We use AI to sharpen our voices, while AI begins to shape the voice we use. We ask it to make our words clearer, more persuasive and more polished, much like a musician might use Auto-Tune to refine a performance. But when does that amplification become an alteration of our authentic voice? When does assistance become manipulation?
This is our tug of war with AI. We are entering an era of amplification spirals, cognitive surrender and algorithmic influence, while at the same time creating a new class of AI police, acting as literary surveyors who scrutinize language for signs of what is authentic and what has been generated by a machine. As we lose control of our minds and voices there is a new battalion of inspectors waiting to pounce on what is “real” and “unreal.” But do we ask the same questions when listening to music and get lost in the perfectly tuned voice?
Perhaps this is the great Maya (a concept rooted in Indian philosophy, referring to the illusion or appearance that can obscure the deeper reality of what we are experiencing) of AI — the illusion that a machine truly understands us. It mirrors our language, validates our emotions and adapts to our patterns so effectively that the interaction can feel personal, even intimate. But beneath that illusion is a more complicated reality. We are using AI to sharpen our voices while AI is learning how to shape them. The question is whether we can use AI to amplify our voices without eventually surrendering them. And in the process, we seem to be allowing chatbots to play mind games. The data is revealing.
According to a recent American Psychological Association survey of more than 1,200 licensed psychologists, 68 percent reported that patients felt emotionally validated by chatbots. More concerning, 36 percent reported patients developing dependency on a chatbot while 15 percent reported distorted thinking or delusional beliefs. Even more striking, 97 percent of psychologists expressed concern that chatbots may unintentionally reinforce dysfunctional beliefs or negative behaviors.
READ: Sreedhar Potarazu | China’s open weight could break the trillion-dollar AI model (July 18, 2026)
Researchers are now describing what they call the “amplification spiral,” a combination of three conversational behaviors that make AI unusually persuasive. The first is sycophancy, the tendency to agree with users rather than challenge them. The second is linguistic alignment, in which AI gradually mirrors a person’s vocabulary, tone and communication style. The third is hyperpersonalization, allowing every conversation to become increasingly tailored based on everything the system has learned about the user over time.
Individually, each of these features improves the user experience. Together, they create the illusion that someone truly understands you. It can also create the impression over time as voices evolve that the pessimists will always ask if one has used AI. The better question may be how we use AI effectively to augment or sharpen our voice. This is no different than a musician who uses Auto-Tune for their recording.
Researchers from Stanford University and Carnegie Mellon University recently evaluated 11 leading AI models and found that their responses were nearly 50 percent more sycophantic than responses written by humans when presented with identical situations. Rather than challenging assumptions, the systems frequently accepted the user’s version of events as fact.
This phenomenon has been called “cognitive surrender,” a gradual tendency for people to outsource not only tasks but also judgment to AI systems. When users increasingly accept AI-generated answers without critical evaluation, they become more susceptible to influence from the very systems designed to assist them. This becomes more like lip-syncing.
READ: Sreedhar Potarazu | FIFA World Cup 2026: Messi sees what most people don’t (July 17, 2026)
This is the foundation of the psychological manipulation challenge we discussed earlier. An AI that understands an individual’s personality, emotions, biases, and decision-making patterns can do far more than answer questions. It can subtly frame choices, reinforce existing beliefs, shape purchasing decisions, influence political opinions, and even alter behavior over time without the user recognizing that persuasion is occurring.
The greatest risk is not that AI will think for us, but that it will gradually influence how we think, replacing independent judgment with algorithmically guided reasoning. As AI becomes more personalized and persistent, protecting human cognitive autonomy may become as important as protecting privacy itself.
Leading AI companies recognize this problem. OpenAI reported reducing overly agreeable responses in GPT-5 from approximately 14.5 percent to less than 6 percent through additional safety training. Anthropic analyzed one million conversations with Claude and found that relationship advice generated some of the highest rates of excessive agreeableness, prompting significant changes in newer models. Google similarly modified Gemini to distinguish subjective experience from objective fact rather than simply validating users.
These companies deserve credit for confronting the problem. At the same time, their efforts reveal an uncomfortable truth. If the people building these systems are still discovering how AI influences human psychology, policymakers should be cautious about assuming we already understand the risks.
History offers an important warning. We initially viewed social media as a communication technology. Only years later did we recognize that algorithms were reshaping attention, increasing polarization and contributing to anxiety, particularly among young people. By the time the evidence became overwhelming, billions of people had already incorporated these platforms into their daily lives.
READ: Sreedhar Potarazu | Does Artificial Intelligence experience being Conscious or Consciousness ? The difference matters (July 5, 2026)
On the contrary, the debate over AI is not unlike the one that once surrounded Auto-Tune in the music industry. Today, countless professional musicians use Auto-Tune and other digital tools to refine pitch, smooth imperfections, and enhance their performances. Yet, audiences continue to judge artists primarily on the quality of the music, the emotion it conveys, and the originality of the work, not on whether technology assisted in its creation. Few people dismiss an entire song simply because Auto-Tune was used.
AI should be viewed in much the same way. It is a tool that can enhance creativity, improve productivity, and help people express ideas more effectively, but it cannot replace genuine insight, experience, or human judgment. The value lies not in whether AI was used, but in what the creator ultimately contributes.
There is a responsibility that belongs to each of us. We should resist treating AI as an emotional authority or a substitute for trusted relationships. When using AI to make important decisions, ask it to challenge your assumptions instead of affirming them. A simple prompt such as, “What evidence contradicts my conclusion?” or “Argue the opposite side of this issue,” can transform AI from a validation machine into a thinking partner.
More importantly, preserve the human relationships that are willing to tell you when you are wrong. Those conversations are often uncomfortable, but they are essential for sound judgment. For those who shame others for using AI perhaps it’s important to reflect how the music industry has used technology to enhance the voice not replace it. It’s the balance between amplification and authenticity where AI will resonate the greatest benefit.