If AI can perform some of the work of a critical friend, does the relationship need to be a friendship for that work to matter?
Kenan Dikilitaş and Thomas Farrell place one word in quotation marks.
Their article is called Using ChatGPT as a “Friend” in reflective critical friendships.
The punctuation is useful.
It allows the possibility of valuable interaction without quietly deciding what kind of relationship the interaction has become.
For language teachers, the appeal is not difficult to see.
A human colleague may be unavailable.
They may not have time.
The question may arrive late at night.
The teacher may want to rehearse an uncertainty before taking it to somebody else.
ChatGPT can respond immediately.
It can ask structured questions.
It can offer different ways of framing a problem.
It can generate alternatives and provocations.
And because the challenge comes from a machine, some users may experience it as less personally threatening.
Those are real affordances.
Dikilitaş and Farrell do not treat them as proof of friendship.
Their argument keeps returning to what is missing or changed.
Human critical friendship is situated in shared professional worlds. Colleagues bring histories, memories, embodied experiences, emotions, cultural knowledge and practical judgement to the exchange.
The authors use the idea of life capital to describe part of this difference.
A human colleague does not only know facts about teaching.
They have taught.
They have misread rooms.
Changed plans.
Been surprised by learners.
Regretted decisions.
Remembered a previous conversation differently from you.
Their response can draw on a life that did not begin when you asked the question.
ChatGPT does not bring that kind of life into the exchange.
It can generate language about experience.
It does not thereby acquire one.
The distinction is important because a fluent reflective conversation can make the absence easy to forget.
ChatGPT can sound patient.
It can sound curious.
It can maintain a supportive tone.
It can challenge an assumption when asked.
But the authors caution that its contribution remains strongly dependent on what the user supplies and how the interaction is prompted.
The reflective partner is partly being constructed by the person seeking reflection.
That changes the meaning of challenge.
A human critical friend can surprise us because they have their own commitments, observations and history.
An AI can surprise us too—but surprise alone does not establish that it arrived from another lived position.
And yet it would be equally careless to conclude that nothing useful has happened.
A study by Arefian, Çomoğlu and Dikilitaş followed four novice and four experienced Iranian EFL teachers using ChatGPT in reflective practice.
The teachers worked with ChatGPT individually and also reflected collaboratively within a community of practice.
The interesting finding for this journey is not that AI replaced the community.
It did not.
Individual reflection with ChatGPT could act as a preparatory step for the later human conversation.
Something happened before the meeting.
A teacher could articulate a problem, encounter possible interpretations and arrive at the community with more material to think with.
The AI exchange mattered precisely because it did not have to become the whole relationship.
That changes the question again.
Perhaps the usefulness of AI as a critical friend does not depend on successfully reproducing human friendship.
Perhaps some of the value lies in making certain reflective functions more available:
a question before the meeting;
an alternative when nobody else is present;
a rehearsal of an argument;
a prompt that makes an assumption visible;
a first pass that gives the human conversation somewhere else to begin.
That is a smaller claim than friendship.
It may also be a more useful one.
Dikilitaş and Farrell ultimately recommend combining AI-mediated reflection with human interlocutors. ChatGPT can supplement the process, but they do not treat its speed, breadth or articulate responses as substitutes for trust, lived experience and human contextual judgement.
Function and friendship can therefore be separated without making either irrelevant.
A relationship does not have to be human friendship for an exchange to improve thinking.
But if we call every useful exchange a friendship, we lose the distinction we have spent this journey trying to preserve.
So perhaps the next step is to stop asking about friendship for a moment.
Because another body of work reaches similar territory without needing the word at all.
What if challenge, dialogue and alternative perspective can be designed without calling the relationship friendship?
References
Dikilitaş, K. and Farrell, T.S.C. (2026; first published online 2025) ‘Using ChatGPT as a “Friend” in reflective critical friendships’, ELT Journal, 80(1), pp. 117–125. doi:10.1093/elt/ccaf053. Open-access article.
Arefian, M.H., Çomoğlu, I. and Dikilitaş, K. (2026; first published online 2024) ‘Understanding EFL teachers’ experiences of ChatGPT-driven collaborative reflective practice through a community of practice lens’, Innovation in Language Learning and Teaching, 20(2), pp. 318–333. doi:10.1080/17501229.2024.2412769. Open-access article.