Would OpenAI's GPT-5 and Other LLMs Be More Approachable with Avatars?

Aug 5, 2025

If you referred to Google as your best friend, most people would assume you were telling a really corny joke. The same logic would apply if you were to swap out “Google” for “ChatGPT” or another large language model. 

For many, these models are merely tools. Some folks have turned to systems like ChatGPT, Jasper, and Cursor to increase their efficiency and enhance their output at work. Others use LLMs to explore trending news stories, learn new languages, and delve into hobbies like running, fishing, and gaming when they have a spare moment. While some people continue to use LLMs after leaving the office, the majority of overall LLM usage is productivity-focused. 

Even with the recent spike in LLM adoption, users are still finding these systems emotionally distant. Sure, they can answer questions, complete tasks, and simulate conversation. But something’s missing. 

What if they could feel less like a chatbot and more like a guide, collaborator, or even — a friend?

How do AI Companions change the way people interact with LLMs?

AI companions introduce a sense of presence to LLM interactions, transforming what previously felt like a transactional exchange into a more natural, human-like conversation. 

When discussing the impact of AI companions, it’s important to underscore the significance of conversational learning in general. Conversation, after all, is one of the best learning tools humans have. As Dartmouth College professor of psychological and brain sciences Thalia Wheatley, PhD, so eloquently states, conversation is “ancient technology for aligning our brains so that we can be on the same page.” 

This raises the question: how do AI companions help human brains align with LLMs?

Imagine if you could only text your friends, never to see their faces in person. After a while, would you really consider them friends? Perhaps just virtual acquaintances? Traditional LLM interfaces, at their best, resemble a static text message thread. AI avatars provide LLMs with the opportunity to enhance usability through personalized experiences. 

There are a couple of key design elements that make talking to AI companions more natural than conversing with a standard LLM. 

For starters, AI companions put a face to the name. Users can see what digital assistants like ChatGPT, Gemini, and Perplexity “look like” through their avatar representations. Avatars leverage familiar human expressions, in turn creating a more comfortable exchange. 

A user’s comfort level is critical to their ability to learn, as proven in a 2024 study by researchers from China’s Zhejiang Sci-Tech University. The study, which assessed user interactions with virtual avatars in educational videos, concluded that highly expressive avatars can “significantly improve” a user’s ability to learn. 

That said, AI companions can’t rely solely on facial expressions to make users feel comfy. A 2025 research study from the University of Aveiro found that avatars powered by both AI-generated visuals and audio were the most successful in increasing user engagement. Participants in the study were asked to watch a series of four learning resource videos, with varying combinations of human and AI-generated visuals and audio. The AI voice and AI avatar combination achieved superior performance, with users experiencing high engagement and low extraneous cognitive load in 96.7% of cases

Speech-to-Text (STT) and Text-to-Speech (TTS) technologies, which are used by most modern AI avatars, have largely fueled the development of influential AI voices. As the former system transcribes human speech into text while LLMs detect the context and intent of the user’s input, the latter converts the responses into natural-sounding speech. 

The end result? 

Fluid, human-esque dialogue that mirrors a real conversation.

Do avatars make AI more relatable and emotionally engaging?

Yes, fine-tuned avatars can decrease the cognitive load associated with AI interactions, making them more relatable and increasing emotional engagement among users. 

Trust plays a major role in this phenomenon. When engaging in conversations with LLMs, users need to trust that the system’s outcomes will be accurate. If the outputs are inaccurate, the user will perceive the LLM to be unreliable. Trust becomes even more important with AI avatars, which must combine visual expression, vocal tonality, factual accuracy, and emotion for users to perceive them as a trustworthy companion. 

When an AI avatar misses the mark with one or more of these characteristics, it’s pretty easy to tell. Instead of instilling trust, the digital characters enter into the uncomfortable realm of uncanny valley, where they look and feel like humans, but something about them is just… unsettling

On the flip side, an avatar that masterfully blends tone, expression, and accurate information to form coherent responses can make AI engagements more relatable and reduce end-user resistance. As it currently stands, the impersonal nature of today’s text-heavy LLMs can make for a rather intimidating user interface, especially for those who are new to AI and LLMs. A high-functioning AI avatar can remove those barriers and make chatbot discussions feel more “alive” across a variety of scenarios. 

📄 A once dull conversation with an LLM about an upcoming exam can be converted into a highly motivational and energizing chat with an avatar ahead of test day. 

🏖️ Vacation planning sessions that previously felt disconnected now feature avatars expressing authentic curiosity alongside their human counterparts. 

🏋️ Post-workout recaps evolve from routine data entry to inspiring, back-and-forth exchanges with an avatar that seems genuinely interested in your progress. 

Can avatars expand AI access to a wider range of user types? 

Avatars present a tremendous opportunity for LLM developers to attract wider audiences. Users with different communication styles or needs can interact with AI avatars to reduce any friction, confusion, or challenges that they may face with LLMs today. 

Whereas neurodiverse individuals or non-native speakers may have previously struggled to use ChatGPT or Perplexity, these groups can now speak to an AI avatar as they would a friend or tutor. The same goes for overwhelmed first-time users or folks who are not technologically inclined. 

No more confusing onboarding processes or required technical know-how — just straightforward conversations with a digital companion. 

Try an AI companion out for yourself and let us know your thoughts on X @genies.

Sign up to get the latest updates from Genies

Sign up to get the latest updates from Genies

Sign up to get the latest updates from Genies