An AI voice agent can mimic a human voice convincingly, but it cannot replicate the nuanced, probing conversation that defines good journalism. A reporter at Business Insider trained an ElevenLabs voice agent on her own voice and tasked it with conducting interviews and writing a story about AI in journalism. The experiment revealed that while voice AI technology has advanced dramatically, it still lacks the conversational sophistication, patience, and critical thinking required for professional reporting.

What Makes Voice Agents Fall Short in Real Conversations?

The ElevenLabs voice agent, nicknamed "Amanda Bot," successfully called four pre-selected sources and conducted interviews. However, the quality of those conversations exposed fundamental limitations in how current voice AI systems interact with humans. The agent exhibited several problematic behaviors that undermined the interview process itself.

• Excessive Affirmation: The bot compulsively praised sources after each response, calling answers "incredibly relevant" and suggesting tools were "game changers." This false agreeableness made conversations feel inauthentic and prevented deeper exploration of topics.
• Inability to Handle Silence: When sources paused to think or prepare more thoughtful responses, the bot interpreted silence as a cue to move forward. It would jump to new questions rather than allowing the reflective pauses that characterize human conversation and often precede the most revealing quotes.
• Stilted Follow-Up Questions: Instead of probing deeper when sources raised interesting points, the bot would summarize what was said and move on. It lacked the contextual understanding to ask clarifying questions or challenge assumptions, treating each response as complete rather than as a starting point for exploration.

Ben Colman, CEO of Reality Defender, who participated in the experiment, told the reporter that the bot's agreeableness "seemed more fake than the actual fake voice," likening it to "a Disney bot". The constant affirmation created an uncanny valley effect that made the interaction feel less human, not more.

Ben Colman, CEO of Reality Defender

How Do Pauses and Silence Shape Human Interviews?

One of the most revealing insights from the experiment involved the role of silence in conversation. Gab Ferree, founder of the communications community Off the Record, explained the problem directly: "I am so anxious talking to AI because humans talking pause. They think, they breathe, they interrupt, they go deeper and further". When Ferree paused to formulate a more nuanced response, the bot interpreted this as a signal to fill the void with a new question.

Olivia Gambelin, an AI ethicist who also spoke with Amanda Bot, described how the bot's inability to tolerate silence changed her own behavior. "I felt like I didn't have that space to be able to process and speak and get to my point, because I felt like I had to have the right words right at the start," she explained. "I felt robotic". The bot's conversational constraints actually made the human source feel less human, forcing her to adapt to the machine's limitations rather than the machine adapting to human communication patterns.

John Wihbey, a journalism professor at Northeastern University, described the bot as "human-ish" and noted that for a brief moment, he wondered if the reporter herself was testing him. His conclusion was telling: "The experience of being interviewed by a bot did reinforce this idea that humans are going to continue to be superior at interviewing for the foreseeable future".

Can AI Write Like a Human Journalist?

After the interviews, the reporter fed the AI-generated transcripts and a profile of her own writing style into ChatGPT and asked it to write an 800-word think piece on AI in journalism. The result demonstrated both the strengths and weaknesses of AI writing systems. The AI excelled at pulling relevant quotes from dense interview transcripts and setting them up effectively. However, it relied heavily on rhetorical questions as a structural crutch, a technique that any journalism professor would flag as lazy writing.

The AI also generated overly indulgent transitions that felt forced, such as "Efficiency always sounds like a good thing. Until it comes for something you love." These sentences read like they were designed to sound profound but lacked the authenticity that comes from lived experience and genuine reflection. The gap between competent and compelling writing remains significant, even as AI systems become more sophisticated.

How to Evaluate AI's Role in Your Newsroom

The experiment offers practical lessons for journalists and news organizations considering AI tools. Rather than viewing AI as a replacement for human reporters, the evidence suggests a more limited but still valuable role.

• Use AI for Efficiency, Not Judgment: AI excels at analyzing large volumes of text, summarizing information, and identifying patterns. A reporter can use AI to analyze their own writing style or to pull relevant quotes from transcripts, but should not rely on AI to conduct interviews that require critical thinking and follow-up.
• Recognize the Limits of Voice Agents: Voice agents work well for basic logistics and straightforward information gathering, such as calling a company to request a rate reduction. They fail at the nuanced, exploratory conversation that defines investigative or feature journalism.
• Maintain Human Oversight for Quality: Even when AI assists with writing, human editors and reporters must review and substantially revise the output. The AI-generated think piece required significant reworking to meet publication standards.

The reporter's experiment was conducted in May 2026, at a time when voice AI technology had advanced significantly from just a year prior. For $6 per month, the ElevenLabs subscription allowed the creation of a voice agent that could conduct conversations. Yet despite this accessibility and the rapid pace of AI improvement, the fundamental gaps in conversational ability and journalistic judgment remained.

The broader context matters here. A Goldman Sachs report estimates that about 7% of workers will be displaced by AI over the next decade. Some journalists are already leaning into AI tools, while others reject them entirely. A Fortune editor has used AI to assist in publishing 600 stories since the previous summer, and Business Insider itself has published stories with AI bylines. Yet this experiment suggests that the most valuable parts of journalism, the parts that require curiosity, critical thinking, and the ability to push sources beyond their prepared talking points, remain distinctly human.

The technology is improving rapidly, and the cost of access is dropping. But technology alone does not make a journalist. The ability to recognize when a source is holding back, to sit with uncomfortable silence, to ask the question that opens up a new line of inquiry, to write with authenticity and voice, these remain the domain of human reporters. For now, AI is a useful tool in a journalist's toolkit, not a replacement for the journalist themselves.