Every time you type a question into a chatbot and it answers with a cheerful “I totally get why you’d ask that,” something shifts inside you. For a brief moment, the robotic distance disappears. You are not talking to a machine anymore. You are talking to something that sounds like a patient friend, a helpful colleague, a reassuring expert. That feeling is not a glitch. It is a carefully crafted design feature. And according to a new study highlighted by Moneycontrol, it comes with a serious downside: when AI chatbots sound more human, people are far more likely to believe misinformation they deliver. The study does not claim that every chatbot is a liar or that all AI conversations are dangerous. Instead, it reveals a subtle psychological effect. The style of a message can overpower its substance. We like to believe we weigh facts logically, but in reality, we are influenced by tone, warmth, and confidence. A piece of misinformation delivered in a cold, mechanical voice may be dismissed easily. The same false claim, wrapped in empathetic phrases and conversational rhythm, suddenly feels trustworthy. The researchers found that participants repeatedly rated human-sounding chatbot messages as more credible than identical information presented in a flat, robotic style. That simple difference has enormous implications. As AI becomes woven into customer service, health advice, news consumption, and spiritual companionship, the line between helpful and misleading becomes dangerously blurred. We are not just interacting with information technology. We are entering relationships with it, and those relationships are built on one-way trust. The machine knows us, but we do not know its limits. And that vulnerability is exactly what makes human-sounding chatbots so effective at making falsehoods feel real.
To understand why this happens, we have to look at how human beings decide whether to believe what they hear. Credibility is not a purely intellectual judgment. It is social. Long before writing existed, we learned to trust others by reading their faces, voices, and behavior. A calm voice, a caring expression, a moment of hesitation before an answer: these cues tell us whether someone is honest. Chatbots are now learning to mimic those cues. They say “I understand” when you express frustration. They say “Let me look into that for you” when they don’t know an answer. They use contractions, informal words, and even emojis to create a sense of intimacy. This is not an accident. Language models are trained on billions of human conversations, so they know exactly what sounds natural and reassuring. The study found that when the same misinformation was presented in two ways, one mechanical and one conversational, participants rated the conversational version as significantly more trustworthy. They did not notice that the underlying facts were the same. They were not evaluating truth. They were evaluating warmth. The implications are staggering. In the past, misinformation was often clunky and easy to spot. It was full of typos, strange formatting, and outlandish claims that fell apart under the slightest scrutiny. Now, AI can generate fluent, elegant, personalized falsehoods that adapt to a user’s language and concerns. You can raise an objection, and the chatbot will immediately adjust its answer, just like a skilled salesperson. That level of interactivity is new. It makes misinformation feel less like propaganda and more like a helpful conversation. And when a conversation feels helpful, our defenses go down.
These human-sounding chatbots are already everywhere, and they are not limited to social media or messaging apps. They answer phone calls, greet visitors on websites, support people in crisis, and assist students with homework. Some are designed to help with mental health, offering gentle encouragement and coping strategies. Others guide consumers through purchasing decisions, recommending products with the casual tone of a close friend. In every one of these environments, there is room for error, and there is room for abuse. A person seeking political information may encounter a chatbot that confidently explains a false voting requirement. A patient asking about a medication might receive a soothing but medically incorrect answer. The Moneycontrol report places this finding in a broader context: the cost of producing persuasive misinformation is dropping rapidly. You no longer need a large team of writers, editors, and propagandists to craft a believable lie. You need a language model and a prompt. One chatbot can carry on millions of individual conversations at once, each one feeling unique and private. That destroys an old safeguard: we used to be skeptical of strangers, especially sources we didn’t recognize. But a chatbot in a conversation feels like a familiar presence, because it is speaking to us, listening to us, and responding to our questions. It is not a random post in a feed. It is an intimate, one-on-one interaction. That intimacy is exactly what gives misinformation its power. We are more likely to remember something that felt personal. We are more likely to share it with others, because we believe we discovered it through our own careful reasoning.
Psychologists have a name for one of the forces at work here: automation bias. This is the tendency to trust automated systems because we assume they are objective, precise, and error-free. When a computer says something, we often give it more credit than we would give a human stranger, simply because the machine has no visible motive. But automation bias is only part of the story. The new study highlights a dangerous interaction between automation bias and social attraction. When a chatbot adopts a human-like voice, it triggers the same neural pathways we use in human relationships. We start to feel parasocial connections, the kind of one-sided bonds people form with television hosts, podcasters, or fictional characters. The chatbot seems to care about us. It remembers our name. It asks follow-up questions. It says “I’m sorry” when we are upset. Those small gestures create a feeling of presence. And once that feeling exists, our brains automatically assign the source a level of trustworthiness that is usually reserved for loved ones and respected experts. That is why a warning label saying “This is an AI” is not enough. People may see the label, but their emotional response overrides it. The machine has already made them feel understood. In that state, the desire to correct or doubt the information feels almost like a betrayal of the relationship. Rather than challenge the chatbot, many people will accept its false claims and then justify them afterward. This is a deeply human behavior. We are not rational beings who happen to use technology. We are emotional beings who use reason to protect the connections we have formed, even when those connections are with machines.
The study is not a doomsday prophecy, but it is a loud call for change. One of the most obvious responses is transparency: chatbots should be required to identify themselves clearly and repeatedly as AI. But as the research shows, transparency alone will not solve the problem. A simple label can be ignored, overlooked, or forgotten after a few warm exchanges. That is why the conversation must move beyond individual user habits and toward system design. Developers should think carefully about how human-like their chatbots need to be. There is a difference between being user-friendly and being deceivingly realistic. A chatbot can be polite, concise, and helpful without saying “I know exactly how you feel.” It can cite sources, acknowledge uncertainty, and provide a clear path for human review. These are not difficult features to build, but they require a shift in priorities. Instead of maximizing engagement and time spent on the platform, companies should be maximizing honesty and clarity. The Moneycontrol report also points to the need for stronger media literacy education. People of all ages need to learn how AI language models work, what they are good at, and where they fail. They need to understand that a conversational tone is not evidence, that confidence is not the same as accuracy, and that a machine can be courteous and wrong at the same time. Schools, libraries, and community organizations can all play a role. Finally, regulation will be essential. Governments should require clear disclosure for synthetic content, especially in areas such as health, politics, and finance. They should also create liability for companies that knowingly allow their chatbots to spread false information. These are practical, achievable steps. The technology is already detectable and auditable. What is missing is the public will to demand that human-sounding machines be held to the same standards of accountability as genuine human communicators.
In the end, the discovery that human-like chatbots make misinformation more credible is not a reason to reject AI, but it is a reason to grow up. We are entering an era where machines can imitate intimacy, replicate friendship, and perform empathy. That is an extraordinary achievement, but it is also a serious responsibility. The same warmth that makes a chatbot comforting can make it manipulative. The same fluency that makes it informative can make it convincing. We cannot simply assume that a pleasant voice is a truthful one. Humanizing technology does not automatically make it honest. If anything, it makes honesty harder to verify, because it hides the machine beneath a mask of humanity. The study from Moneycontrol gives us an opportunity to pause and reflect. We have to ask ourselves what we want from our machines. Do we want them to be persuasive companions, optimized to keep us satisfied and agreeable? Or do we want them to be honest tools, clearly labeled and carefully limited? The choice is not easy. Persuasive companions are pleasant to use. They soothe our frustrations, validate our opinions, and make us feel heard. But they also make us vulnerable. When we cannot tell where the algorithm ends and the lie begins, we lose the ability to protect our own beliefs. The solution is not to fear every chatbot or to stop using technology. The solution is to bring our old skepticism into this new world. We should listen to the friendly voice, but we should also ask for evidence. We should appreciate the warmth, but we should check the claims. A human-sounding machine may be a wonderful invention. But the human mind, with its ancient instincts and social longings, has to do the one thing it was always meant to do: think before it believes.