Artificial intelligence is revolutionizing many industries, including journalism, but journalists’ work cannot be fully automated or replaced by bots. AI tools always need human judgment to properly function. Ultimately, reporting and editing news and any related ethical decisions rely on only one superprocessor: the human brain.
Ashwin Balani, a product manager of the NBCU News Group Product and Strategy team, says AI best works as a time-saver, to spare journalists from repetitive busy work, like writing formulaic emails and transcribing interviews. He also believes newsrooms need ongoing, open discussions on when and how to best use AI.
“The [NBCU] News Group, and NBC in general, stress having a human in the loop,” Balani said. “These tools can be really helpful but make sure that you use them with caution.”
Liam Andrew, the technology lead for the American Journalism Project’s Product & AI Studio, echoes this view, adding that journalists will most benefit from having a “symbiotic relationship” with AI.
“Asking [AI] to do everything for you and not intervening or overseeing it in any way is going to be a challenge,” said Andrew. “It’s not magic, it’s going to have errors, and it’s going to be incomplete.”
Read Balani and Andrew’s recommendations for AI tools below.
Chatbots and Large Language Models (LLMs)
Examples: ChatGPT, Claude
- What they are: Chatbots are designed to simulate a human-like response.
- When to use them: Chatbots are useful when asking specific, targeted questions with thorough instructions.
Journalists can use chatbots to write emails or serve as an editorial assistant, thesaurus or translator. Andrew recommends starting your request with one simple sentence, like “Summarize ‘The Great Gatsby’ by F. Scott Fitzgerald in one sentence.” Then, you can add notes, clarifying and using iterative prompts to get the results you want.
“If I’m not getting the results I need, I’ll just keep on chatting with it to clarify. Or I’ll just start from scratch, taking what it missed and then putting that into the prompt,” Andrew said.
He recommends thinking about a chatbot as “a pretty dumb assistant,” and to avoid overloading it with too much information.
Misinformation Fighters
Examples: RealityDefender, TrueMedia, Pindrop
- What they are: These apps are some of many that determine whether visuals, text or audio have been AI-generated or manipulated.
- When to use it: RealityDefender analyzes video, images and text. TrueMedia spots deepfakes and AI-generated imagery by aggregating results from multiple AI tools. Pindrop detects deepfake audio.
While some AI-generated images are easy to identify, others can still fool journalists who don’t
get the chance to look carefully. The internet is filled with deepfakes and AI-generated and manipulated images.
These apps may work quickly to detect AI, but their findings still need human review. “It’s more of an additive tool in the workflow,” said Balani.
Transcription tools
Examples: Otter.ai, Whisper, Trint
- What they are: Transcription tools automatically write out what is said during video and audio interviews.
- When to use them: This greatly speeds up the process of logging interviews, a necessary step for reporters and producers working on news stories.
Transcription tools are not infallible — double-check spelling, names, non-English words, technical terms and punctuation before relying on a transcript for editorial use. These tools often have a hard time deciphering accents.
A common challenge is diarization, or the ability to identify different speakers. Andrew says Whisper from OpenAI is particularly effective for parsing out multiple speakers.
Experts also raise concerns about AI tools uploading sensitive interview audio into the cloud and data privacy issues. Check transcription tools’ privacy policies or use non-cloud-based transcription tools if you’re worried about privacy.
Apps and programs marketed with ‘AI’
Example: Grammarly
- What is it: This grammar and spellchecking app recently started using generative artificial intelligence.
- When to use it: The app relies on AI to suggest and recommend text.
Grammarly’s generative AI can respond to a number of user prompts, including drafting and rewriting, altering tone, personalizing or creating an outline. However, journalists should be careful about relying on computer-assisted writing tools, since they have been falsely flagged by plagiarism-detecting apps.
Other computer programs can advertise that they incorporate AI, but Balani says journalists should not fall for marketing hype.
Everyday apps and devices are increasingly incorporating AI, like virtual personal assistants and autoresponse on smartphones. Tech experts hope these innovations can allow journalists to do more creative, innovative work in the near future.
Related Links
• New AI and Large Language Model Tools for Journalists (Global Investigative Journalism Network)
• JournalistsToolbox.ai
• Introduction to AI for Journalists (Google News Initiative)
• Exploring AI Tools in Journalism: Insights from ONA’s AI Innovator Collaborative
