Within the ecosystem of commonly-used or new language learning applications, I have been witnessing a trend. Duolingo, a major language learning application for instance, has dedicated fully to pivoting towards the use of generative artificial intelligence (GAI) in its course design, and replacing human translation teams in favour of GAI. Similarly, on the App Store, I have been noticing some language learning applications popping up featuring AI-generated voiceovers, sprites, and perhaps even translations. Some of these have been sponsoring YouTube channels as well, using influencer marketing to push their usually-unknown brand. Thus, I think there is a need to investigate the use of GAI in language learning a lot further, and assess if GAI-integration into language learning methods is really necessary or effective as some might purport.

I have covered the use of GAI in academia, and once on its use in constructed languages, mostly putting them in a strongly negative light. And so, I want to dedicate a little series of essays discussing the use, attitudes, and discourse surrounding the use of GAI in language learning, and perhaps biting the bullet and trying out first-hand, one of these language learning methods as well.

The main discourse surrounding the use of GAI in language learning applications has been overwhelmingly negative, and understandably so. After all, there have been publications discussing the erosion of critical thinking through the overreliance on GAI to fact check certain things. You may have noticed the general replies on Twitter quoting the handle @grok, @PerplexityAI, or some language model of the sort asking if a certain thing is true, if a certain event happened, or to just explain the entire tweet. Sometimes, bots would just jump in and explain the entire tweet without even having a reply prompt it at all. This, combined with the threat GAI presents to creatives and academia, as well as human thought and social connection, the arguments against GAI especially in these contexts would be extremely strong.

Going back to the Duolingo example, there has been some use of AI to tailor lessons to the user, but each exercise was still created by humans. How this AI worked was to observe how the user responds to questions, and assessing the user’s strengths and weaknesses exhibited in each lesson. Using these data, the AI would determine the difficulty level of exercises to be shown to the user. This sort of AI could be compared to social media algorithms as they decide what to show to the user, based on the user’s behaviour (and in Duolingo’s case, performance).

However, in 2023, Duolingo announced that they are removing humans from the equation when it comes to designing exercises, moving instead to prompt large language models to push these exercises. They cited ‘speed’, ‘productivity’, and ‘convenience’ as motivations for this pivot, which did not go too well with many users. In late April of 2025, they also announced that Duolingo has ‘more than doubled’ the number of language courses offered on their platform thanks to the use of GAI. Now of course, this has been met with much backlash, such as ‘quality over quantity’ arguments, quality concerns, and the use of GAI in what is fundamentally a human process, language teaching and language acquisition.

As such, I have been thinking of several topics I want to cover in a series, perhaps with some intermissions to break up the streak. I plan to space this over a couple of months or so, and capping it off with a little reflection of my experience taking on a language learning application, and giving it a review. These topics include the impact of GAI in the language landscape of academia, the effectiveness of GAI in language acquisition, and we’ll also take a look into some pedagogy or education studies that pertain to this field of language learning. As we progress into the mid 2020s, there would have been a growing body of research discussing these topics, and perhaps more so down the line as we learn about the long-term impacts.

To clarify, I have been generally biased against the use of GAI, and made it rather clear of these biases, but I want to put these aside when I bite the bullet in picking up a language learning application with heavy GAI use, and conveying my honest opinions about these applications. And so, I hope you look forward to this new series of experiences and readings that I will be putting up over the next couple months, and share your thoughts on these topics as well.