Editorial and AI policy

How AI is used

AI synthesises salary bands, summaries and outlook language. It does not operate unconstrained: prompts are anchored to imported market signals and research observations, and deterministic validation rejects outputs that break basic plausibility rules.

Where research is enabled, the pipeline searches for public salary sources, fetches the pages, keeps quoted observations with their URLs and classifies source quality. Where evidence is thin, the system says so and widens the range.

Human oversight

Reports are generated automatically. We do not claim that a human editor reviews every figure before publication. We do review reported errors, investigate corrections and improve validation rules when a failure is found.

If you spot a problem, email [email protected] with the report URL and evidence.

Transparency rules

• Every report labels its confidence, data methods and sources.
• Modelled and imputed figures are described as estimates, not official statistics.
• Forecasts are labelled as scenarios and show nominal and real growth.
• Take-home estimates state the tax method and what is excluded.
• Machine-readable Markdown and JSON versions carry the same canonical facts as the HTML page.

Independence

We do not accept payment to change a salary figure, suppress a source or rank an employer. Sponsored placements, if introduced, will be clearly labelled and kept separate from the benchmark.

Corrections and updates

Corrections that change a published figure are reflected in the report refresh date. Source citations are updated with the report. If a correction is material, we may add a note to the report page.

Contact

Editorial and accuracy questions: [email protected]. See also the methodology and disclaimer.