AI Can Be More Comforting Than a Person: What the Science Reveals and What It Means for India

Generative artificial intelligence has rapidly expanded beyond mundane workplace automation and coding tasks. Millions of users now turn to large language models (LLMs) to process complex internal states, vent after difficult workdays, and manage personal anxiety. What was once seen as a gimmick is developing into a primary emotional outlet, particularly among young adults.

While traditional intuition suggests that human interaction is inherently superior for emotional support, empirical research presents a far more complex picture. Recent academic studies demonstrate that AI-generated messages are frequently rated as more empathic, structured, and comforting than those written by real people. However, this phenomenon—termed the “AI advantage”—comes with significant psychological caveats, disclosure paradoxes, and distinct systemic implications for emerging digital health landscapes like India.

Table of Contents

The Emotional Shift: How Common Is AI-Driven Comfort?

Data indicates that using generative AI as an emotional sounding board is no longer a niche behavior. According to a 2025 survey of 1,058 individuals aged 12 to 21 in the United States, 13% of respondents reported seeking advice from generative AI when feeling sad, angry, or nervous. That proportion rose to 22% among young adults aged 18 to 21, establishing that over one in five young adults regularly leverages chatbots for emotional support.

Rather than relying solely on traditional social networks or professional counseling, users are actively querying LLMs to navigate interpersonal conflicts, anxiety, and daily distress. To understand why users report high levels of comfort from these interactions, behavioral researchers have begun benchmarking LLM outputs directly against human responses.

Quantifying the ‘AI Advantage’: Why Chatbots Score Higher in Empathy Tests

To evaluate the efficacy of conversational AI in emotional situations, researchers from the University of Manchester and Durham University conducted a series of five experiments comparing human-written supportive messages with those produced by LLMs. As detailed in research reported by Yahoo News UK, the team examined how individuals respond to different sources of guidance when experiencing emotional distress.

In one experiment involving 390 participants who imagined scenarios provoking anger, sadness, or fear, individuals read either a human-written or AI-generated response without knowing the source:

Anger and Fear Scenarios: Participants consistently rated AI-generated responses as significantly more emotionally supportive than human-written alternatives.
Sadness Scenarios: There was no statistically reliable difference in perceived support between human and AI responses.
Emotional State Impact: In fear-inducing scenarios, AI messages increased feelings of calm more effectively than human messages, though they did not lead to a greater direct reduction in overall fear.

These findings align with a systematic review of 23 separate studies, which observed that users routinely rate generative AI outputs as more empathic—appearing noticeably understanding, caring, and attentive—than human interactions. Researchers refer to this trend as the “AI advantage.”

Actionable Support: The Mechanism Behind the Scores

Why do algorithms excel at communicating comfort? The Manchester and Durham research isolated key variables to explain why LLMs perform so well in controlled experiments:

Consistency and Structure: Humans experience fatigue, distraction, and uncertainty when attempting to comfort others. AI systems deliver structured, coherent, and attentive responses on demand without emotional exhaustion.
Actionable Guidance: In follow-up experiments, explicit validation of feelings alone did not account for the greater emotional improvement delivered by AI. Instead, the critical factor was “actionable support”—offering specific, practical, and realistic steps that the recipient can execute immediately.
Restraint and Framing: Practical help requires careful execution. A 2024 study cited in the research noted that excessive or overwhelming lists of suggestions can backfire, making individuals feel less heard. AI models trained to provide clean, manageable advice strike a effective balance between listening and solution-building.

When human participants offered equally specific and manageable practical steps, evaluators judged their support as just as comforting as the AI outputs. However, in real-world scenarios, humans frequently fail to provide structured, practical steps in moments of acute stress.

The Human Paradox: Disclosure Changes Everything

Despite the high ratings given to AI-generated text in blind tests, human psychology exhibits a sharp contradiction when the source is revealed. This phenomenon highlights a fundamental paradox in human-AI interaction.

Across nine studies involving 6,282 participants, researchers found that when the exact same AI-written response was explicitly labeled as coming from an artificial intelligence rather than a person, ratings of empathy and emotional support dropped significantly. Furthermore, a 2024 study demonstrated that while AI messages initially made participants feel more heard, this benefit eroded the moment they learned the message originated from a machine.

When given a choice, people consistently prefer human interaction for emotional engagement—even when choosing a human means enduring longer wait times. Scientists suggest this stems from relational dynamics: human responses signal that someone is willing to expend time, cognitive energy, and emotional effort on the relationship. A chatbot’s response, while technically flawless, costs the machine nothing.

This tension underscores the importance of transparent system deployment. As explored in our analysis on why organizations must govern AI agents with clear boundaries rather than pretending they are human, attempting to disguise automated systems as genuine human empathy degrades trust over time.

The Indian Context: Healthcare Spending Gaps and Digital Frontlines

The findings around AI-driven emotional support take on heightened importance when contextualized within developing healthcare infrastructures. In India, access to formal mental health services and psychiatric resources remains constrained by structural expenditure gaps.

According to World Health Organization data published by The New Indian Express, high-income countries spent approximately $7,300 per person on healthcare in 2023 (adjusted for purchasing power parity), compared to $346 per person in India. While India’s spending is nearly three times the low-income country average ($125), it represents less than one-twentieth of the high-income average and aligns closely with the lower-middle-income average ($360).

Data from India’s National Health Accounts (2021-22) highlights the specific financial pressure on households:

Total Health Expenditure: ₹9.04 lakh crore, representing 3.83% of GDP.
Per-Capita Spending: ₹6,602 at current prices.
Financing Breakdown: The government financed 48% of total health expenditure (up from 28.6% in 2013-14), while direct out-of-pocket household spending accounted for 39.4% (down from 64.2% in 2013-14), equaling ₹2,600 per person.

Given that direct household payments still constitute nearly 40% of healthcare costs, affordable mental healthcare remains out of reach for a vast segment of the population. Global life expectancy data underscores this divide, showing that individuals in poorer nations live an average of 16 years less than those in richer nations.

At the same time, international perception of India’s development trajectory remains globally relevant. As highlighted by Pew Research Center survey data across 24 countries, global audiences hold a net-positive view of India’s growth and expanding international reach. In this landscape, accessible digital tools—including generative AI applications—are emerging as informal first-line mechanisms for emotional triage, filling critical gaps where institutional mental health infrastructure is under-resourced.

Decision Framework: Evaluating AI for Emotional Support

To safely navigate the benefits and limitations of using generative AI for emotional guidance, users and developers can use the practical decision criteria outlined below:

Evaluation Criterion	AI Support Strengths	Human Support Strengths	Recommended Use Case
Availability & Speed	Instantaneous, 24/7 availability with zero queue times.	Subject to schedule, delay, and emotional availability.	Immediate crisis de-escalation, late-night anxiety, initial venting.
Actionability	Generates structured, clear, and realistic step-by-step guidance.	May become reactive, unscripted, or unsure of advice.	Organizing thoughts, drafting difficult messages, step-by-step grounding.
Relational Value	Lacks real relationship, effort, or mutual personal context.	Demonstrates genuine emotional effort, care, and shared history.	Deep personal bonding, long-term emotional recovery, shared mourning.
Perceived Empathy	High technical empathy scores in blind evaluations.	Highly valued when authentic; diminished if distracted.	Reframing negative self-talk, identifying emotional triggers without judgment.

Final Verdict: A Useful Tactical Tool, Not a Replacement for Connection

The research is clear: generative AI possesses a measurable advantage in delivering consistent, highly structured, and practical emotional support during moments of stress, particularly for feelings like anger and acute fear. Its ability to provide actionable, step-by-step guidance makes it a valuable self-regulation tool.

However, AI cannot replicate human care. The moment users recognize an emotional response as algorithmic, its comforting effect declines, highlighting our fundamental desire for genuine human effort and real relationships. In regions like India, where per-capita healthcare spending ($346) leaves significant gaps in mental health access, AI offers a low-cost, accessible option for initial self-help and emotional grounding. Ultimately, AI should be viewed as a helpful tool for personal emotional management, rather than a replacement for human relationships and professional healthcare.