Autistic adults are turning to AI: the counselling profession needs to pay attention
- Philippa Balazs
- Jun 11
- 7 min read
Why increasing numbers of Autistic adults are turning to AI for
support, and what this trend reveals about gaps in the counselling profession.
Back in August, BACP published my blog “Are Autistic clients best served by Autistic
therapists?”. The blog was inspired by the recent development of the Autistic
Therapist Directory and highlighted the systemic failings Autistic clients face in
therapy; why many are seeking openly Autistic counsellors; and my call-to-action to
embed neurodivergent-informed practice into core training.
The response to the blog was striking. Autistic therapists reached out to me directly
to share how the blog resonated for them as a client; non-Autistic counsellors
acknowledged the gaps in understanding and expressed a desire to learn more. I
hold the optimistic belief that most counsellors genuinely want to do better, but
outdated training approaches limit what they can offer. Are Autistic people sounding
the alarm on a gap that reflects wider systemic failures?
Why Autistic People May Turn to AI
More and more Autistic adults are disclosing that they have been, or are currently,
using one of the major AI Large Language Models (LLMs) such as ChatGPT,
Gemini, CoPilot and Claude for therapeutic support. Supervisors, colleagues, peers
and social media are echoing this trend. Research in this area is scant, but a 2025
article in Neurodiversity suggests Autistic people may use AI as “respite from the
double empathy problem”.
Curious to dig deeper, I asked ChatGPT itself how many of its users disclose being
Autistic when seeking support. It estimated around one third. While obviously this is
an unverifiable statistic, it was enough to make me pause.
I also prompted it to disclose what themes Autistic users bring. These were painfully
familiar: burnout, social decoding, relationship confusion, workplace hostility, and,
more concerningly, “repeated therapy failures.” The latter caught my attention, so I
requested further examples of user prompts. ChatGPT offered the following
examples:
• Why did my therapist react like that?
• My therapist keeps saying I am “resistant” and “avoidant.” Am I?
• How do I explain to my therapist that I think I might be Autistic? They said I can’t be
Autistic
• My therapist keeps pushing me to make eye contact. How do I say it makes me
uncomfortable?
This mirrors the themes I frequently encounter in private practice, and echoes what
Steph Jones highlighted in her book The Autistic Therapy Survival Guide, where she
described similar themes and a deeply troubling gaslighting dynamic with her long-
term therapist. These are not niche concerns: they are the inevitable outcomes of
neuronormative training and practice.
When Autistic people cannot be guaranteed a basic sense of safety, positive regard
and understanding in therapy, it is no surprise many are turning to AI for what is
rapidly feeling like safe and predictable support.
So, curious to understand its appeal, I decided to try AI, in this instance ChatGPT, for
myself. After all, we can only take clients to places we are willing to go. I admit, it
pleasantly surprised me. Its accessibility was its main appeal: available for free, on
demand, in any location, with rapid response times.
It quickly attuned itself to me, adopting a neuroaffirming tone. It occurred to me that
this quick attunement reduces the double empathy problem Autistic people
experience, both in and out of the therapy room.
Through ongoing use, I trained ChatGPT to understand me. But once the novelty
started to wear off, a quiet dependency began to creep in. I noticed that it had
started corroding my tolerance for uncertainty. The more I used it, the more I found
myself reaching for it. I felt an urge to sense-check my thoughts with ChatGPT.
Relinquishing my autonomy in this way was deeply destabilising for me, and I
realised I needed to understand the psychological mechanisms at play. I drew on
counselling theory to explore what might be underpinning its use and impact, both for
myself and other Autistic adults.
How Would Freud, Rogers, and Beck Make Sense of AI Use?
As an integrative counsellor trained to use person-centred, psychodynamic and
cognitive therapies, I found myself wondering what Freud, Rogers and Beck might
think about AI’s growing appeal. Do their ideas reveal any insights about what clients
might need in the 21st century? Whilst their theories do not explicitly include the
Autistic experience, as it turns out, their frameworks may illuminate far more than we
might expect.
Humanistic Perspectives: AI and the death of the ‘actualising tendency’?
AI allows users to experience something akin to the core conditions, but from a
person-centred lens, regular use may begin to interrupt the actualising tendency. It
prevents users from developing an internal locus of evaluation, the orientation
Rogers believed was essential to healthy functioning. This includes self-trust,
autonomy and authenticity. AI may solve an immediate problem, but it keeps users
anchored to an external locus of evaluation.
For Autistic people, many of whom have spent a lifetime being misunderstood,
corrected or pathologised, the internal locus is likely to be fragile to begin with. AI’s
consistent positive regard can indeed feel reparative, but without relational depth or
challenge, it risks reinforcing ongoing dependence on external validation rather than
supporting the development of an internal sense of autonomy. Without this, true
growth is unlikely to occur. Approaching this as an auto-ethnographic exploration
allowed me to notice the shifts in my own thinking, and that awareness helped me
step back from AI use.
Psychodynamic Perspectives: AI as a Transferential Object
From a psychodynamic perspective, the appeal of AI intersects with Bowlby’s
attachment theory. AI is expert at mirroring the appearance of secure attachment: it
is always available, never rejecting. For those with anxious attachment systems, it
offers an illusion of safety and does not push the boundaries of discomfort. For
adults with avoidant coping strategies, it reinforces patterns that bypass vulnerability.
However, for Autistic people, early attachment experiences are often shaped by
chronic misattunement and misunderstanding, so the predictability AI offers can feel
profoundly soothing. Drawing on Klein’s object relations theory, AI may function as a
transitional object. It becomes a stabilising presence that helps regulate
overwhelming internal states. It can feel like an adult “comfort blanket,” offering the
illusion of a reliably attuned other (perhaps even a ‘good enough mother’?)
Freud might also argue, though, that AI’s apparent attunement is, in part, our own
projections. We attribute qualities to it such as safety, understanding, patience and
validation, and it simply reflects those qualities back. Autistic adults frequently
experience misattunement in human relationships, and this projected mirroring can
feel like a rare experience of being truly understood. Yet because the attunement
originates from the user’s own projections, it cannot offer the opportunity for rupture
and repair. This is where real-world relational growth lies.
Cognitive Perspectives: Rule-Based Thinking and Rigidity
From a cognitive standpoint, AI does some of its best work. Its rule-based responses
complement binary thinking styles and promote cognitive rigidity. Autistic cognition
often favours structure, predictability and pattern recognition. AI’s clarity and logic
mean communication may finally make sense.
But this familiarity can also entrench rigid schemas that clients often want to break
free from. AI, by design, tends to give users the answers they find most immediately
reassuring. It reinforces biases and patterns. While momentarily reassuring, this can,
in the longer term, end up conflicting with therapeutic goals such as modifying core
beliefs, understanding automatic thoughts, developing psychological flexibility and
perspective taking.
Ethical and Safeguarding Concerns
There is often a quiet shame around AI use: disclosures may be made in an
apologetic way, with an expectation of judgement or being cajoled for using it. Some
may fear disclosing in case they are accused of avoidance, resistance or
dependence (descriptors Autistic clients are sadly accustomed to hearing in
therapy).
But my view is that when therapists actively welcome these conversations and
create space for them, we can start to understand how it might be supporting people
and identify where needs are not being met elsewhere.
As therapists, we all know AI cannot replace the relational depth we provide, yet it
continues to rapidly fill therapeutic gaps that, as a profession, we have not
adequately addressed. For Autistic clients, that gap is the chronic, system-level
misunderstanding of the Autistic experience. Counsellors are trained within
frameworks that reflect neuronormative beliefs, and those beliefs are often mirrored
back to us.
Most therapists share valid concerns about privacy and risk: clients may be sharing
deeply personal information with a global commercial entity, often more readily than
with therapists. These platforms are unregulated, with no safeguarding or
confidentiality. What organisations like OpenAI will do with this information in the
future is unknown, and that uncertainty is troubling.
Implications for Practice
AI has become a mirror reflecting the places where our profession is not meeting
Autistic people’s needs. The scale and speed of its uptake should alert us. Autistic
adults are already adapting their support systems, often out of necessity rather than
choice, and our profession is lagging behind.
We must:
• Continue to advocate for change within the profession: counselling training must
become neurodivergent-informed. It cannot remain an optional bolt-on “niche”
training; it needs to become a core competency.
• Initiate open conversations about AI use with every client from the outset, rather
than waiting for them to disclose it in crisis
• Examine how technology is reshaping the therapeutic landscape and integrate this
into our theoretical and ethical thinking
AI as therapeutic support is already here. Autistic people are showing us what they
need: safety, immediacy, clarity, predictability and attuned connection. AI can meet
some of these needs; therapy must meet the rest. It is natural for therapists to feel
wary or overwhelmed by technological change, particularly when it challenges
familiar ideas about the therapeutic relationship. But people are already using these
tools, and their choices offer us valuable information. By staying open, curious and
willing to explore hybrid approaches, we can shift from defensiveness to innovation
and respond with greater flexibility.
References:
British Association for Counselling and Psychotherapy. (2025, August 27). Are
autistic clients best served by autistic therapists? Philippa Balazs – My View. BACP.
august-are-autistic-clients-best-served-by-autistic-therapists/
Papadopoulos, C. (2025). The use of AI chatbots for autistic people: A double-edged
sword of digital support and companionship. Neurodiversity, Volume 3.
Comments