With over 1.4 million followers on TikTok, Malaysian content creator and entrepreneur Bella Khann recently became one of the first creators to deploy an AI avatar of herself on her secondary TikTok account, which has over 500,000 followers. The AI generated avatar, Khann's digital twin, will primarily be used for live commerce streams.
Khann is one of the first recruits to AnyLive's Creators, a program where AI avatars are able to host livestreams on behalf of creators on creators' official accounts on YouTube, TikTok and Facebook, enabling them to passively earn income through affiliate marketing campaigns delivered through livestreams.
The introduction of AI avatars into live commerce comes at a time when the market in APAC is booming. Live commerce in the region is projected to exceed US$77 billion by 2030. At the same time, the region’s creator economy is on track to surpass US$75 billion by 2032. As these two spaces converge, new challenges emerge such as longer streaming demands, limited creator bandwidth, and a lack of data-driven feedback loops.
To try and tackle these challenges, Japanese tech company AnyMind Group have launched AnyLive for Creators, offering a new model where creators can develop AI avatars to host livestreams on their behalf.
"We see AI avatars as a means to augment human capacity and capability, rather than a replacement for what creators can currently do, especially in the live commerce space," says Akinori Kubo, managing director of global ecommerce at AnyMind Group. "We’re an advocate for brands to take on a hybrid model, where humans front livestreams during the most optimal periods, and AI avatars take over while the humans rest, creating a 24/7 sales channel."
AnyLive gives creators the ability to develop AI avatars that can host livestreams on their behalf.
The ultimate ambition is to transform how creators approach live commerce.
"Instead of being tied to an individual’s time and energy, creators can now create a digital twin of themselves and extend their influence into new platforms and languages," Kubo explains. "This without doubt allows creators to maintain relevance and presence without burning out, and enables brands to work with them at a much greater scale."
Unlimited scalability
Remington Ji, chief digital officer at Ogilvy China, foresees rapid democratisation and scaling of live commerce via AI avatars over the next two to three years.
"For example, in China, we’re already seeing a clear split: 'prime-time humans, always-on avatars,'" Ji says. "Avatars can livestream simultaneously across Douyin, Taobao Live, Kuaishou, and other platforms, extending reach and sales windows far beyond human limits. Top KOLs will use avatars to extend their brand presence, letting avatars handle smaller promotions while they focus on high-value engagements."
Unlimited scalability is a key advantage of AI avatar livestreams, permitting brands to run 24/7, multi-platform, multi-language sessions at a fraction of the cost.
"They allow precise data-driven personalisation for different audience segments, while reducing reliance on a single influencer’s schedule or stamina," Ji adds. "For emerging brands, this means faster market entry and the ability to test content formats at scale."
AI avatars lower entry barriers drastically. Tasks once requiring large teams and seven-figure budgets can now be done quickly and affordably.
"Brands and creators can respond to trends in real time, launching new characters or campaigns within days to captivate their audiences," says Chye Yong Hock, chief innovation officer at Dentsu Creative Singapore. "This has opened an exciting frontier previously limited by high costs and long development cycles."
For instance, virtual influencers can now convincingly interact with products thanks to advanced video AI tools such as VEO or WAN. Another significant benefit is cost savings. AI avatar-driven livestreams reduce costs relative to fully human-hosted events, allowing funds to be allocated to attracting more viewers.
In a test last year, AnyMind collaborated with Evian in Thailand to deliver over 800 hours of livestreaming on Shopee. Humans fronted just 45 hours, while AI hosted 820 hours, leading to a 90% reduction in hourly costs and a 200% year-on-year sales increase.
No substitute for human hosts
Despite promising metrics, some brands remain sceptical, insisting that live commerce must be "truly live, human, and interactive."
"AI avatar hosts for live commerce don’t work," says Dominic Powers, CEO of APAC & LATAM and director at Stickler. "At least not if your goal is genuine engagement and commercial success."
Stickler tracked a recent TikTok livestream from Innisfree in Thailand featuring an AI avatar. The stream earned a Stickler Engagement score of just 29, deep in the 'Below Average' range. Top streams often exceed 180, with Southeast Asian averages between 50-70. The stream attracted only 17 comments over nearly 8 hours, showing poor audience engagement.
Powers attributes the failure to a human connection deficit, not technology.
"TikTok users come for conversation, shout-outs, answers to product questions, and the feeling that someone on the other side sees them. The best streams aren’t just digital QVCs, they’re social, always on, places where viewers can drop in and hang out," he explains. "By contrast, AI hosts are literally tone-deaf. They are monologue, and in this case on loop, and not in conversation. They mimic presence but can’t replicate that connection."
A stream from Innisfree in Thailand featuring an AI avatar achieved an engagement score of just 29, deep in the 'Below Average' range.
Questions linger about whether AI avatars can match human authenticity. Kubo is focused on three pillars: relevance, balance, and transparency, and continually optimises AI avatar interactions.
"We analyse both human and AI live commerce data such as viewer retention, drop-off points, and comments, and adapt scripts to match different shopping occasions," says Kubo. "At the same time, we believe trust still comes from human connection. A hybrid model ensures that while AI avatars bring consistency, scalability, and 24/7 presence, human hosts remain vital for building emotional resonance and empathy."
Transparency remains paramount.
"All AI-fronted sessions on AnyLive are clearly labelled so audiences understand they’re interacting with an avatar," Kubo adds. "This openness helps set expectations and builds long-term trust, while also showcasing what AI can add to the shopping experience."
Amid heightened regulatory scrutiny, clear disclosure that AI-generated content is artificially created is mandatory. China's AI labelling regulation, effective September 1, 2025, requires all AI-generated content—text, images, audio, and video—to carry explicit labeling.
"Some platforms like Douyin and Kuaishou don’t even allow pure AI avatar broadcasts," Ji notes. "And as more brands enter the space, standing out creatively will become harder, making strong storytelling and brand personality even more critical."
Additionally, some viewers instinctively distrust AI-generated content, viewing virtual influencers as "fake" or inauthentic. Such scepticism can hinder engagement and provoke backlash, especially if avatars lack development and livestream execution is poor.
Studies reveal that globally, less than half (46%) of people trust AI despite 66% using it regularly, with older generations tending to distrust AI content more, favouring clear labelling.
"Responsible AI avatar use in live commerce means clear disclosure, truthful content, data protection, and influencer consent, all under strict compliance with deepfake and livestreaming regulations," Ji says. "The winning approach is AI efficiency with human oversight, ensuring trust, safety, and brand integrity."
Revenue sharing
The revenue-sharing model for AI avatar livestreams is evolving. AnyLive follows a model where creators and the platform share sales-generated income.
"This creates a win-win-win situation, where both the creator and AnyLive are invested in and incentivised by the success of the brand," Kubo explains. "Additionally, branding-focused livestreams run on a fixed-fee model shared by platform and creator."
Hock anticipates that the model will grow to include direct virtual influencer participation and potentially fully AI-managed content systems with small creative teams for brand consistency.
"As the market matures, performance-based models will likely emerge where all parties share extra revenue growth from stronger sales, similar to today’s ecommerce landscape," he adds.
While AI avatars promise unlimited scalability, enabling brands to run 24/7, multi-platform, multi-language sessions cost-effectively, some experts emphasise the necessity of human hosts.
"If brands are serious about scaling live commerce, they need to scale authenticity," Powers cautions. "While AI can support scripting, data tagging, and analysis of vast structured and unstructured data from live sessions, the host needs to be human. Live commerce is not passive viewing; it’s real-time, brand-owned engagement. And you can’t fake that with AI. Not now. Maybe not ever. Not unless we’re ready to change what 'live' really means."
