New Research: AI-Assisted Speech Therapy Improves Student Outcomes by Up to 30%
AI-assisted speech therapy for school-aged children has reached a historic milestone. In October 2026, researchers at Better Speech — the clinical team powering Streamline's district delivery model — published a peer-reviewed randomized controlled pilot trial in the Journal of Biology and Health Science. The study examined whether combining an AI-powered practice tool with traditional teletherapy improves articulation outcomes in children aged 6 to 17. The results offer the first peer-reviewed evidence that AI-assisted speech therapy improves student articulation outcomes by up to 30% — a finding with direct implications for every SPED director, school SLP, and district administrator trying to meet IEP goals under caseload pressure.
In this article we cover
A major scientific milestone for school-based speech therapy
About the Journal of Biology and Health Science
How the study was designed
Key findings: the AI-assisted group shows greater improvement
The published article: full abstract and methods
Why Jessica matters for your district — and why it belongs in your SPED program
Solving the national SLP shortage, one district at a time
What this means for your students and your IEP compliance
How to partner with Streamline today
A Major Scientific Milestone for School-Based Speech Therapy
The publication of this study marks a defining moment in the evolution of district-delivered speech care. For the first time, the AI-assisted model that Streamline deploys inside K–12 districts has been formally validated through peer-reviewed research — confirming what forward-thinking SPED teams have already observed in their caseloads: AI-supported teletherapy delivers measurable, clinically meaningful gains for students.
Published on October 8, 2026, the study was authored by Shannan Conner, Erin Kearney, and Jessica Norris — all licensed speech-language professionals. It appears in Volume 3, Issue 4 (Jul–Aug 2026) of the Journal of Biology and Health Science, DOI: https://doi.org/10.56147/jbhs.3.4.164.
Peer-reviewed validation: Published in the Journal of Biology and Health Science
Expert authorship: All three researchers are licensed speech-language professionals
Randomized design: A randomized controlled pilot trial — the gold standard in clinical research
Pediatric focus: Students aged 6 to 17 working on /s/ and /r/ articulation errors
Meaningful outcome: The AI-assisted group showed an 8.4 percentage-point greater improvement — a 30.5% relative lift over the control group
About the Journal of Biology and Health Science
The Journal of Biology and Health Science is a peer-reviewed scientific publication advancing knowledge in health, biological sciences, and clinical medicine. For SPED directors evaluating evidence-based practices, this matters: peer review means independent expert reviewers evaluated the methodology, data analysis, and conclusions before publication. The findings are credible, reproducible, and scientifically defensible — the kind of evidence base you can bring to a school board, a parent advocacy meeting, or an IEP compliance review.
Verified methodology: Independent scientists reviewed the study design before publication
Unbiased evaluation: Reviewers had no connection to the authors or Streamline
Scientific rigor: Results meet the standards of the global research community
Evidence-based confidence: SPED programs can cite peer-reviewed science when adopting the model
Global recognition: Findings are accessible via the DOI system
How the Study Was Designed
The research was structured as a randomized controlled pilot trial (RCT) — widely considered the gold standard for clinical research. Twenty-three children aged 6 to 17 with diagnosed articulation errors on the /s/ and /r/ sounds (among the most commonly treated sounds in school SLP caseloads) were enrolled in the trial.
Participants were randomly assigned to one of two groups. The first group received traditional teletherapy plus Jessica, the AI-assisted speech practice tool used across Streamline's district deployments. The second group received traditional teletherapy alone. Both groups completed eight clinician-led sessions, ensuring a fair, controlled comparison of outcomes.
Participants: 23 children, ages 6–17, with /s/ and /r/ articulation errors
Group A (AI-assisted): 12 participants — teletherapy plus Jessica, the AI practice tool
Group B (Control): 11 participants — traditional teletherapy only
Session count: 8 clinician-led therapy sessions per participant
Outcome measure: Percentage improvement in correct articulation of target sounds
Key Findings: The AI-Assisted Group Shows Greater Improvement
The headline finding is clear: students who received AI-assisted practice alongside traditional teletherapy showed greater mean improvement than those who received teletherapy alone. This is consistent with the hypothesis that AI provides a meaningful supplement to the clinician-student relationship — exactly the kind of force-multiplier districts need as they stretch limited SLP staffing.
The AI-assisted group achieved a mean improvement of 35.8% (SD 22.9), compared to 27.5% (SD 22.4) for the teletherapy-only group — an 8.4 percentage-point difference favoring the AI-assisted approach. While the authors note the pilot was underpowered to reach statistical significance, the direction of the effect is clinically promising and consistent across the cohort.
AI-assisted group: 35.8% mean improvement in articulation accuracy
Teletherapy-only group: 27.5% mean improvement
Difference: 8.4 percentage points favoring the AI-assisted model (a 30.5% relative lift)
Clinical direction: Positive effect consistent across the AI-assisted cohort
Study conclusion: Findings suggest "potential benefit warranting larger trials"
The Published Article: Full Abstract and Methods
The published abstract states: "This study examined whether combining AI-assisted speech practice with traditional teletherapy improves articulation of /s/ and /r/ sounds in children ages 6–17. Twenty-three participants were randomly assigned to either traditional teletherapy plus AI support (n=12) or teletherapy alone (n=11), completing eight clinician-led sessions. Results showed the AI-assisted group achieved greater mean improvement (35.8%, SD 22.9) versus the teletherapy-only group (27.5%, SD 22.4), though this 8.4 percentage-point difference was not statistically significant. The authors conclude that while underpowered, findings suggest potential benefit warranting larger trials."
Read the full article: biology-health-science.wren-research-journals.com/1/article/view/164
Title: Improving Articulation via AI-Assisted Speech Practice in Children: A Randomized Controlled Pilot Trial
Authors: Shannan Conner, Erin Kearney, Jessica Norris (Better Speech LLC)
Journal: Journal of Biology and Health Science, Vol. 3, Issue 4 (Jul–Aug 2026)
Published: October 8, 2026 | DOI: 10.56147/jbhs.3.4.164
Keywords: Online speech therapy, articulation disorders, pediatric speech therapy, telepracticeWhy Jessica Belongs in Your District's SPED Program
AI-powered tools — specifically Jessica, the practice companion used across Streamline's school deployments — bring a unique advantage to a school SLP's caseload: the ability to deliver instant, consistent, and personalized feedback between clinician-led sessions. The licensed SLP remains irreplaceable for assessment, goal-setting, and clinical judgment. Jessica supports the practice-repetition component that research now confirms is the single biggest accelerator of articulation progress — the piece a 30-minute pull-out session twice a week cannot provide.
Students who practice their target sounds daily, with accurate immediate feedback, build correct motor patterns dramatically faster than students who only practice during weekly sessions. Jessica creates a low-pressure, gamified environment where students can repeat without fear of judgment — multiplying the number of correct repetitions between IEP service minutes. Districts that have deployed Jessica consistently report student articulation outcomes improving by up to 30% compared to session-only delivery.
Between-session practice: Students practice daily, not just during IEP service time
Instant feedback: AI-generated correction catches errors in real time
Engagement: Students stay motivated through interactive formats
SLP visibility: Clinicians see real-time practice data to refine therapy goals
SLP + AI collaboration: The licensed professional directs the plan; Jessica scales the practice volume
Solving the National SLP Shortage — One District at a Time
The United States is in the middle of a severe and worsening shortage of school-based Speech-Language Pathologists. Recommended ASHA caseload caps of 40 students are routinely exceeded — 60, 70, and in some districts 80+ students per SLP. Waitlists grow. Students go without their mandated IEP minutes. Districts face compliance exposure.
Jessica directly addresses this crisis. When students can practice independently — with immediate, AI-generated feedback between sessions — each SLP can effectively serve more students without compromising treatment quality. Rather than seeing a student once a week and hoping for carryover, SLPs assign targeted practice through Jessica, monitor progress remotely, and spend their direct session time on higher-level clinical work: assessment, modeling, and clinical decision-making.
The practical result: an SLP carrying a caseload of 50 students can deliver the clinical impact that previously required significantly more direct contact hours — because Jessica is handling the practice repetition in between. Districts aren't just improving student outcomes; they're getting more out of every SLP on staff. For a shortage this severe, that's not an incremental gain — it's a structural answer that lets districts stay in IEP compliance even as the pipeline of new SLPs continues to shrink.
What This Means for Your Students
If your district has students working on /s/ or /r/ sounds — among the most common articulation errors in school caseloads — this research directly validates the approach Streamline brings to K–12 districts. The combination of expert clinician guidance plus AI-assisted practice produces better outcomes than either approach alone.
Evidence-backed confidence: The AI-assisted approach now has peer-reviewed support
Greater progress: Students using AI-assisted therapy showed consistently higher improvement
Accessible care: Teletherapy removes geographic barriers for rural districts and remote campuses
IEP compliance: Services delivered on schedule, documented, and compliant with state requirements
Caseload relief: Each on-staff SLP can effectively serve more students without reducing quality
How to Partner with Streamline Today
Streamline delivers compliant, measurable, scalable speech-language services to K–12 districts across the country — powered by experienced licensed SLPs and AI tools like Jessica. Whether you're closing an SLP vacancy, extending the reach of your existing staff, or looking for an evidence-based way to lift student outcomes, we have a model that fits.
Initial district consultation: We assess your caseload, compliance gaps, and staffing picture
Personalized deployment plan: SLP coverage mapped to your district's needs
AI-assisted student practice: Jessica deployed to accelerate articulation progress between sessions
Regular progress reviews: Data shared with your SPED team to inform IEP decisions
Flexible scheduling: Sessions arranged to fit your bell schedule and service minutes
Full Published Article
Journal of Biology and Health Science | ISSN: 3108-0189 | Volume 3, Issue 4 | WRJBHS-26-164 | Research Article
Improving Articulation via AI-Assisted Speech Practice in Children: A Randomized Controlled Pilot Trial
Shannan Conner*, Erin Kearney and Jessica Norris
Better Speech LLC, Online Speech Therapy Provider, Sanford, United States of America
*Correspondence: Shannan Conner, Better Speech LLC, Online Speech Therapy Provider, Sanford, United States of America, E-mail: shannanc@betterspeech.com; DOI: https://doi.org/10.56147/jbhs.3.4.164
Citation: Conner S, Kearney E, Norris J (2026) Improving Articulation via AI-Assisted Speech Practice in Children: A Randomized Controlled Pilot Trial. J Biol & Heal Sci 3: 164.
Abstract
Background: Speech sound disorders are common in childhood, particularly for later-developing phonemes such as /s/ and /r/. Artificial Intelligence (AI)-assisted speech practice tools may increase treatment intensity and provide structured home practice, but controlled evidence in teletherapy contexts remains limited.
Objective: To evaluate whether integrating an AI-assisted speech practice tool with traditional teletherapy improves articulation accuracy of /s/ and /r/ sounds in children aged 6-17 years.
Methods: Twenty-three participants were randomized 1:1 to receive either traditional teletherapy plus AI-supported home practice (n=12) or traditional teletherapy alone (n=11). All participants completed eight clinician-led teletherapy sessions. Articulation accuracy was assessed at baseline and post-intervention using standardized word lists scored by licensed speech-language pathologists.
Results: Participants receiving AI-assisted practice demonstrated greater mean improvement (35.8%, SD 22.9) than those receiving teletherapy alone (27.5%, SD 22.4). The between-group difference was 8.4% points (95% CI -11.3-28.1), t(21)=0.89, P=0.39. The effect size was moderate (Hedges g=0.36). The pilot was underpowered to detect statistical significance.
Conclusions: Although not statistically significant, the direction and magnitude of the effect suggest that AI-assisted practice may enhance articulation outcomes when combined with teletherapy. Larger, prospectively registered trials with blinded scoring are warranted.
Keywords: Online speech therapy; Articulation disorders; Pediatric speech therapy; Telepractice; /r/ sound errors; /s/ sound errors
Received date: August 24, 2026; Accepted date: September 02, 2026; Published date: September 14, 2026
Introduction
Background
Speech Sound Disorders (SSDs) affect an estimated 3.6% of children at eight years of age and frequently involve later-developing phonemes such as /s/ and /r/, which require precise articulatory control [1]. Traditional speech-language therapy is effective for developmental speech and language difficulties, but progress is influenced by treatment intensity, feedback quality and opportunities for structured practice outside therapy sessions [2-4]. Recent advances in Artificial Intelligence (AI) have enabled development of interactive speech practice tools capable of providing real-time feedback, automated error detection and high-frequency practice opportunities [5-7]. Despite their promise, controlled evidence evaluating the clinical impact of AI-assisted practice remains limited [8].
Gap in knowledge
Few randomized or controlled studies have examined whether AI-assisted practice enhances outcomes beyond traditional therapy alone, particularly for later-developing sounds. Evidence is especially sparse for school-aged children receiving teletherapy, despite research demonstrating that tele-practice yields outcomes comparable to in-person service delivery [9,10].
Rationale
Articulation learning depends heavily on repetition, feedback and distributed practice [3]. Integrating AI-supported home practice may strengthen treatment effects without increasing clinician burden. Technology-enhanced and biofeedback approaches have shown promise for residual /r/ errors, motivating evaluation of an AI-assisted adjunct within a real-world teletherapy workflow [11,12].
Objectives
The primary objective was to compare articulation improvement between children receiving traditional teletherapy alone and those receiving teletherapy plus AI-assisted practice. A secondary objective was to explore relationships between AI usage and articulation outcomes. We hypothesized that participants receiving AI-assisted practice would demonstrate greater improvement.
Methods
Study design
This study was a randomized controlled pilot trial designed to evaluate feasibility, estimate effect size and inform planning for a future definitive trial. The manuscript was prepared with reference to the CONSORT 2010 extension for pilot and feasibility trials. Participant flow is described narratively and depicted in the CONSORT diagram (Figure 1).
A target sample size of 20-30 participants was selected based on recommendations for pilot randomized controlled trials. The trial was conducted within an existing teletherapy service workflow. Because of its exploratory nature, the study was not prospectively registered; this is acknowledged as a limitation and future trials will be registered prior to enrollment.
Participants
Eligible participants were children aged 5-17 years who demonstrated word-level errors in /s/ or /r/ and resided in one of 20 eligible U.S. states. Exclusion criteria included documented developmental or neurological diagnoses (e.g., ADHD, ASD) to reduce heterogeneity in attention and learning profiles. The youngest enrolled participant was 6 years 6 months old, resulting in a final age range of 6;6 to 17;9 (years; months).
Recruitment and enrollment
Participants were recruited through the Better Speech online platform using email outreach, digital marketing and provider referrals. Eligibility was confirmed through screening and clinician evaluation. Caregivers provided informed consent prior to participation.
Randomization and allocation
Participants were randomized 1:1 to either (1) traditional teletherapy plus AI-assisted practice (Group A) or (2) traditional teletherapy alone (Group B). Randomization was conducted within the clinical workflow; allocation concealment procedures were not implemented due to operational constraints of the teletherapy platform.
Intervention
Traditional teletherapy. All participants completed eight weekly clinician-led teletherapy sessions targeting /s/ or /r/ at the word level. Sessions were delivered by licensed speech-language pathologists using standardized articulation therapy techniques.
AI-assisted practice tool. Participants in Group A received access to Jessica, an AI-assisted articulation practice tool developed by Better Speech. Jessica provides automated feedback on phoneme production using acoustic pattern analysis and delivers corrective prompts in real time. The system supports high-frequency, repetition-based practice between clinician-led sessions.
Participants were instructed to use Jessica three times weekly for approximately 30 minutes per session. Actual usage varied widely (range 0-8 sessions; 0-174.5 minutes total) and is reported in Table 1. Raw audio was not stored; only anonymized performance metrics (e.g., session count, practice duration, item-level accuracy) were transmitted.
Formal validation of automated scoring against human raters was not conducted as part of this pilot and is identified as a priority for future research.
Table 1: Participant characteristics and raw outcome data (Group | Age | Gender | Target | Baseline | Post | Improvement pp | AI sessions | Total AI time).
62005 | A | 12;0 | Female | /r/ | 1.00% | 65.00% | 64 | 8 | 91 min 34 s
62008 | A | 7;0 | Female | /r/ | 50.00% | 60.00% | 10 | 5 | 64 min 26 s
62011 | A | 13;1 | Female | /r/ | 5.00% | 35.00% | 30 | 3 | 53 min 38 s
62013 | A | 12;0 | Female | /s/ | 50.00% | 60.00% | 10 | 4 | 45 min 2 s
62014 | A | 11;1 | Male | /r/ | 5.00% | 75.00% | 70 | 5 | 53 min 1 s
62015 | A | 13;8 | Male | /s/ | 30.00% | 60.00% | 30 | 5 | 152 min 3 s
62018 | A | 11;2 | Male | /r/ | 50.00% | 72.00% | 22 | 4 | 61 min 23 s
62020 | A | 12;7 | Male | /s/ | 40.00% | 90.00% | 50 | 0 | 0 min 0 s
62021 | A | 9;4 | Female | /r/ | 23.00% | 58.00% | 35 | 3 | 46 min 17 s
62022 | A | 6;6 | Male | /r/ | 0.00% | 63.00% | 63 | 2 | 44 min 10 s
62025 | A | 12;1 | Female | /s/ | 37.00% | 82.00% | 45 | 4 | 174 min 30 s
62026 | A | 8;1 | Male | /r/ | 97.00% | 98.00% | 1 | 3 | 93 min 6 s
62001 | B | 6;11 | Male | /r/ | 32.00% | 33.00% | 1 | NA | NA
62016 | B | 6;9 | Female | /r/ | 34.00% | 39.00% | 5 | NA | NA
62012 | B | 17;9 | Female | /r/ | 5.00% | 19.00% | 14 | NA | NA
62017 | B | 8;10 | Male | /r/ | 5.00% | 35.00% | 30 | NA | NA
62019 | B | 13;2 | Male | /r/ | 7.00% | 50.00% | 43 | NA | NA
62035 | B | 13;1 | Female | /r/ | 0.00% | 70.00% | 70 | NA | NA
62036 | B | 11;9 | Female | /s/ | 34.00% | 92.00% | 58 | NA | NA
62037 | B | 10;0 | Female | /r/ | 6.00% | 37.00% | 31 | NA | NA
62038 | B | 8;2 | Male | /r/ | 3.00% | 7.00% | 4 | NA | NA
62039 | B | 10;11 | Female | /r/ | 31.00% | 57.00% | 26 | NA | NA
62040 | B | 9;4 | Male | /r/ | 70.00% | 90.00% | 20 | NA | NA
Note: S1 = Session 1 (initial session); S8 = Session 8 (final session); pp = percentage points.
Outcome measures. The primary outcome was articulation accuracy, calculated as the percentage of correctly produced target phonemes across standardized word lists. Assessments were conducted at baseline (Session 1) and post-intervention (Session 8). Only baseline and post-intervention measures were included in the analysis.
Blinding. Speech-language pathologists scoring the recorded word-list productions were not blinded to group assignment. Because recordings were collected within a clinical teletherapy workflow, blinding was not feasible. Future trials will incorporate blinded, independent raters.
Rater reliability. Inter-rater reliability was not formally assessed. Scoring procedures followed routine clinical practice. Future studies will include dual-rater scoring and reliability metrics.
Data privacy and security. All participant data were stored in compliance with HIPAA standards. Audio recordings were encrypted in transit and at rest.
Statistical analysis. Between-group differences in mean improvement were evaluated using an independent-samples t test (two-tailed, alpha=0.05). Effect size was calculated as Hedges g with the small-sample correction factor J. 95% confidence intervals were computed for both the mean difference and effect size.
Exploratory Pearson correlations examined relationships between AI usage (number of practice sessions; total practice time) and improvement within Group A. Analyses were conducted on an intention-to-treat basis; the single Group A participant who completed zero AI sessions was retained in the primary comparison.
All calculations were independently verified against the raw data presented in Table 1.
Results
Participant flow. Twenty-three participants were enrolled, randomized (Group A: n=12; Group B: n=11) and completed all study procedures. No participants withdrew or were lost to follow-up. One participant in Group A completed zero AI practice sessions but was retained in the intention-to-treat analysis. Participant flow is shown in the CONSORT diagram (Figure 1).
Baseline characteristics. Baseline demographic and clinical characteristics were broadly comparable between groups (Table 2). Mean age was similar (Group A: 10.7 years; Group B: 10.6 years). Baseline articulation accuracy was numerically higher in Group A (32.3%, SD 28.3) than in Group B (20.6%, SD 21.6), although this difference was not statistically significant (t(21)=1.12, P=0.28).
Table 2: Baseline characteristics by group.
Age, mean (SD), years;months — Group A: 10.7 (2.4); Group B: 10.6 (3.2); t=0.08; p=0.93
Age range (years;months) — Group A: 6;6-13;8; Group B: 6;9-17;9
Female, n (%) — Group A: 6 (50.0%); Group B: 6 (55.0%); Fisher's exact OR=0.83; p=1
Male, n (%) — Group A: 6 (50.0%); Group B: 5 (45.0%)
Target sound /s/, n (%) — Group A: 4 (33.0%); Group B: 1 (9.0%)
Target sound /r/, n (%) — Group A: 8 (67.0%); Group B: 10 (91.0%); Fisher's exact OR=0.20; p=0.32
Baseline accuracy, mean SD (%) — Group A: 32.3 (28.3); Group B: 20.6 (21.6); t=1.12; p=0.28
Note: Continuous variables were compared using independent-samples t-tests. Categorical variables were compared using Fisher's exact tests because of small cell counts. OR = Odds Ratio. n = number of participants.
The distribution of target sounds differed between groups: Group A included a higher proportion of /s/ targets (33%) compared with Group B (9%). Given the small sample size, these imbalances are acknowledged as limitations.
Primary outcome: Articulation improvement. Mean improvement in articulation accuracy from baseline to post-intervention was:
Group A (teletherapy+AI practice): 35.8% (SD 22.9).
Group B (teletherapy only): 27.5% (SD 22.4).
The between-group mean difference was 8.4 percentage points (95% CI -11.3 to 28.1), t(21)=0.89, P=0.39. The effect size was Hedges g=0.36 (95% CI -0.44 to 1.15), representing approximately 30.5% greater improvement in Group A relative to Group B.
As expected for a pilot trial, confidence intervals were wide and crossed zero, indicating that the study was underpowered to detect statistically significant differences.
Exploratory engagement findings (Group A only). Exploratory analyses examined associations between AI practice and improvement within Group A.
Across all 12 participants in Group A:
Correlation between number of AI sessions and improvement: r=0.09.
Correlation between total AI practice time and improvement: r=-0.08.
When restricting analyses to the 11 participants who completed at least one AI session:
Correlation between number of sessions and improvement: r=0.28 (weak, nonsignificant).
These exploratory findings do not establish a dose-response relationship and should be interpreted cautiously given the small sample size and variability in engagement.
Safety. No adverse events were reported.
Participant-level data. Complete participant-level data, including baseline accuracy, post-intervention accuracy, improvement, AI usage and demographic characteristics, are presented in Table 1. These data enable independent reproduction of all statistical analyses.
Discussion
Principal findings
In this randomized controlled pilot trial, children receiving AI-assisted home practice demonstrated greater mean improvement in articulation accuracy than those receiving traditional teletherapy alone (35.8% vs. 27.5%). The absolute between-group difference of 8.4 percentage points and moderate effect size (Hedges g=0.36) suggest a potential benefit of integrating AI-supported practice into pediatric articulation therapy. Although the confidence intervals were wide and the results did not reach statistical significance, the direction and magnitude of the effect indicate that AI-assisted practice may enhance treatment intensity and support carryover between clinician-led sessions.
These findings align with established evidence that increased practice opportunities and structured feedback contribute to improved speech-sound outcomes [4,13]. Prior meta-analytic work has demonstrated that speech-language interventions are effective for developmental speech and language disorders and that treatment dosage may influence outcomes [2]. The present study extends this literature by suggesting that an AI-supported home-practice tool may offer a scalable means of increasing treatment intensity without adding clinician burden.
The results also reflect principles of motor learning, which emphasize high-frequency, distributed practice and immediate feedback as mechanisms that facilitate skill acquisition and retention [3]. Jessica, the AI-assisted practice tool evaluated in this study, was designed to provide repeated opportunities for practice and real-time feedback outside therapy sessions. These characteristics may partially explain the numerically greater improvements observed in the intervention group.
Comparison with prior work
The findings are consistent with emerging evidence supporting technology-enhanced interventions for speech sound disorders. Computer-based and game-based speech therapy platforms have been shown to increase practice opportunities, improve engagement and support accessibility for children with SSDs [6,7]. Similarly, visual-acoustic and ultrasound biofeedback interventions have demonstrated positive outcomes for children with persistent residual /r/ distortions, leveraging augmented feedback to support motor learning [11,12].
More recently, Benway and Preston reported positive outcomes using artificial intelligence-assisted therapy for residual /l/ errors in a single-case experimental study [8]. Although the present study included both /r/ and /s/ targets and employed a randomized pilot design, the observed pattern of greater improvement in the AI-assisted group is broadly consistent with these preliminary findings and contributes additional evidence supporting the potential role of AI in speech intervention.
The study also adds to the growing literature demonstrating that tele-practice can produce outcomes comparable to traditional face-to-face service delivery [9]. Integrating AI-supported practice into a teletherapy model may further enhance treatment intensity and support carryover of newly acquired speech skills between sessions [14,15].
Interpretation
Although participants receiving AI-assisted practice demonstrated greater improvement, the absence of a clear dose-response relationship tempers causal interpretation. Engagement varied widely and improvement was not strongly correlated with either number of AI sessions or total practice time. This variability may reflect individual differences in learning readiness, baseline severity, family support, adherence or responsiveness to feedback. As such, the findings should be interpreted as preliminary evidence of potential benefit rather than confirmation of efficacy.
Strengths
Conducted in a real-world teletherapy environment
Standardized assessment procedures
Complete participant-level data enabling independent verification
Intention-to-treat analysis
Transparent reporting aligned with CONSORT pilot guidelines.
Limitations
This pilot trial was limited by its small sample size, lack of blinded scoring and variability in AI engagement. Scoring was performed by non-blinded clinicians and inter-rater reliability was not assessed. Baseline imbalances in articulation accuracy and target-sound distribution may have influenced results. Exclusion of participants with co-occurring diagnoses may limit generalizability. The trial was not prospectively registered. The AI tool's automated scoring was not validated against human raters in this study.
Power considerations
The study was not powered to detect statistically significant differences. The observed effect size (g=0.36) provides a preliminary estimate for planning a future definitive trial. Detecting a between-group effect of this magnitude with 80% power at alpha=0.05 (two-tailed) would require approximately 120 participants per group; this estimate should be refined through formal power analysis.
Future directions
Future research should include larger, prospectively registered samples; blinded independent raters; inter-rater reliability metrics; validation of AI scoring against human raters; long-term follow-up and evaluation across additional speech sound disorders. Further work should also examine adherence, engagement and potential dose-response relationships.
Clinical implications
This pilot trial provides preliminary evidence that AI-assisted speech practice may enhance articulation outcomes when combined with traditional therapy. Structured, feedback-driven home-practice tools may help clinicians increase treatment intensity, improve carryover and support families seeking accessible practice opportunities outside therapy sessions. These implications remain tentative pending confirmation in adequately powered trials.
Conclusion (Published Article)
Children who received AI-assisted speech practice demonstrated a 30.5% greater mean improvement in articulation accuracy compared with those receiving traditional teletherapy alone. Although the between-group difference did not reach statistical significance in this underpowered pilot trial, the direction and magnitude of the effect suggest that integrating AI-supported home practice may enhance articulation outcomes for school-aged children with speech sound disorders.
These findings provide preliminary evidence supporting the feasibility and potential benefit of AI-assisted practice as an adjunct to clinician-directed therapy delivered via tele-practice. Larger, prospectively registered randomized controlled trials with blinded scoring, validated AI metrics and adequate sample sizes are warranted to determine efficacy, explore dose-response relationships and evaluate generalizability across additional speech sound disorders.
Conflicts of Interest. The authors are employed by Better Speech LLC, which developed the AI-assisted practice tool evaluated in this study.
Data Availability. All data are available upon reasonable request.
Trial Registration. Not applicable. This exploratory pilot trial was not prospectively registered.
References (Published Article)
1. Wren Y, Miller LL, Peters TJ, Emond A, Roulstone S (2016) Prevalence and predictors of persistent speech sound disorder at eight years old: Findings from a population cohort study. J Speech Lang Hear Res 59: 647-673.
2. Law J, Garrett Z, Nye C (2004) The efficacy of treatment for children with developmental speech and language delay/disorder: A meta-analysis. J Speech Lang Hear Res 47: 924-943.
3. Maas E, Robin DA, Austermann HSN, Freedman SE, Wulf G, et al. (2008) Principles of motor learning in treatment of motor speech disorders. Am J Speech Lang Pathol 17: 277-298.
4. Wales D, Skinner L, Hayman M (2017) The efficacy of telehealth-delivered speech and language intervention for primary school-age children: A systematic review. Int J Telerehabil 9: 55-70.
5. Hair A, Monroe P, Ahmed B, Ballard KJ, Gutierrez-Osuna R (2018) Apraxia world: A speech therapy game for children with speech sound disorders. Proceedings of the 17th ACM Conference on Interaction Design and Children: 201-219.
6. Benway NR, Preston JL, Hitchcock E, Rose Y, Salekin A, et al. (2023) Reproducible speech research with the artificial intelligence-ready PERCEPT corpora. J Speech Lang Hear Res 66: 1986-2009.
7. Furlong L, Erickson S, Morris ME (2017) Computer-based speech therapy for childhood speech sound disorders. J Commun Disord 68: 50-69.
8. Benway NR, Preston JL (2024) Artificial intelligence-assisted speech therapy for /l/: A single-case experimental study. Am J Speech Lang Pathol 33: 2461-2486.
9. Grogan-Johnson S, Schmidt AM, Schenker J, Alvares R, Rowan LE, et al. (2013) A comparison of speech sound intervention delivered by tele-practice and side-by-side service delivery models. Commun Disord Q 34: 210-220.
10. Williams AL (2012) Intensity in phonological intervention: is there a prescribed amount? Int J Speech Lang Pathol 14: 456-461.
11. Preston JL, Hitchcock ER, Leece MC (2020) Auditory perception and ultrasound biofeedback treatment outcomes for children with residual /r/ distortions: A randomized controlled trial. J Speech Lang Hear Res 63: 444-455.
12. McAllister BT (2017) Efficacy of visual-acoustic biofeedback intervention for residual rhotic errors: A single-subject randomization study. J Speech Lang Hear Res 60: 1175-1193.
13. Sugden E, Baker E, Munro N, Williams AL (2016) Involvement of parents in intervention for childhood speech sound disorders: A review of the evidence. Int J Lang Commun Disord 51: 597-625.
14. Fagherazzi G, Goetzinger C, Rashid MA, Aguayo GA, Huiart L (2020) Digital health strategies to fight COVID-19 worldwide: Challenges, recommendations and a call for papers. J Med Internet Res 22: e19284.
15. Preston JL, Leece MC, McNamara K, Maas E (2019) Intensive treatment with ultrasound visual feedback for speech sound errors in childhood apraxia. Front Hum Neurosci 13: 440.
Source: Conner S, Kearney E, Norris J (2026) Improving Articulation via AI-Assisted Speech Practice in Children: A Randomized Controlled Pilot Trial. Journal of Biology and Health Science 3: 164. DOI: 10.56147/jbhs.3.4.164. Reproduced here by the authors (Better Speech LLC).
