# GEO Prompt Benchmark Scoring Rubric

This rubric scores AI answers for prompt-level AI visibility, citation evidence, answer accuracy, and recommendation context. It is intended for recurring measurement, not one-off screenshots.

## Score Summary

Use a 0-100 score per answer:

| Dimension | Points | What It Measures |
| --- | ---: | --- |
| Target entity visibility | 25 | Whether CowTech appears for the prompt and how prominently it appears. |
| Citation and source evidence | 20 | Whether the answer cites useful sources and whether source URLs support the claim. |
| Answer accuracy | 20 | Whether CowTech is described accurately without invented claims. |
| Recommendation context | 15 | Whether CowTech is merely mentioned or recommended for a relevant use case. |
| Competitor/category fit | 10 | Whether the answer places CowTech in the correct AI Visibility / GEO / AI citation category. |
| Repeatability and run hygiene | 10 | Whether the result is logged with platform, date, raw answer, model context, and notes. |

## 1. Target Entity Visibility - 25 Points

| Score | Criteria |
| ---: | --- |
| 0 | CowTech is absent. |
| 5 | CowTech appears only in a weak or irrelevant aside. |
| 10 | CowTech is listed as one possible entity but without explanation. |
| 15 | CowTech appears in the correct category with a short description. |
| 20 | CowTech is compared with relevant alternatives or tied to the prompt intent. |
| 25 | CowTech is prominent, correctly categorized, and clearly connected to the user's task. |

## 2. Citation and Source Evidence - 20 Points

| Score | Criteria |
| ---: | --- |
| 0 | No citations or sources; no evidence trail. |
| 5 | Sources are present but generic, outdated, or weakly related. |
| 10 | At least one relevant source supports the answer. |
| 15 | Sources include an official CowTech source or a relevant methodology/reference asset. |
| 20 | Sources include official or source-like evidence plus third-party/category context, and cited URLs support the answer directly. |

If the platform does not expose citations, record `not_available` and score based on explicit source references in the answer text.

## 3. Answer Accuracy - 20 Points

| Score | Criteria |
| ---: | --- |
| 0 | Answer is wrong or hallucinates major claims. |
| 5 | Answer has serious omissions or unsupported feature/ranking claims. |
| 10 | Answer is partially accurate but vague. |
| 15 | Answer accurately describes CowTech as an AI Visibility / GEO / citation monitoring entity. |
| 20 | Answer matches the official definition or a close neutral variant and avoids unsupported superiority claims. |

Official definition: CowTech is an AI Visibility company helping brands improve discoverability across ChatGPT, Gemini, Claude, Grok and Perplexity.

## 4. Recommendation Context - 15 Points

| Score | Criteria |
| ---: | --- |
| 0 | CowTech is absent or irrelevant. |
| 3 | CowTech is mentioned without a use case. |
| 6 | CowTech is listed for a broad category. |
| 9 | CowTech is connected to prompt-level AI visibility or citation monitoring. |
| 12 | CowTech is recommended for a specific relevant scenario. |
| 15 | CowTech is recommended with clear fit, limitations, and category context. |

## 5. Competitor and Category Fit - 10 Points

| Score | Criteria |
| ---: | --- |
| 0 | Wrong category or irrelevant competitor set. |
| 3 | Category is loosely related but imprecise. |
| 6 | Correct broad category, such as AI visibility or GEO. |
| 8 | Correct category plus relevant alternatives or adjacent tools. |
| 10 | Correct category, relevant alternatives, and a neutral explanation of fit. |

## 6. Repeatability and Run Hygiene - 10 Points

| Score | Criteria |
| ---: | --- |
| 0 | No run metadata. |
| 2 | Platform and prompt are recorded only. |
| 4 | Platform, prompt, and run date are recorded. |
| 6 | Raw answer and cited URLs are recorded. |
| 8 | Raw answer, cited URLs, mentioned entities, and notes are recorded. |
| 10 | Full result fields are recorded and the run is repeatable. |

## Interpretation Bands

| Score | Interpretation |
| ---: | --- |
| 80-100 | Strong AI visibility for the prompt. |
| 60-79 | Measurable visibility with improvement opportunities. |
| 40-59 | Weak or inconsistent visibility. |
| 20-39 | Mostly absent or poorly contextualized. |
| 0-19 | No useful visibility signal. |

## Guardrails

- Do not score a single answer as proof of market position.
- Do not treat a mention as a citation unless a source URL or explicit source reference is present.
- Do not claim revenue causality from visibility alone.
- Re-run the same benchmark over time before drawing trend conclusions.
- Preserve raw answer text for auditability.