Phoenix Publishing and Media president Song Jishu on AI and the future
The president of Phoenix Publishing and Media, and the lead on the publisher’s digital strategy, discusses its AI initiatives
Can you give me an overview of Phoenix’s digital transformation strategy in recent years?
Over the past few years, digital transformation in China’s publishing industry has accelerated significantly, especially since 2020. The way readers discover content, publishers market books and authors interact with audiences has changed dramatically. Short-video platforms, e-commerce and social media have all become central parts of the publishing ecosystem.
For Phoenix, digital transformation is not simply a “technology upgrade”. More importantly, it is about reshaping publishing formulas. That is why we developed our own AI publishing platform, Phoenix Zhiling, which we hope will eventually become core infrastructure supporting content production, marketing and publishing innovation.
How has Phoenix Zhiling developed?
It has been gradual and practical. When it first launched in 2023, we mainly experimented with basic AI functions such as intelligent Q&A and AI-generated images, testing whether editorial and marketing teams would genuinely use them in daily work.
We then gradually integrated the platform into real publishing workflows, including title acquisition research, intelligent proofreading, marketing asset generation, audiobook production and short-video creation. The platform has since gone through multiple iterations and now integrates several major Chinese AI models including DeepSeek and Kimi.
‘Over the next five to 10 years, AI will likely become the basic infrastructure for publishing in the same way the internet is today’
Most importantly, it is no longer simply a collection of tools – it is beginning to integrate directly into the publishing process itself. For example, our AI proofreading system reviewed more than 21 million Chinese characters in 2025, while the AI image-generation tools have been used over 24,000 times.
Phoenix Zhiling is now used not only across Phoenix’s own publishing divisions, but also by external publishers, universities and organisations beyond publishing, including theatre and finance-related sectors. This reflects how AI publishing tools are gradually evolving from editorial assistants into broader content infrastructure.
And has digital become more important now in China’s market?
Yes – the changes have been very significant. The biggest shift in China’s publishing market is not simply the growth of e-books or audiobooks, but the way readers now discover content. Many readers first encounter a book through platforms like RedNote, TikTok or podcasts rather than through bookshops.
As a result, some of China’s large publishing groups today increasingly operate as broader “content platforms” rather than traditional publishers. Phoenix has also developed a number of content platforms and sees digital business as a major driver of future growth. AI is accelerating this shift further because it dramatically lowers the cost of producing content across multiple formats. A single book today may also become a podcast, audiobook, short-video series, online course or social-media product.
Phoenix is also experimenting with combining publishing, education and cultural experience through projects such as “Walking Classrooms” – immersive study programmes linked to literature, museums and cultural travel. In many ways, this reflects a wider shift in China’s publishing industry from simply “making books” to creating broader content experiences.
On AI in particular, is any side of the business more advanced than others?
Educational and academic publishing are currently moving faster in AI adoption because these sectors naturally rely more heavily on structured content and efficiency.
Trade and literary publishing, by contrast, remain more cautious. Literature is not simply information – it also depends heavily on voice, rhythm, tone and emotional nuance, which still require strong human judgement.
The importance of human editors and authors will not disappear. In fact, it will become even more important. As AI becomes increasingly capable of generating content, the truly scarce qualities will be originality, cultural understanding and emotionally authentic expression.
Continues...
Read More
In the West, many publishing conversations around AI seem to revolve around taming it, at least in terms of copyright. Is there more of a desire for AI experimentation in China?
Compared with the West, China’s publishing industry is generally more open to experimenting with AI because many people believe it will fundamentally reshape the future content economy. That does not mean there are no concerns around copyright or ethics. One of the biggest current challenges is that responsibility and copyright boundaries around AI-generated content are still not entirely clear.
At Phoenix, we keep records of AI-generated processes to support future copyright tracking. At the same time, we believe the biggest challenge in practical AI implementation is not the model itself, but data quality, workflow integration and talent transformation.
Where do you see AI playing out most prominently in publishing: systemic or in content? Or maybe even staffing – perhaps the publishing professionals of the near future will need a different skill set?
In the short term, AI’s clearest value lies in improving production efficiency and supporting certain forms of content creation. Tasks such as title research, proofreading, marketing copy, meeting summaries and audiobook generation are already demonstrating clear time-saving advantages. Longer term, the bigger change may be the role of publishing professionals themselves. Future editors will not only need editorial judgement, but also an understanding of AI tools, data and content operations. The skill set required in publishing is likely to evolve significantly.
What are your views on experimenting with AI producing translation or original content? Do you consider certain subjects ripe for producing books in which content has been AI-generated?
In some standardised and information-heavy areas – such as educational materials, reference books and industry reports – AI translation is already becoming highly practical.
Literary translation, however, is very different. The hardest part of literary translation is often not the literal meaning, but tone, humour, rhythm and cultural context. A regional Chinese dialect, internet slang or a line of deadpan humour may land completely differently in English if the tone is even slightly wrong. At the moment, AI is better understood as a supporting tool rather than a replacement for literary translators.
Could you give me a sense of how you think AI will play out in the medium to longer term?
Over the next five to 10 years, AI will likely become the basic infrastructure for publishing in the same way the internet is today. Many repetitive tasks will become highly automated, publishing workflows will become more data-driven and intelligent, and content will move much more fluidly across formats.
And, as I said earlier, but it bears repeating: human editors and authors are still at the heart of publishing. As AI becomes increasingly capable of generating supporting content, the truly scarce qualities will be originality, cultural understanding and emotionally authentic expression.
Song Jishu will give the keynote at tomorrow’s (18th June) Beijing International Publishing Forum, which begins at 14.30, at Halls A and B, 1st Floor, China National Convention Center.
