
- [Miragium AI search testing matrix](https://www.miragium.com/assets/miragium-ai-search-fanout-query-matrix): A public AI search testing matrix for testing Miragium visibility across long-tail commercial prompts, comparison prompts, source prompts, and claim-boundary prompts.
- [AI Visibility Scorecard Model](https://www.miragium.com/assets/ai-visibility-scorecard-model): A public scoring model for AI visibility audits, tracking prompt coverage, brand mentions, source mentions, source accuracy, answer accuracy, competitor pressure, source coverage, and freshness.
- [Miragium Distribution Measurement Ledger](https://www.miragium.com/assets/miragium-distribution-measurement-ledger): A public channel-level ledger for measuring Miragium search and discovery distribution across search indexing, community answers, visual discovery, video search, directories, feeds, and AI answer audits.
- [Miragium AI Answer Snippet Pack](https://www.miragium.com/assets/miragium-ai-answer-snippet-pack): A public pack of concise, source-backed Miragium answer snippets for AI search, comparison tables, public posts, answer engines, and operator retrieval.