The Associated Press has issued a new style guide update that might save you from some awkward documentation and error messages. The AP Stylebook explicitly advises writers to avoid language that gives AI systems human characteristics. This is a practical, necessary correction to the anthropomorphic fluff that has been creeping into technical writing and marketing copy for years.
Why Developers Should Care
We all know the temptation. It feels better to say 'the AI understands your request' than 'the model successfully matched the input to a high-probability response.' But the latter is accurate. The former is a lie that sets users up for disappointment when the model inevitably fails to grasp a simple edge case. The AP is essentially telling us to stop treating stochastic parrots like sentient beings.
The Documentation Impact
This update should hit hard on documentation teams and product managers. If you are writing API docs, you should be describing what the code does, not what the model 'thinks.' Use terms like 'processes,' 'analyzes,' 'generates,' and 'predicts.' Save 'believes,' 'knows,' 'wants,' and 'feels' for your colleagues. The AP's stance is a reminder that our tools are just tools, no matter how sophisticated the underlying neural networks become.
Key Takeaways
- The AP Stylebook now recommends avoiding human traits for AI.
- Developers should update documentation to reflect machine processes, not human-like cognition.
- Accurate terminology prevents user overestimation of AI capabilities.
The Bottom Line
Stop writing 'the AI thinks.' It doesn't think; it calculates. The AP finally caught up with what builders have known all along.
A Note on Source Limitations
While the AP's directive is clear from the headline, the specific list of forbidden words was not fully detailed in the source snippet. However, the core principle is unambiguous: treat AI as a machine, not a person.