The Making of a Virtual Influencer Revisited, 2026

In March 2021, I published The Making of a Virtual Influencer in Container Magazine. It was research born out of my dissertation for my MA in Virtual Reality — an exploration into the social impact of CGI personas, built from the inside out.

Six months after losing my sister to a brutal brain tumour, her treatment compromised by the chaos of the pandemic. Her daughter, Ashleigh, was on the frontline nursing COVID patients. As we lived through one of the hardest periods of our lives, I struggled to find ways to support Ashleigh and process the grief myself. Evie was born out of that struggle: an avatar modelled on Ashleigh's likeness and a project built from empathy and compassion forcing me to solve creative and technical problems while focusing on storytelling that mirrored that struggle.

Five years on, I wanted to hold my original predictions up to the light, sit with what I got right, and be honest about what I didn't see coming.

Five Predictions. Five Years On.

Back in 2021, I made five core predictions. Here is where they stand today, in 2026.

First, I predicted the avatar economy would grow at a massive rate. This was correct — beyond anyone's projections. According to data tracking from SNS Insider, the global virtual influencer market has climbed to an estimated value of 9.75 billion dollars, with brand spending on AI personas jumping dramatically year-on-year. Further data points can be found via Digital Applied's 2026 influencer marketing statistics. However, the real story isn't the corporate titans like Lil Miquela. It's the total collapse of the barrier to entry. What once required a studio of 3D artists can now be done by a single creator with a laptop. The tools have democratised, and the growth has come from everywhere: bedroom creators, YouTube, and independent artists.

Second, I predicted that digital assets would replace .png-based cryptoart, and third, that cryptoart would migrate to sustainable blockchains like Tezos. On this, the underlying tech exploded, but the financial vessel changed. The NFT boom peaked around 2021 and 2022 and sharply contracted — the market's obsession with static, tokenised .png files completely ran out of steam. But if we look past the speculative crypto bubble, the prediction about the assets themselves was directionally spot on. The true evolution shifted away from the blockchain and directly into advanced pipelines like DCC 3D modelling, generative AI, and Gaussian Splatting.

We moved from trading flat jpegs to creating genuine "phygital" assets — where digital garments and objects designed in tools like CLO 3D or Substance 3D have exact physical counterparts, mapping fabric physics and real-world textures directly onto digital avatars. Instead of traditional, slow rendering methods, the rise of Gaussian Splatting means we can now capture real-world items and spaces instantly in 3D with a mobile phone. AI didn't replace the meticulous pipeline of traditional 3D DCC tools; it integrated with it, transforming how quickly we can iterate, simulate fabric, and populate virtual environments. The cultural energy didn't die — it just grew up and moved into the production pipeline.

Fourth, I predicted that Amazon Alexa would get a face and a body. On this, the form factor shifted completely from 3D space to the 2D video frame. While Alexa still doesn't have a physical or fully rendered 3D body in our living rooms, the broader concept of voice assistants gaining a human face completely materialised through a different medium. Instead of waiting for real-time 3D environments to mature, platforms like Synthesia, D-ID, and ElevenLabs completely bypassed the 3D pipeline. They proved that embodiment didn't need a game engine; it just needed a video player. By mapping neural text-to-speech onto flat video generation, they turned scripts into hyper-realistic digital presenters in minutes. With the rise of interactive, conversational visual agents, the face of AI finally arrived — its platform is just streaming video, not 3D engine architecture.

Fifth, I predicted that virtual influencers would become as numerous as VTubers. This was spot on. The proliferation happened, but the original faces got complicated.

Lil Miquela — who earned an estimated $11.7 million (around £9 million) for her creators in 2020 — proved that a well-constructed virtual identity functions more like a legacy cartoon character or a Hollywood brand than a human influencer. When her parent company, Brud, faltered, the character didn't vanish — because her value wasn't tied to a single founder's business model. It was tied to her audience equity and her cultural footprint. It's a structure the VTuber industry understood from the start: agencies like Cover Corp and Anycolor built rosters where the character outlasts any single business decision made behind the scenes.

By signing with major talent agencies like CAA and securing blue-chip campaigns with BMW and high fashion houses, she proved that corporations actually prefer the ultimate benefit of virtual creators: complete risk management. A virtual influencer doesn't get caught in an organic, late-night scandal, they don't age out of a demographic, and they can be in three places at once.

The pivot from "edgy tech experiment" to "stabilised corporate asset" is exactly how the industry managed to scale alongside VTubers, even if the pioneering companies behind them had to collapse to pave the way.

What I Didn't See Coming

For one, the tools became entirely native. In 2021, I wrote about Unreal Engine's MetaHuman Creator as an emergent, cloud-based technology — and while the direction of travel was clear, Epic themselves weren't certain how they would productise it, or whether they would charge. Today, as of Unreal Engine 5.6, it is fully integrated directly inside the editor. The friction is gone. It is no longer an emerging experimental feature; it is a core foundational baseline. Let me get back to you on how many MetaHumans I currently have...

Hand-in-hand with native engines came the absolute explosion of live virtual performance. When I first built Evie, the goal was to eventually find a way to live-perform her — to close the lag between the digital skin and the human soul behind it. That ambition is exactly what led me to my work at Disguise. In 2021, the infrastructure required to drive high-fidelity digital humans live on stage, in real-time, with zero latency was painfully complex. Today, the real-time virtual production workflow has matured into an industry powerhouse. By connecting live performance hardware, advanced motion capture data, and LED volumes, the boundary between the virtual asset and the physical space has dissolved. Avatars are no longer trapped in frozen, pre-rendered social feeds; we can exist, react, and perform live in the room.

One paradigm shift I completely missed on an architectural level was the infrastructure that would connect these personas to the wider digital world. In 2021, a virtual influencer was essentially a static script or a curated feed. Enter Model Context Protocol, or MCP. This development has quietly dismantled the siloed nature of AI. Instead of virtual beings existing purely inside their own proprietary codebases, MCP provides an open standard that allows avatars and agents to securely fetch data, interact with development tools, and interface across different platforms and environments. It took virtual influencers from being isolated 3D animations and turned them into dynamic, tool-wielding participants in the broader internet ecosystem.

At the same time, the bias scaled, and the ethics curve flattened. Five years ago, I flagged documented racial bias in GPT-3. At the time, I was even considering pursuing a PhD in AI Ethics to properly parse these structural fault lines. I didn't end up taking the doctoral route, but the problems I wanted to study have only amplified.

GPT-3 has been superseded many times over, but the bias wasn't engineered out — it just scaled up. Gender, racial, and cultural bias remain live, heavily documented flaws across every major AI model in 2026. The technology curve has been steeper than almost anyone anticipated; the ethics curve has simply not kept pace. We are nowhere near past this.

Meanwhile, regulation moved at a glacial pace. The legal gaps flagged by Truth in Advertising have marginally narrowed but haven't closed. The FTC now requires sponsored content to be clearly marked as both paid and AI-generated. Most established virtual personas carry disclosure tags in their bios, but how this holds up under global advertising codes remains largely untested.

But perhaps the most unexpected thread involves my original research with Kuki, the Pandorabots conversational AI. Back then, I spent a month talking to Kuki for 15 minutes a day alongside fellow researcher Anna Xygkou-Tsiamoulou.

Today, Dr Anna Xygkou-Tsiamoulou is the Director of Space Environmental and Behavioural Systems at Titans Space Industries — and the world's first female Greek astronaut candidate. Her COSMIC project uses neuroscience and AI to build conversational companions for deep space missions. In the brutal isolation of space, an empathetic AI isn't a marketing novelty; it's vital psychological infrastructure. Seeing a chatbot experiment evolve into a support system for astronauts makes this entire five-year journey feel profoundly worth it.

Five Years Later

The avatar economy grew entirely outside the lines of corporate expectation, fuelled by the grassroots energy of indie creators. We have achieved the technical wizardry to build flawless digital humans, wire them into interconnected open protocols, perform them live on stage, and launch them into deep space.

And yet, after all the market projections, YouTube experiments, and increasing technical capability — despite the AI, despite the innovation curve — do we need human connection more than ever?

With the overwhelming deluge of AI slop, with deep fakes and unlicensed digital replicas eroding both identity and ownership of our likeness — are we actually seeking something real?