The Evidence Behind AI-Supported Writing Assessment

When school leaders ask us about AI assessment, they’re usually asking the same three questions:

These are fair questions and they deserve a straight answer grounded in evidence, not reassurance.

So we worked with Ben Lawless, assessment researcher, University of Melbourne lecturer and independent consultant focusing on learning progressions, to look honestly at what the research actually says.

Ben and our Head of Education, Sarah Bakker (one of Australia’s Most Influential Educators 2026), have written about it for Education Matters.

The article takes on the three concerns directly, including one confronting stat: research shows human markers only agree with each other around 50% of the time, while the best AI rubric-marking systems are achieving above 80% agreement (King et al., 2001; Koretz et al., 1992; Li et al., 2025).

Ben and Sarah also look at the reality of current marking practice (the findings are more confronting than most people expect!).

Big picture: it’s critical that we consider AI-supported assessment from the perspective of what evidence shows us, not fear. 

If you’re a school leader trying to get ahead of this conversation, or just trying to answer the questions your staff are already asking, it’s definitely worth a read: 

Ben is also hosting a free webinar in partnership with Seven Steps: AI & Assessment: A Pragmatic Case for Schools.

Find upcoming dates for this and other free events here: Free Literacy Events