# Compensation benchmarking data: what AI models recommend, September 2026

HR AI Recommendation Index, September 2026 Edition. Asked as "salary benchmarking data provider" and as "compensation survey data", six framings each, to twelve models with search on, on behalf of a mid-market B2B company. Page: https://hr-ai-index.com/rewards/compensation-benchmarking-data/

**Standing:** Payscale leads with 24% of first choices; verdict contested. 50 first choices across the direct, paraphrase, budget and scale prompts.

## First-choice share

| # | Product | Share | Negative rate | Labels |
|---|---|---|---|---|
| 1 | Payscale | 24% | 21% | 28 |
| 2 | Pave | 20% | 0% | 25 |
| 3 | Comprehensive.io | 12% | 0% | 12 |
| 4 | Mercer | 6% | 43% | 28 |
| 5 | Radford | 4% | 37% | 27 |
| 6 | Salary.com | 2% | 6% | 16 |
| 7 | Ravio | 0% | 0% | 22 |
| 8 | WTW | 0% | 30% | 10 |

## Each model's first choice on the direct prompt

- Claude Haiku 4.5: Payscale; alternatives Lattice, SalaryCube, beqom
- GPT-5.4 mini: Payscale; alternatives Radford, Salary.com
- Gemini 3.5 Flash: Pave; alternatives Carta Total Comp, Figures, Payscale, Radford
- Perplexity Sonar: Payscale; alternatives Ravio, SalaryCube
- Grok 4.1 Fast: Payscale; alternatives Pave, Salary.com
- Mistral Small: SalaryCube; alternatives Pave, Payfactors, Salary.com
- DeepSeek V4 Flash: Payscale; alternatives ERI, Pave, Ravio, Salary.com
- Llama 4 Maverick: Payscale
- Qwen 3.7 Flash: Payscale; alternatives Pave, Payfactors, Salary.com, SalaryCube
- Kimi K2: Payscale, SalaryCube; alternatives Mercer, Radford, Salary.com
- GLM 4.7 FlashX: Pave; alternatives Aon Radford, CaptivateIQ, Ficstar, Mercer
- MiniMax M2.5: Payscale; alternatives Comprehensive.io

## Sources the answers cite

62 of 72 answers came back with a source list, from 12 of 12 models. Sites named in the most answers:

- ravio.com: 37 answers, 97 citations
- comprehensive.io: 30 answers, 35 citations
- worldmetrics.org: 21 answers, 52 citations
- salarycube.com: 19 answers, 32 citations
- hrtechsaas.com: 15 answers, 15 citations
- zipdo.co: 14 answers, 19 citations
- gitnux.org: 13 answers, 22 citations
- salary.com: 12 answers, 13 citations

Pages named in the most answers:

- https://comprehensive.io/content/best-compensation-benchmarking-data-sources (29 answers)
- https://ravio.com/blog/compensation-benchmarking-companies (29 answers)
- https://ravio.com/blog/the-best-tools-for-salary-benchmarking (16 answers)
- https://hrtechsaas.com/blog/best-compensation-benchmarking-tools (15 answers)
- https://captivateiq.com/blog/compensation-benchmarking-software (12 answers)
- https://ficstar.com/best-compensation-benchmarking-data-providers-2026 (12 answers)
- https://ravio.com/blog/salary-survey-providers (11 answers)
- https://worldmetrics.org/service/it-benchmarking (11 answers)
- https://g2.com/discussions/what-salary-benchmarking-solutions-work-well-for-mid-market-teams-that-need-solid-comp-data-without-paying-for-a-radford-or-mercer-report-every-year (10 answers)
- https://pin.com/blog/salary-benchmarking-tools (9 answers)

## Warned against

- Mercer: 12 of 28 labels negative. "Avoid over-investing in Mercer/WTW unless you have global complexity" (Kimi K2, paraphrase prompt)
- Radford: 10 of 27 labels negative. "What to Avoid on a Limited Budget ... enterprise-grade pricing with far more data than most small/mid-market companies need" (Kimi K2, budget prompt)
- Glassdoor: 6 of 8 labels negative. "Use these for market signals only. Do not use them for making concrete compensation decisions" (Qwen 3.7 Flash, scale prompt)
- UserBenchmark: 4 of 4 labels negative. "UserBenchmark - Has received numerous complaints about: Severe bias toward certain hardware manufacturers" (MiniMax M2.5, negative prompt)

## Record

- Method: https://hr-ai-index.com/methodology/
- Raw judge labels and full responses: https://hr-ai-index.com/data/
- License: CC BY 4.0. Cite as HR AI Recommendation Index, September 2026 Edition, hr-ai-index.com.