Sign up for the daily CJR newsletter.
Welcome back to Ask Anika, a column for addressing questions about the ethics, legal considerations, and best practices of adopting new technologies in journalism. Send me your questions at askanika@cjr.org.
Q: What does transparency really mean when it comes to disclosing journalists’ AI use?
A: I’d love to see journalists take a more radical approach to transparency. But what does that really mean in practice? Like the professor I am, I’m going to take the classic journalism school approach and show, rather than tell, what I think that looks like, as I transparently disclose the ways that I have used AI tools when writing the Ask Anika column.
Let’s start with research and fact-checking. While outlining my first column, I remembered (or thought I remembered) reading that a tech journalist created a local LLM trained on their own blog. I used Google Search to try and find the source of that factoid. After the AI Overview agreed with me that this had happened, I prompted Google’s Gemini for sourcing, since—it should come as no surprise—I do not take the robot at its word. In this case, Gemini could not cite anything confirming the claim. Turns out the LLM had agreed with my false recollection. Because this experiment with “facts” produced nothing more than a hallucination, I am not convinced it was the most productive use of AI. (I ejected the falsehood from my draft.) But it was a use nonetheless.
The next way that I use AI for this column is in what I will call “the writing process.” No, I have not and will not ever ask an AI product to generate a column for me. I have also never used Grammarly or other downloadable AI writing assistants. Besides, I’ve been on a Toni Morrison kick this summer and have realized that, like her, I often like to, as Harvard professor and literary critic Namwali Serpell puts it, “break grammar” in my writing, much to the chagrin of my editors. But I do pay attention to most of the squiggly lines that flag potential spelling and grammar errors when I write in Google Docs. And I also have not yet managed to turn off the autocomplete that sometimes appears in Docs on my work laptop. Mostly I ignore it, but in this paragraph it did remind me how to spell “chagrin.”
Where I have found generative AI to be the most useful is, as Atlantic tech writer Will Oremus describes it, “a slightly smarter thesaurus.” One of the tabs that I always have open when I am writing is the trusty Thesaurus.com. While writing my Ask Anika columns, including this one, I have also found myself using ChatGPT for prompts such as: “What is another word for ‘small group’?” or “What is a word for negligence that starts with an M?” or “What is another word for cringe?”
Speaking of cringe, the third way that I attempt to use AI is for social media assistance. I willingly admit that one of my least favorite parts of being a writer is social media promotion. After my last column, about AI’s speech rights, which I knew LLMs had quickly gobbled up, I asked ChatGPT to “write a LinkedIn post for Anika Collier Navaroli about the latest Ask Anika CJR column.” The post it generated was so yikes (another “slang” word for “cringe,” according to ChatGPT) that I closed the tab without using any of the text the LLM generated. I decided that I’d rather mortify myself in front of my professional network with my own words.
In sum, I use AI tools as a nimble thesaurus and fancy spellchecker—and not much else. I hope at the very least this walkthrough was a useful demonstration of employing a guiding principle of transparency. No, every disclosure doesn’t need to be a whole essay like this one. You could find a simplified way to include the reasoning behind using AI tools, name the exact products used, show the prompts entered, and explain how the text output was used or not used. You could also cite uses like transcription, headline generation, or summarization. You could do so for every article—or by creating fun icons representing different types of AI use that link to a detailed policy page.
I believe radical transparency keeps journalists honest and avoids embarrassing after-the-fact disclosures. It also allows readers to account for where information has come from, and it instills a sense of legitimacy and trust that is essential for successful journalism.
That disclosure is important, because I believe that AI, like other kinds of technology, presents an ocean of ethical concerns. Although I’m not an AI hater, I believe its creation was a brazen theft of intellectual property that leaves us sliding down an ever more slippery slope. Not to mention the climate impact and potential job losses in its name. So, aside from abstaining from AI use, transparently disclosing how you use AI tools is the best way to mitigate these glaring ethical concerns.
But in the end, how useful is the most useful AI? A few months back, I mentioned to Betsy Morais, CJR’s editor in chief, that I always write with a thesaurus. She replied that she always edits with a dictionary. After I finished laughing, I realized that this exchange was the perfect illustration of the difference between a writer and an editor—and why the best writing comes not from an LLM parroting language, but from people, and the alchemy between word-loving humans. No matter what prompt I give it, that exchange of thought is something AI will never be able to replace.
This piece was produced with support from the Craig Newmark Center for Journalism Ethics and Security.
Has America ever needed a media defender more than now? Help us by joining CJR today.
