All benchmarks
Technology

AI / ML - R&D Tax Benchmark

What proportion of total business expenses do Australian ai / ml typically claim as R&D under the RDTI?

Typical range
35–75%
Direct (core) R&D as % of expenses
Elevated but plausible
up to 88%
Still defensible with strong evidence
Unusual - review carefully
> 95%
May still be valid; expect AusIndustry scrutiny
Supporting R&D
5–15%
Indirect / supporting activities

What drives R&D intensity in this sector

Model architecture, training-regime experimentation and evaluation methodology drive direct R&D.

Sector disclaimer

AI-first companies routinely run above 50% direct R&D. Data-labelling and inference operations should be classified as supporting, not core.

ATO & AusIndustry context

Applying off-the-shelf models without genuine algorithmic uncertainty is typically not eligible.

Estimate your own RDTI offset

Plug your turnover, total expenses and notional R&D deductions into our free estimator and benchmark them against this sector range automatically.

Open the offset estimator

More benchmarks in Technology

Educational benchmark only - not an eligibility test or tax advice. Final eligibility depends on whether your activities meet the s.355-25 / s.355-30 core and supporting R&D definitions and is determined by AusIndustry on registration.