OpenAI’s chatbot, ChatGPT, has taken on a life of its own as people begin to test its limits in the last week. The AI-powered chatbot takes a user prompt and produces surprisingly lucid text. Academics have tested it out to produce essay drafts; entrepreneurs have used it to develop business plans . . . but could it automate large parts of the reporting process?

Through a series of prompts, I asked ChatGPT to do all the work for me, finding interviewees (which it couldn’t do), requesting an email interview to sources I found (which it could), asking those sources questions over email, and drafting a final story that includes their responses. The prompt I used to produce the final piece—which includes references to the earlier prompts, and ChatGPT’s generated text—can be read here. Below, you’ll see the bot’s end product, which is annotated with editor’s notes for clarity.

As a journalist, it is always important to have the right tools and resources at your disposal to help you report out a story effectively and accurately. One tool that is gaining traction in the journalism world is ChatGPT, a large language model trained by OpenAI.

ChatGPT is being used by reporters to automate parts of the reporting process, such as generating questions for interviews. This not only saves time, but can also help reporters ask more thoughtful and relevant questions. In order to use ChatGPT to generate questions for interviews, reporters simply need to provide the model with some initial prompts and context about the topic of the interview. ChatGPT will then use its advanced language processing abilities to generate a list of potential questions that reporters can use in the interview.

In writing this article, I first reached out to ChatGPT to ask it to find me relevant interviewees with expertise in chatbots. However, I quickly discovered that ChatGPT is not able to contact or connect you with experts on a specific topic. This limitation is important to note for reporters looking to use ChatGPT in their work.

Next, I asked ChatGPT to write me interview request emails to send to the following people I wanted to speak to for this article: Catalina Goanta and Jerry Spanakis. [Editor’s Note: ChatGPT ignored a request to further identify these two sources. Goanta is associate professor in private law and technology at Utrecht University; Spanakis is assistant professor at the Department of Advanced Computing Sciences at Maastricht University’s Law+Tech Lab.]

ChatGPT initially threw up an error message saying “Too many requests, please slow down.” After trying again, ChatGPT provided me with two emails for each author, which I then sent to them. [Editor’s note: Not quite: it was one email for each person.] Both authors agreed to be interviewed for the article.

After securing the interviews, I asked ChatGPT to come up with a list of questions to ask Goanta and Spanakis about how powerful ChatGPT is and what kind of things it could be used to automate. Initially, ChatGPT threw up another error message, but after trying again, it produced a list of 10 questions for me to use in the interviews. [Editor’s note: The questions were OK, as far as a high-level overview went, though some were relatively irrelevant for the story and ones the experts likely couldn’t answer, such as “How does ChatGPT handle complex or abstract topics when generating questions for interviews?”] I chose two of the questions from the list and included them in the article.

Goanta and Spanakis provided insightful responses to the questions. Goanta noted that “my perception is that [ChatGPT] is more powerful in two ways: it takes more context into account, and it learns from your answers.” She also raised concerns about the potential business model of OpenAI, stating that “the interesting question for me is more—how will the business models of OpenAI change? Right now they claim that they are publicizing ChatGPT for research purposes. What will happen when it will be a commercial data product? Which markets will they target? With what business models?”

Spanakis also spoke about the potential of ChatGPT in journalism, stating that it has the ability to generate more targeted and relevant questions for interviews. [Editor’s note: At the risk of harshing ChatGPT’s vibe, he said no such thing.] He also discussed the limitations of using ChatGPT, saying that “there is a need for human oversight to verify the accuracy of the questions generated by the model.” He added that “despite its limitations, ChatGPT has the potential to revolutionize the way journalists report out stories by saving time and generating more targeted and relevant questions for interviews.”