The research team, led by Şükriye Kayhan-Aktürk of Muğla Sıtkı Koçman University together with Beyza Nur Dükar of Anadolu University and Şevket Özdemir, recruited 262 SLTs working across Türkiye. The survey ran between February and April 2025 and used a mixed-method, cross-sectional design, meaning it combined closed-ended quantitative questions with open-ended qualitative responses. Participants were reached through convenience sampling via the national professional association’s mailing list, social media channels, and clinical and educational settings. The questionnaire, developed by the authors from recent AI literature in the field, covered demographics, self-rated AI knowledge, patterns of use, perceptions of AI’s future, and ethical concerns, with Likert-type scales and open-ended items analyzed through descriptive statistics, non-parametric tests, and thematic analysis respectively.

The headline quantitative finding is striking: a substantial proportion of participants reported already using AI in professional practice, and usage varied significantly by years of experience. Less experienced therapists showed significantly higher levels of AI use across disorder groups and clinical stages, with the difference reaching statistical significance at p < .01. In other words, the newest members of the profession are not waiting for institutional permission. They are folding large language models and other generative tools into assessment preparation, therapy material creation, and academic writing, while their more senior colleagues remain largely on the sidelines.

What exactly are these therapists using AI for? The benefits participants highlighted cluster around efficiency. Material preparation, a notoriously time-consuming task in speech and language therapy, where clinicians must tailor picture cards, reading passages, and language exercises to individual clients, emerged as a major use case. Time management and academic tasks, such as drafting reports and research writing, also featured prominently. For a profession where a single session may require hours of bespoke preparation, tools that compress that labor carry obvious appeal, particularly for early-career therapists juggling heavy caseloads with the demands of building a practice.

Yet the qualitative responses, examined through thematic analysis, tell a more complicated story. Participants repeatedly framed AI as a supportive tool rather than a replacement for the therapist, emphasizing the irreplaceable importance of human interaction and clinical expertise. This is not a trivial distinction. Speech and language therapy is fundamentally relational work: therapy sessions depend on rapport, real-time observation of communication attempts, and nuanced adjustment to a client’s emotional state and communicative context. The therapists in the study seemed acutely aware that no current system can replicate the moment-to-moment clinical judgment that unfolds between a therapist and a child with autism or an adult recovering from a stroke.

Concerns about limited individualization surfaced repeatedly in the open-ended responses. AI-generated therapy materials, however convenient, risk being generic, and communication disorders are anything but generic. Each client presents a unique profile of strengths and deficits, and the study’s participants worried that over-reliance on AI could flatten the personalization that effective therapy demands. There were also worries about potential impacts on professional roles, a familiar anxiety in the AI era but one with particular weight in a field where the therapist’s own communication is the primary therapeutic instrument.

Ethics emerged as perhaps the deepest fault line. Participants expressed concerns about ethical issues and, crucially, about their own limited ethical confidence, a self-assessment that the authors flag as a key limitation of the current workforce. The study’s conclusions are blunt on this point: while AI shows considerable potential, its safe and effective integration requires improved AI literacy, clear ethical and professional guidelines, and stronger evidence-based practices. In a profession dealing with sensitive clinical data, including recordings of clients’ speech, ethical confidence is not an abstract virtue. Questions of data privacy, consent, and the provenance of AI-generated clinical recommendations remain unresolved, and the therapists in this study clearly sensed that the rules of engagement have not been written.

Limitations in knowledge and proficiency also ran through the findings. Even among enthusiastic users, self-rated AI knowledge was uneven, and the authors note that limitations remain in knowledge, proficiency, and ethical confidence across the sample. This gap between use and understanding is a familiar pattern in healthcare AI research: clinicians adopt tools faster than formal training can keep up, creating a workforce that is experimentally competent but formally unprepared. The study participants generally viewed AI as useful and supportive, but that positive perception coexisted with an admission that they lacked the depth of knowledge needed to evaluate outputs critically or troubleshoot failures.

The methodological choices deserve attention for what they reveal about the state of the evidence. The authors acknowledge that evidence on SLT perspectives toward AI remains limited despite rapid developments, which is precisely why they designed this questionnaire from the recent AI literature in the field rather than borrowing instruments from other professions. The use of non-parametric statistics reflects the distribution of the survey data, and the decision to pair quantitative scales with thematic analysis of open-ended responses allows the study to capture both the pattern of adoption and the texture of professional anxiety. The study was approved by the Committee on Research and Publication Ethics in Health Sciences of Anadolu University, with informed consent obtained from all participants, and it was presented as an oral presentation at the 9th National Interdisciplinary Congress on Early Childhood Intervention in Kuşadası in April 2026.

What does this mean for the future of the profession? The study’s participants generally view AI as a useful and supportive tool, and the generational gradient in adoption suggests that AI use will only deepen as less experienced therapists advance through their careers. But the authors are careful not to read this as a green light. The safe and effective integration of AI into speech and language therapy, they conclude, depends on three pillars: improved AI literacy so therapists can understand what these systems actually do, clear ethical and professional guidelines so clinicians know where the boundaries lie, and stronger evidence-based practices so adoption follows demonstration of clinical benefit rather than novelty. Until those pillars are in place, the promise of AI in this field will remain balanced, precariously, against risks that the profession itself is only beginning to articulate. For a discipline built on the science of human communication, the study suggests that the most important conversation about AI is still the one happening between therapists, not between a therapist and a machine.

Subject of Research: Balancing promise and risk: artificial intelligence use, knowledge, and perceptions among speech and language therapists in Türkiye

Article Title: Balancing promise and risk: artificial intelligence use, knowledge, and perceptions among speech and language therapists in Türkiye

Article References: Kayhan-Aktürk, Ş., Dükar, B. N., & Özdemir, Ş. (2026). Balancing promise and risk: artificial intelligence use, knowledge, and perceptions among speech and language therapists in Türkiye. BMC Medical Education. https://doi.org/10.1186/s12909-026-10426-0

Image Credits: AI Generated

DOI: 10.1186/s12909-026-10426-0

Keywords: Balancing, promise, risk, artificial, intelligence, knowledge, perceptions, speech, language, therapists, rkiye, scientific research