# Are general and best-of prompts more volatile than comparisons? | Trakkr Research

Canonical URL: https://trakkr.ai/trakkr-research/model-divergence/answers/are-general-and-best-of-prompts-more-volatile-than-comparisons
Published: 2026-03-11
Last updated: 2026-03-11
Author: Mack Grenfell

Yes. Comparison prompts average 50.4% agreement, while general prompts average 42.2% and best-of prompts carry a 14.8% high-divergence rate.

## Methodology

Built from 797,644 valid comparisons across 44,088 reports and 8 models, covering 6,439,133 model responses in the observed window.

## Direct Answer

Yes. Comparison prompts average 50.4% agreement, while general prompts average 42.2% and best-of prompts carry a 14.8% high-divergence rate.

## What this means

Operators must allocate resources differently based on query type, as high-divergence categories require multi-model optimization rather than single-platform focus.

## Evidence table

| Metric | Value | Why it matters |
| --- | --- | --- |
| Comparison-query agreement | 50.4% | Comparison prompts produce the highest average agreement. |
| General-query agreement | 42.2% | General prompts are less stable across models. |
| Best-of high divergence | 14.8% | Best-of prompts frequently split models. |

## Frequently Asked Questions

### Which prompt type produces the highest agreement across models?

Comparison prompts produce the highest average agreement at 50.4%.

### How often do best-of prompts cause models to split?

Best-of prompts carry a 14.8% high-divergence rate, frequently splitting models.

## What to do next

- [Track visibility across multiple models instead of using one platform as a proxy for the whole market.](https://trakkr.ai/trakkr-research/model-divergence/answers/are-general-and-best-of-prompts-more-volatile-than-comparisons#next-step-1)
- [Prioritize query classes where disagreement is highest because that is where share can move fastest.](https://trakkr.ai/trakkr-research/model-divergence/answers/are-general-and-best-of-prompts-more-volatile-than-comparisons#next-step-2)
- [Treat consensus as a benchmark, but treat divergence as the operating reality.](https://trakkr.ai/trakkr-research/model-divergence/answers/are-general-and-best-of-prompts-more-volatile-than-comparisons#next-step-3)

## Related pages

Continue through the same study cluster.

- [what does an average top three overlap of two point eight mean](https://trakkr.ai/trakkr-research/model-divergence/answers/what-does-an-average-top-three-overlap-of-two-point-eight-mean) - Related answer page
- [should you use one model as a proxy for all ai visibility](https://trakkr.ai/trakkr-research/model-divergence/answers/should-you-use-one-model-as-a-proxy-for-all-ai-visibility) - Related answer page
- [comparison prompts are the most stable query class](https://trakkr.ai/trakkr-research/model-divergence/facts/comparison-prompts-are-the-most-stable-query-class) - Related fact page
- [query class agreement tracker](https://trakkr.ai/trakkr-research/model-divergence/trackers/query-class-agreement-tracker) - Related tracker page

## Data And Sources

- [Same Question, Different AI, Different Answers](https://trakkr.ai/trakkr-research/model-divergence) - Flagship study behind this page
- [Page JSON](https://trakkr.ai/data/research-answers/model-divergence/answers/are-general-and-best-of-prompts-more-volatile-than-comparisons.json) - Machine-readable companion file