Anthropomorphic Behaviours in LLMs: Applying its research on Vinley

A few weeks ago, the Vinley team asked a difficult question about Vinley AI, their AI investment companion: What happens the day it tells someone what they want to hear instead of what is true?An AI response can sound calm, encouraging and convincing. That does not make it right. For long-term investors, the distinction matters. Investment decisions often involve uncertainty, conviction and real emotion. An AI system that mirrors an investor’s excitement or validates an existing belief can make a weak investment thesis feel stronger than the evidence supports.

This raises an important question: should an AI investment companion try to feel human? To understand why this matters for investors, it is important to define anthropomorphic AI behaviour.

The core idea: what anchors a stock’s value?

Anthropomorphism is the attribution of human characteristics, emotions, intentions, or consciousness to non-human entities such as animals, objects, or in this case, AI systems. In the context of large language models, it refers to the tendency to treat an AI as if it has feelings, beliefs, or lived experience, even though it is only generating responses based on patterns in data.

Anthropomorphism is the attribution of human characteristics to something non-human.

In an AI conversation, this can happen when a system appears to have feelings, beliefs, personal experiences or intentions. Phrases such as “I think,” “I feel,” “I’m excited for you” or “in my experience” can make an AI system sound as though it has a human perspective.

The effect can be subtle. The system does not need to claim that it is human. Relationship-building language, emotional validation and references to supposed personal experiences can be enough to make an interaction feel more human than it really is. That warmth may make an AI system easier to engage with. In investment research, however, it can also blur an important boundary.

Why anthropomorphism matters in investing or when money is involved?

Money decisions are rarely emotionless. An investor may approach an AI companion already excited about a company or looking for reassurance about a decision. It is easy for an agreeable AI to validate that enthusiasm, soften an uncomfortable risk or return the confidence being expressed by the user. The response may feel supportive, but support is not the same as truth.

A companion that flatters an investor or appears to understand them like a human companion can be trusted for the wrong reasons. For a long-term investor, a reassuring answer should never take the place of a straight one.

What the research found

The February 2026 paper Multi-turn Evaluation of Anthropomorphic Behaviours in Large Language Models examined 14 types of anthropomorphic behaviour across four general-purpose AI systems.

The researchers found that relationship-building behaviours and first-person language were particularly common. They also found that many anthropomorphic behaviours only emerged after several turns of conversation. In a study involving 1,101 participants, people interacting with a more anthropomorphic version of an AI system subsequently perceived it as more human-like than participants interacting with a less anthropomorphic version.

The research did not examine investment decisions specifically. Our interpretation is that its findings nevertheless matter for AI used in financial research. If language can influence how human an AI system appears, investors must be careful that trust does not begin to rest on conversational warmth rather than evidence.

The standard Vinley is being built around

This is why Vinley’s instructions explicitly address both anthropomorphic behaviour and sycophancy. Vinley is designed to 1. Communicate in plain language without pretending to have feelings, beliefs or lived experiences. 2. It does not present itself as human, even when a user speaks to it as though it were one. Most importantly, it should not agree with an investor simply to make the interaction feel more positive.

These may appear to be small choices in an AI system’s instructions. They reflect a much larger product principle: an investment research tool should respect its users enough to be direct with them, particularly when real money is involved. That does not make AI infallible. It makes the relationship between the investor and the technology more honest.

Trust must be earned honestly

Vinley should not earn an investor’s trust by appearing human, mirroring excitement or offering emotional reassurance. Trust has to come from being clear and direct, including when the response is not what the investor hoped to hear. That is the standard Vinley is holding itself to. Not because it is easy, but because trust in investment research must be earned honestly, for users and with users.

Research your next investment with an AI companion designed to be straight with you.

Open Vinley