Public trust in journalism will require clear disclosure when newsrooms use text-generating AI tools like GPT-3. At first, outlets that use AI tools will be seen by the public as a binary: Either this is a publication that publishes AI-generated news articles, or it isn’t. We’ll also see a debate on which topics make the use of AI tools unacceptable for reporting, versus those where an AI tool might be the best first source for rapid coverage.

These debates exist within certain beats already, but the evolving sophistication of how the public understands issues like media manipulation will eventually inspire demand for transparency in who or what wrote the news.

New safeguards and content moderation practices will be developed to account for the vulnerabilities and terrifying abusability of AI text generators. Editorial practices will evolve to weigh human contributions against that of an AI system, balancing issues such as accuracy, speed, reliability, data sources, and name brand. And after an “AI-content-bad/human-content-good” morality play, those newsrooms resourced to explore these complex tools will recognize and begin to adapt to their pitfalls and tradeoffs.

Newsroom leadership will particularly need to rapidly learn how AI is integrated with human creativity, labor, and existing norms, a set of practices known as “repairing innovation.”

We may see a kind of simple report card or badging system on news articles indicating the degree or role that AI and humans played in creation. Crucially, both newsrooms and news consumers will come to understand an “accountability shorthand” around the use of AI text generators, a new layer in the formulation of trust — or distrust — in media by the general public that, over time, newsrooms will be obligated to communicate to their readers.