#1 Victorian uni for graduate employment1
#1 in the world for sport science2
#1 Victorian uni for course satisfaction3
NEXT UP ON this.
There are some beautiful benefits to learning another language.
Being bilingual or multilingual can bring us closer to other cultures when we travel, allowing us to experience life beyond the tourist trails. It can open up job or study opportunities all around the world in fascinating fields like diplomacy, international finance or education. Speaking a second language might even improve our brain function, potentially helping us process information and learn better.
There’s a problem, though: learning a language can be tough. Depending on our own native language and the one we’re trying to learn, a new language could take thousands of hours of study to master.
Given the heavy time commitment, some are turning to more contemporary language-learning techniques, using apps like Duolingo and other AI-powered tools. Where AI is concerned, there’s a lot of educational optimism – but also cause for scepticism.
So, can AI be used to learn a language? As it turns out, it’s a complex question without a clear answer. With the help of Jack Walton, Lecturer at Deakin University’s Centre for Research Assessment and Digital Learning, let’s explore why.
‘While AI has had renewed press since ‘generative’ models (ChatGPT and its ilk) have become widely available, AI systems are by no means new,’ says Walton.
The point is a good place to start, because the technology of AI is complex, and older than we might think.
AI has very primitive and speculative origins in the 18th century, but it’s the recent wave of generative AI models that are having an impact on language learning. The Australian eSafety Commissioner defines generative AI as being able to ‘create new outputs, instead of just making predictions and classifications like other machine learning systems.’ That means that generative AI can produce text, video, images, sound, and more.
Generative AI gets this knowledge from language models (LLMs), which ‘enable the processing of natural language through algorithms’.
Modern generative AI models like ChatGPT, for instance, can interact in human-like ways, and are designed for tasks like text generation and language understanding.
In this sense, generative AI and natural language processing technologies are very much based on the patterns of human speech and language.
It only takes a quick search to see that AI-based language learning is everywhere in 2025.
Popular apps like Duolingo have recently embraced AI for language teaching, scaling down human staff in favour of AI. Other language apps have either adopted AI or been built specifically for the technology, like Babbel, Gliglish and Langua. These apps often use real-time chat bots (either text or voice/audio) for instruction and practice. Gliglish, for instance, lets users converse with a CGI avatar to practice language learning.
‘The general answer to this line of questioning tends to be “it depends” and “we don’t know yet,” says Walton. ‘It’s a rapidly moving conversation in which many are hoping to find certainties that by and large don’t exist (or if they do, we don’t recognise them yet).’
While it’s difficult to answer with certainty at this stage, some research suggests that AI tools may benefit those learning English as a foreign language. Those benefits include:
Other studies have suggested that AI language learning tools can create helpful cultural contexts for students, helping to ‘shorten the social distance between the two languages, and thus promote [second language] acquisition.’
When it comes to generative AI, the potential benefits seem infinite – at least in theory.
Some studies suggest that AI-based language learning tools do have the potential to help with language learning, but much of that potential depends on the individual and tools being used. In other words, the answer is as Walton suggests: it depends.
‘It is reasonable to expect, for instance, that AI tools will be both more and less effective than traditional approaches to language learning,’ he says. ‘The examination worth doing is to wonder what the benefits and trade-offs are, for particular individuals, in particular situations (and, who has the capability to recognise those benefits and trade-offs — they aren’t always obvious!).’
Walton also raises a valid point about the power of the technology. Of course, AI has potential for helping us learn languages – but potential doesn’t always result in a satisfying outcome.
‘The difference between what the technology can do in theory is often a world away from what it does in practice,’ says Walton. ‘Could AI be a resource for language learning? Certainly! But I could not at this moment speak with confidence about any particular strategies for doing this that would necessarily be generalisable or replicable in terms of the learning achieved. Explaining how it might work would be as difficult as explaining theories of learning themselves. Major influences on learning will always include the characteristics of the learner themselves, as well as the contexts and situations in which they find themselves.’
At this stage, there’s no satisfying answer to whether or not you can use AI to learn a language. As Walton has pointed out, AI is a broad technology with many unknowns, and much of any success in learning will depend on individual tools and learners themselves. How the use of AI in language learning evolves, it seems, is still very much up in the air.
‘We don’t yet have a reliably durable picture as to how uptake of AI tools as pedagogic resources will shape learning,’ Walton says. ‘Sweeping generalisations are unlikely to serve us well in this arena: what will be required is nuance and care, and a willingness to be flexible and iterative in our concept of what AI does with, for, and to learning.’
Research Fellow, Office of the Deputy Vice Chancellor Academic/Centre for Research Assessment and Digital Learning/Research for Educational Impact, Deakin University
