Voice-Based AI Maths Tutoring vs. Text-Based: Why Conversation Matters For Maths Progress

AI tutoring can be a promising solution for school leaders facing limited budgets and stretched resources, particularly when tackling learning gaps in maths. But how students interact with their AI maths tutor – either voice or text-based communication – can significantly impact their learning outcomes, particularly in mathematics. If the AI tutor can’t hear students explain their thinking, it may miss the most important part of maths learning.

This article explores how voice-based AI tutoring offers advantages over more widely recognised text-based alternatives for maths instruction. It examines the research-backed benefits and real-world applications helping schools provide affordable, personalised learning at scale.

Understanding the different approaches to AI tutoring

Artificial intelligence (AI) in education may seem like a relatively new concept, but it can be traced back as far as the 1960s. Today, AI has been used to develop intelligent tutoring systems (ITS) so they can deliver adaptive and personalised learning in classrooms. 

Originally, AI tutors used text-based communication to provide support to learners in the form of generated questions and marking answers right or wrong. However, advancements in AI and large language models (LLMs) are enabling the development of personalised voice-based AI. 

While voice-based AI is not new – automated voice messages have existed for many years – these advancements make it possible to provide real-time voice-based AI tutoring in education.   

What is voice-based AI tutoring?

Voice-based AI tutoring, or conversational AI tutoring, is an intelligent tutoring system (ITS) that delivers a form of online tutoring by creating a dialogue between the student and the artificial intelligence through spoken interaction. It combines: 

• Automatic speech recognition (ASR) to understand what students say
• Natural language understanding to interpret meaning and context
• Adaptive pedagogical strategies to respond appropriately
• Speech synthesis to deliver verbal feedback and instruction

Voice-based tutors like Third Space Learning’s Skye enable students to explain their mathematical thinking aloud, just as they would with a traditional human tutor. The AI listens, analyses the response, and provides immediate verbal feedback tailored to the student’s needs.

What is text-based AI tutoring?

Unlike conversational AI tutors, text or chat-based AI tutoring relies on written communication between the student and the ITS. Students read questions or prompts, type their responses, and receive written feedback. While these systems have improved significantly in recent years, they still operate primarily through:

• Written prompts and questions are presented on screen
• Typed student responses
• Text-based feedback and hints
• Sequential interaction that follows a more rigid pattern

Text-based systems often focus more on providing correct answers rather than developing mathematical thinking processes, with interactions that feel less conversational and more transactional.

While text-based tutoring systems have improved significantly, they still face inherent limitations in maths education, where explaining step-by-step reasoning verbally often leads to deeper understanding than typing answers.

Skye: voice-based AI maths tutoring in practice 

Third Space Learning’s maths experts and teachers developed Skye, a voice-based AI maths tutor, using the same pedagogical principles, effective teaching strategies and curriculum-aligned lessons as our traditional, human-led tutoring.    

67% of secondary students are already using AI tools in education. But what sets Skye apart from other AI maths tools is the application of expert insights from over 2 million one-to-one tutoring sessions delivered to more than 170,000 students across 4,000 UK schools. The most effective AI tutoring is built on human expertise.

Skye follows the seven principles of effective AI maths tutoring identified by education researchers:

1. Comprehensive curriculum crafted by experts: Qualified maths teachers design all of Skye’s lessons. Nothing is generated by AI, ensuring accuracy and pedagogical soundness.
2. Scaffolded learning with gradual hints: Skye follows the “I do, we do, you do” approach. Rather than giving the right answer, it provides up to three targeted hints that address specific misconceptions. 
3. Active recall and spaced repetition: Experts have built Skye on a spiral curriculum, incorporating regular retrieval practice and revisiting previous concepts at strategic intervals to strengthen long-term memory.
4. Interleaving for enhanced understanding: Each Skye lesson includes a variety of question types, including exam-style questions, and opportunities to connect mathematical concepts.
5. Structured challenge problems: Skye includes carefully designed multi-step challenges in every lesson to promote deeper thinking and reasoning opportunities.
6. Reflective feedback: Skye is trained to understand “correct ideas”, not just correct answers. It asks open-ended questions that encourage students to articulate their reasoning.
7. Cognitive load management: Skye’s interface is intentionally simplified, with lessons broken into small, manageable steps and features like selective blurring to focus attention on relevant information.

Voice vs text: 11 educational advantages of voice-based AI tutoring in maths

1. Verbalisation improves mathematical reasoning

Voice-based AI tutoring: When learners talk through problems, they articulate their thought process, strengthening neural connections to make reasoning clearer and easier to recall. Verbalisation also helps the AI tutor to identify misconceptions immediately.

The Education Endowment Foundation (EEF) found that when students explain their mathematical thinking aloud, they engage in metacognitive processes that help consolidate understanding.

For example, when a student verbally reasons how to convert a fraction to a decimal, they must articulate each step, e.g. \frac{3}{4} to 0.75, “I need to divide 3 by 4, which gives me 0.75.”

Skye prompts pupils to explain their thinking aloud by asking open-ended questions such as, ‘How did you work that out?’ This verbal articulation of mathematical processes helps deepen understanding and identify misconceptions immediately.

Text-based AI tutoring: Students often give brief answers without providing the reasoning behind them. A lot of typing can discourage thorough explanations.

2. Reduces mathematical cognitive load 

Voice-based AI tutoring: When students engage in mathematical discussions, they use a different cognitive process from typing. Students can focus entirely on mathematical thinking without the added cognitive burden of reading complex instructions and formulating written responses.

For example, when solving a complex problem about calculating the area of a compound shape, students can concentrate solely on the mathematical reasoning rather than how to express it in writing. Skye presents one idea at a time, using screen blur, highlight and pointing tools to guide focus and reduce cognitive load.

Text-based AI tutoring: Students have to divide their cognitive resources between decoding text, formulating written responses, and solving mathematical problems. This split attention reduces capacity for mathematical thinking.

3. Removes literacy barriers to maths

Voice-based AI tutoring: Takes away reading and writing barriers, allowing students to engage with mathematical concepts regardless of their reading or writing abilities.

For example, younger pupils or those with additional needs such as dyslexia can understand and solve complex mathematical problems without needing to read word problems or write detailed explanations.

Skye uses carefully tested, age-appropriate mathematical language. If students don’t understand a question, Skye can rephrase instructions using simpler language while maintaining mathematical accuracy.

Text-based AI tutoring: Places high demand on reading fluency and writing skills. Students with limited literacy, including EAL students, may struggle to access maths through text.

4. Adaptive maths support through dialogue

Voice-based AI tutoring: Can detect nuances such as hesitation, confusion, or uncertainty and adjust responses accordingly, providing the right level of help.

For instance, when a student struggles to explain how to solve an equation like 2x + 3 = 7, the tutor can hear uncertainty and provide appropriate guidance on that specific step.

Skye provides scaffolded hints in three steps: 

1. Gentle nudge,
2. Clearer clue, 
3. Full explanation if needed.

Text-based AI tutoring: Misses nuances and relies solely on written responses to gauge understanding, missing vocal cues that indicate confusion. This limitation makes it harder to provide the appropriate support.

5. Creates engaging and authentic learning experiences

Voice-based AI tutoring: Mimics the natural back-and-forth of human conversation, creating a presence that many students find more engaging and supportive.

For example, the tutor can use an encouraging tone when a student correctly identifies that the area of a triangle is ½ × base × height, reinforcing positive learning experiences.

“Today, we’re thrilled to be at the forefront of using Third Space Learning’s AI voice tutoring. The children are thoroughly enjoying the experience, and their engagement and focus are clearly evident.”
Chris Harris, Deputy Head
Admirals Academy, Norfolk

Text-based AI tutoring: Often lacks conversational quality, with interactions feeling more mechanical and less personally engaging. The absence of real-time, encouraging dialogue can make the experience feel less supportive.

6. Builds mathematical confidence and reduces anxiety

Voice-based AI tutoring: Creates a safe learning environment where students can focus on the maths without worrying about spelling, grammar or typing skills.

For example, a student who normally feels anxious about written work can confidently explain verbally why 3² = 9 without worrying about how to format the explanation correctly in writing.

“Third Space Learning’s voice-based AI tutoring has reinvigorated some of our most reluctant learners. They are so excited that they ask their teacher to leave class a bit early so that they can get started.”
Ryan Lucas, Principal 
Governor’s Ranch

Text-based AI tutoring: May create additional anxiety for students who struggle with writing or spelling, potentially causing them to limit their responses or avoid fully expressing their mathematical thinking.

7. Easier to support more students simultaneously

Voice-based AI tutoring: Enables schools to provide personalised maths support to many students simultaneously without needing additional staff members or resources.

For example, a school could run 30 simultaneous one-to-one tutoring sessions with just one staff member supervising, allowing many students to receive personalised support.

Skye offers unlimited sessions for a fixed yearly price starting from £3,500, meaning schools no longer have to choose which students receive support based on budget constraints. 

“This innovative one-to-one maths tutoring solution offers an even more cost-effective alternative.”
Chris Harris, Deputy Head
Admirals Academy, Norfolk

Text-based AI tutoring: While it can also operate at scale, it often requires more teacher intervention to address misunderstandings not captured in the text-based interaction. This limits its scalability.

8. Develops mathematical language

Voice-based AI tutoring: Demonstrates and helps students internalise and practise correct mathematical vocabulary through spoken dialogue.

For example, if a student refers to the top part of a fraction, the voice-based AI tutor can repeat the sentence using the correct terminology; numerator. 

Text-based AI tutoring: Typically provides fewer opportunities to model and reinforce mathematical language in real-time, leading to slower development of precise vocabulary.

9. Real-time formative assessment

Voice-based AI tutoring: Allows tutors to instantly assess understanding through tone, language, and explanation, adapting instruction on the spot.

For example, if a student incorrectly says, “I multiplied both sides by 2,” rather than dividing, the tutor can adapt the tutoring to address the misconception or mistake. It might clarify: “Think again – what’s the opposite of multiplying?”

This rapid, responsive feedback loop enhances learning and minimises the persistence of misconceptions.

Skye understands when students are hesitant and when they have misconceptions. It then provides scaffolded support to address the misconception and build confidence. 

Text-based AI tutoring: Cannot interpret tone or spontaneous verbal reasoning, often requiring delayed clarification or follow-up from teachers.

10. Inclusive of EAL learners

Voice-based AI tutoring: Provides flexible support through real-time speech for students with English as an additional language (EAL) or neurodiverse profiles, reducing language and processing barriers.

For example, if an EAL pupil is confused by “subtract,” the tutor can rephrase: “That means to take away. Can you try again?”

When students don’t understand the question or instruction, Skye can rephrase it.  

Text-based AI tutoring: Often provides static instructions or feedback, which can be harder for EAL and neurodiverse learners to decode.

11. Reinforces the working memory

Voice-based AI tutoring: Engages the working memory by allowing students to hear and say mathematical concepts aloud, strengthening retention.

This method supports dual-channel learning, combining auditory and visual, which is particularly helpful for younger learners or those with working memory challenges.

Every Skye session has visual lesson slides to support the learning objective. Students can reference the slides while listening and responding to Skye to reinforce mathematical concepts.

Text-based AI tutoring: Primarily engages visual processing and doesn’t activate auditory learning pathways, limiting reinforcement through spoken repetition.

Ensuring safety and accountability in AI tutoring

Whether AI tutoring is delivered through voice or text, safeguarding must remain a top priority. Both formats involve student interactions with a digital system. 

This means clear AI tutoring policies and oversight are essential to keep students safe and maintain accountability.

For both voice-based and text-based AI tutoring tools, schools should ensure that:

• Sessions are scheduled and supervised by authorised staff
• Conversations are recorded or logged
• Any concerns are quickly flagged and escalated where necessary
• You have an AI policy in school covering use of all AI tools

Unlike many chat tools that rely on unmonitored written exchanges, Skye, the voice-based AI maths tutor records all audio interactions, making it easy to monitor and flag concerns.

Red flags are automatically recorded, reviewed and escalated, where appropriate, by a member of the Third Space Learning team. They are categorised from 1 (most severe) to 5 (least severe):

1. Safeguarding concern
2. Disruptive behaviour (severe)
3. Pupil distress
4. Disruptive behaviour (mild)
5. Student frustration

Implementing voice-based AI tutoring in schools

For schools considering voice-based AI tutoring, here’s a practical implementation roadmap:

1. Assess technology: Ensure you have a reliable internet connection, appropriate devices, and headsets with microphones for each student.
   
   For Skye, you only need basic equipment that most schools already have.
   
2. Identify target students: Determine which students would benefit most from extra maths support via a voice-based AI maths tutor. 
   
   Schools currently using Skye support between 16-120 weekly sessions.
   
3. Schedule tutoring sessions: Allocate specific timeslots for additional maths tutoring that fit into the school day.  
   
   With Skye, you can schedule sessions at 5-minute intervals before, during, or after school hours, meaning you can integrate tutoring into your existing timetable seamlessly.
   
4. Prepare quiet learning spaces: Voice-based tutoring requires a space where students can speak aloud without distraction.
   
   Many schools using Skye opt to use corridors, breakout spaces, or library corners, and dedicated rooms such as computer suites.
   
5. Train supervising staff: Ensure staff understand how to oversee multiple sessions simultaneously.
   
   Skye offers an “observe session” feature allowing staff to join any live session as a muted observer to check progress or assist with technical issues.
   
6. Set clear learning goals: All voice-based AI tutoring sessions must align with curriculum objectives. Diagnostic assessments help to identify each pupil’s knowledge gaps.
   
   With Skye, teachers can reorder lessons to prioritise these gaps or align with classroom teaching.
   
7. Monitor and evaluate: Regularly review progress and adjust implementation as needed.
   Schools can track student attendance, engagement, and progress through detailed session reports.
   
   Early data shows students perform better on post-session questions with Skye, and attendance is higher due to increased flexibility.
   
8. Safeguarding considerations: Ensure your AI tutoring policy addresses data protection and safeguarding requirements.
   
   Skye records all sessions and automatically flags potential safeguarding concerns, categorising them in order of severity.

The future of personalised AI maths support 

Voice-based AI tutoring offers a more effective solution for extra maths support than text-based alternatives. Conversational AI maths tutors encourage students to verbalise their reasoning and problem-solving. This helps the tutor identify and address misconceptions in real-time, just as a teacher would in class.

Verbal communication allows for immediate and personalised feedback in the same way traditional one-to-one tutoring does. This supports reasoning, focus, and confidence in ways that short, typed answers can’t match.

For schools looking to provide high-quality maths support at scale, voice-based AI tutoring combines the best teaching strategies with the consistency and availability of technology. It makes personalised learning accessible to every pupil who needs it, without compromising on safety, oversight or teaching standards.

FAQs