Intro

AI is increasingly becoming a tool for everything, and that includes generating salary data. Consequently I’m seeing a spike in clients asking specifically about AI-generated figures, as candidates and employees are now using these tools to carry out their own research.

I’ve put this page together based on recent reports from AIHR & Ravio and Figures to explain why using AI-generated salary data is currently inappropriate for making compensation decisions.

<aside> 🚨

N.B. This was produced in March 2026. I’ll do my best to keep this page updated as the landscape evolves. While these models may eventually plug into verified data sources, they aren't there yet and rely largely on unverified web snippets.

</aside>

The Challenges with AI Salary Data

Despite the authoritative tone of AI outputs, several structural issues make them an unreliable basis for setting salaries today:

• The American Bias: Most models anchor to US-centric data. In the UK and Europe, this typically leads to a 9% to 23% overestimation of base salaries, as AI fails to account for local tax, pension, and benefit structures.
• Seniority Compression: Research indicates error margins as high as 83% for senior roles. Because executive pay is rarely public, AI "guesses" by flattening the curve between junior and senior positions.
• False Precision: AI generates specific figures to appear helpful. These are probabilistic predictions based on unverified web snippets, not actual payroll records.
• Source Hallucination: It is a documented issue that AI will claim to cite professional databases (such as Mercer or Carta) that it cannot actually access in reality.

For these reasons, I recommend selecting a ’primary data source’ like Ravio, Mercer, Pave, Figures, etc. to run a salary benchmarking exercise and build salary bands. These providers pull anonymised data directly from company payroll and HR systems, which is defensible and auditable.

You can read more about primary data sources here on my guide 2.1 Choosing your data sources.

Example FAQ for your Compensation Philosophy

If it helps, you can add the following to your internal handbook or compensation philosophy to set expectations with employees on why AI sourced data is problematic and unreliable.

Question “I used an AI tool to check my salary benchmark and the numbers are higher than our internal bands. Why is that?”

Answer AI is a language model, not a financial database. It lacks access to the private, subscription-only payroll data we use. Most AI estimates are influenced by US "Big Tech" outliers or unverified job adverts, which do not reflect the specific peer group (industry, size, and location) we benchmark against. We use verified employer-reported data to ensure our ranges are accurate, sustainable, and fair.

Manager Cheat Sheet: Handling AI Benchmarks