Use of AI

October 2026

AI software can’t do shoe-leather journalism, write award-winning essays or interview members of parliament over a pint.

We are experimenting with ways different software can assist us in tasks such as summarising transcripts or data, but our default position is that any generative AI output requires human oversight and constant, rigorous checking.

The words you read on The Spinoff have been written by Spinoff writers (and sometimes, readers).

We do not and will not publish work that has used generative AI to produce the words you read on The Spinoff. There are a couple of exceptions:

• We utilise an AI tool to assist us in generating alternative text for images. These must be reviewed by a human before publishing.
• Parody. There is an obvious exception when AI-generated images or GenAI in writing are used very obviously as a parody of AI-generated images/GenAI writing. Images created using these tools will be acknowledged in the caption, and the use of GenAI writing or parody purposes will be very apparent and noted.

Other uses:

• We use design software that utilises a form of basic AI to perform simple tasks, such as background removal on images. We do not use generate imagery via prompts in AI image generators. We do not use AI image tools to generate or edit human figures, create hyper-realistic scenarios or manipulate news photography. Best efforts and care are taken to avoid using imagery generated via prompt-based AI tools when using media from image libraries.
• Our staff can use generative AI to brainstorm, analyse and summarise publicly available documents and datasets. Outputs, such as facts, data analysis, or quotes, are reviewed by writers and editors. It is well-documented that generative models of large language models, such as Google Gemini, Claude, and Chat GPT, make many mistakes.