The VOXEO Method

Almost everyone learning a language lives in one of these two worlds.

World one

Tests without a plan

A placement test, a letter from the European framework, and then a course that’s identical for everyone who got that letter. The test tells the person where they are; it doesn’t tell them what to do on Tuesday.

World two

Plans without a diagnosis

An app with a fixed syllabus, a daily streak and lessons on ordering food in a restaurant that have nothing to do with the sounds and structures that are actually holding them back on a call with a client.

VOXEO starts with assessment, and that changes everything. Because we’ve already measured in detail how a person speaks, listens and writes, we know their specific weaknesses: the /b/ and /v/ contrast they can’t hear, the past-tense endings they swallow, the forty words from their job that they’re missing. The improvement plan is built from that profile, and rebuilt every time the profile changes.

Seven principles, and the evidence behind each one.

We didn’t invent a theory of how languages are learned. We read the research, above all the meta-analyses, which are studies that each combine dozens of studies, and built the product around the findings that hold up. Here they are, in plain language.

01

Having the rule explained works better than waiting to absorb it

The most cited review in the field (Norris & Ortega, 2000), and its later replication (Goo et al., 2015), found that explicit instruction produces large, lasting gains, and that telling the student what the pattern is beats leaving them to work it out alone. That’s why VOXEO’s tutor explains, in two sentences, the two or three things you’re working on this week, instead of trusting immersion to sort it out.

What VOXEO doesEach weekly plan names two or three specific goals taken from your assessment, and the tutor explains each one in a sentence or two before you practice it.

02

You learn by producing, not just by understanding

Understanding a sentence and being able to say it are two different skills. Studies comparing comprehension-based and production-based teaching (Shintani, Li & Ellis, 2013; Shintani, 2015) found that if you want someone to produce the language, which is what a job demands, they have to practice producing it. Speaking and writing force you to notice the gap between what you meant to say and what came out.

What VOXEO doesThe center of every session is you speaking. Listening and reading are there to feed your speaking, not to replace it.

03

Correction works, and the best kind makes you correct yourself

Correction has a solid, lasting effect on learning (Li, 2010; Lyster & Saito, 2010), but the type matters. Repeating the sentence back correctly is kind and very easy to miss. Asking “which tense do we use for something that already happened?” and waiting for you to fix it produces bigger gains, because you do the mental work. Correcting everything at once, on the other hand, overwhelms and kills confidence.

What VOXEO doesThe tutor asks first and explains after; it only corrects your current goals and whatever stops you being understood; and during fluency exercises it stays quiet and saves corrections for the final review.

04

Memory needs spacing and retrieval, not cramming

Across 48 experiments with more than three thousand language learners, spaced practice beat massed practice, and longer intervals won on long-term tests (Kim & Webb, 2022). Pulling something out of memory beats rereading it (Roediger & Karpicke, 2006), and doing it with correction beats doing it without. A new word needs ten or more meaningful encounters before it sticks (Webb, 2007).

What VOXEO doesEvery word, sound and structure you struggle with goes into your personal bank and comes back at widening intervals, roughly days 1, 2, 4, 8, 15 and 30, always as a recall task: you say it and the system scores it, until you’ve retrieved it correctly enough times to graduate it.

05

Practice real tasks, and repeat them

Programs built around meaningful tasks, like resolving a complaint, explaining a policy or politely disagreeing with a colleague, outperform lessons built around grammar points (Bryfonski & McKay, 2019). And doing the same task a second time, faster or with less support, is one of the best-documented ways to build fluency (Nation, 1989; De Jong & Perfetti, 2011). It also lowers speaking anxiety, and with an AI it drops even further because there’s nobody to be embarrassed in front of.

What VOXEO doesEvery session has a role-play built from the real context of your job, done twice: once to get through it, and once under time pressure or with a twist. The second run is where fluency comes from.

06

Pronunciation starts in the ear, and needs many voices

You can’t reliably produce a sound you don’t hear. The best-supported pronunciation method is high variability phonetic training: hearing the contrast from many different speakers and in many contexts, choosing what you heard and getting immediate feedback (Uchihara, Karas & Thomson, 2025; 79 studies). Hearing a single voice doesn’t generalize; hearing five to ten does.

What VOXEO doesExercises use several synthetic voices, target the specific contrasts your assessment flagged (for Spanish speakers, almost always /b/ and /v/, /i/ and /ɪ/, dropped -ed endings, the “e” that slips in before an s, and flattened stress), and combine listening with recording and comparing yourself.

07

What you read and hear only teaches you if you already understand almost all of it

Reading and listening to real English matters, because that’s where you meet a word for the tenth time and where natural rhythm and phrasing come from. But the research sets a hard threshold: you need to know around 95% of the words to follow it, and 98% to learn from it (Nation, 2006; van Zeeland & Schmitt, 2013). Vocabulary size explains between a third and almost half of how well someone understands what they hear and read (Zhang & Zhang, 2022). Content that’s too hard is noise.

What VOXEO doesWe estimate your vocabulary size from the assessment and prescribe material at the right level, including which kinds of shows and podcasts will actually help you.

The VOXEO learning loop

The seven stages aren’t seven separate features; they’re a circle that never ends. Nobody graduates and leaves: you keep going around, a little higher each time.

AssessSpeaking, listening and writing, always scored with the same rubric.
DiagnoseThe scores become a profile: level, vocabulary estimate and a ranked list of specific errors.
PlanA weekly plan with two or three goals and a daily session of about twenty minutes.
PracticeTasks that force you to produce the language around those goals.
CorrectQuestions first and explanations after; selective, never exhaustive; and held back during fluency work.
RetrieveEverything that failed comes back on a spaced schedule until it comes out on its own.
ReassessA short weekly check with the same rubric, so progress is visible and the profile stays current.

A daily session, in twenty minutes

This is what a student does on an ordinary Tuesday. Each block exists because of one of the seven principles, and none of them is filler.

Block	Min	What happens	Principle
Warm-up	4	Spaced recall of the words, sounds and structures due today in your bank, out loud and scored.	04
Ear and mouth	4	A minimal-pairs exercise with several voices on one contrast, then recording yourself and comparing.	06
Task	8	A role-play from your work context, done twice, the second time faster or with a twist.	02 05
Review	4	The tutor leads you to correct two or three things yourself, explains the rule and adds new material to your bank.	01 03 04

On top of that, each week adds one session of real content and a three-minute reassessment.

Where vocabulary fits

Vocabulary isn’t a separate module. It’s the blood that runs through the whole method.

The words come from you. The bank fills up from the assessment, from words you reached for and didn’t have, and from your role-plays, with the ones the tutor had to give you or correct. Generic lists are the last resort.
Frequency first. The 2,000 to 3,000 most common word families cover between 90% and 95% of everyday speech (Nation, 2006). Gaps there get filled before anything rarer.
Chunks, not single words. Words are stored and retrieved inside the phrase they live in, “I’ll look into that” and not “look”, because that’s how fluent speakers use them and how memory stores them.
Recall, and out loud. You’re shown the Spanish or the situation and you say the English, and the system scores it. Recognizing a word on a multiple-choice card isn’t the same as having it available on a call.
Ten or more encounters, spaced. Each item is retrieved at widening intervals and reappears in different contexts, in the drill, the role-play and the listening, until it graduates.

Where shows, podcasts and music fit

This is the real material, and it’s what takes someone from “practicing” to “living” in the language. It only works under certain conditions, so VOXEO prescribes it instead of leaving it to chance.

Shows and video

At the right level. Comedies and dialogue-heavy dramas sit around 3,000 word families to reach 95% coverage; news, documentaries and technical content ask for far more (Webb & Rodgers, 2009). A B1 student gets prescribed a comedy, not a documentary.
English subtitles, not native-language ones. Subtitles in the language being learned improve both listening comprehension and vocabulary; subtitles in your first language don’t train the ear the same way (Montero Perez, Van Den Noortgate & Desmet, 2013).
Stick with one show. Vocabulary repeats far more within a single show than across different ones, and that repetition is exactly what delivers the ten encounters (Rodgers & Webb, 2011).
Wrap it. Before watching, VOXEO gives eight to ten likely words for that genre and topic. Afterward, the student tells the tutor about the episode in two minutes. That spoken summary is the step that turns watching TV into learning, because it converts passive material into production, correction and new items in the bank.

Podcasts and audio

The same rules, plus reading while listening, with the transcript on screen, for anyone below B2. It’s better supported than audio alone.

Music

The evidence is weak for comprehension and for vocabulary, because lyrics are compressed, poetic and rarely conversational. It’s good for two things: rhythm, linking and stress, and motivation. VOXEO treats it as an optional pronunciation aid, never as a source of learning.

The human part, and why your school isn’t optional

There are things an AI tutor can’t do and shouldn’t pretend to: negotiate meaning with an unpredictable human, read a room, provide social accountability, teach the pragmatics of a real relationship. That’s what classes, teachers and conversation groups are for, and VOXEO actively pushes students toward them.

What VOXEO adds is that the teacher no longer starts blind. With the student’s permission, the school receives:

their current level and skills profile,
their error inventory ranked by importance, with sounds, structures and vocabulary gaps,
this week’s goals, so the class hour reinforces exactly what the app is drilling,
and progress over time, so the teacher can see what’s moving and what’s stuck.

In return, whatever the teacher observes, whether it’s a pragmatics problem, a crisis of confidence or a leap forward, goes back into the student’s plan. The app handles repetition, spacing and drilling, which is exactly what a person’s time is too valuable for; the person handles the interaction only a person can give. Neither replaces the other.

What this means for your school →

What we deliberately don’t do

A short list matters as much as the method, because almost the whole market does exactly the opposite.

We don’t run a fixed syllabusThe plan is generated from the student’s profile, every week.

We don’t correct everythingSelective correction is a design decision, not a limitation.

We don’t interrupt fluencyCorrections wait for the review at the end.

We don’t host a video libraryWe prescribe the material and wrap it. The world already has enough TV.

We don’t measure success in streaksWe measure it in the reassessment. A streak is a reminder, not a goal.

We don’t pretend to be your friendWe’re a coach with a very good memory for your mistakes. Your friends, colleagues and teachers are the people you’re learning to talk to.

References

Everything above comes from here. If you run a school and want to check it, that’s exactly the idea.

Bryfonski, L., & McKay, T. H. (2019). TBLT implementation and evaluation: A meta-analysis. Language Teaching Research, 23(5), 603–632.

De Jong, N., & Perfetti, C. A. (2011). Fluency training in the ESL classroom. Language Learning, 61(2), 533–568.

Goo, J., Granena, G., Yilmaz, Y., & Novella, M. (2015). Implicit and explicit instruction in L2 learning. In P. Rebuschat (Ed.), Implicit and explicit learning of languages (pp. 443–482). John Benjamins.

Kim, S. K., & Webb, S. (2022). The effects of spaced practice on second language learning: A meta-analysis. Language Learning, 72(1), 269–319.

Li, S. (2010). The effectiveness of corrective feedback in SLA: A meta-analysis. Language Learning, 60(2), 309–365.

Lyster, R., & Saito, K. (2010). Oral feedback in classroom SLA: A meta-analysis. Studies in Second Language Acquisition, 32(2), 265–302.

Lyu, B., Lai, C., & Guo, X. (2025). Effectiveness of chatbots in improving language learning: A meta-analysis of comparative studies. International Journal of Applied Linguistics.

Montero Perez, M., Van Den Noortgate, W., & Desmet, P. (2013). Captioned video for L2 listening and vocabulary learning: A meta-analysis. System, 41(3), 720–739.

Nation, I. S. P. (1989). Improving speaking fluency. System, 17(3), 377–384.

Nation, I. S. P. (2006). How large a vocabulary is needed for reading and listening? Canadian Modern Language Review, 63(1), 59–82.

Nation, I. S. P. (2007). The four strands. Innovation in Language Learning and Teaching, 1(1), 2–13.

Norris, J. M., & Ortega, L. (2000). Effectiveness of L2 instruction: A research synthesis and quantitative meta-analysis. Language Learning, 50(3), 417–528.

Rodgers, M. P. H., & Webb, S. (2011). Narrow viewing: The vocabulary in related television programs. TESOL Quarterly, 45(4), 689–717.

Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning. Psychological Science, 17(3), 249–255.

Shintani, N. (2015). The effectiveness of processing instruction and production-based instruction on L2 grammar acquisition. Applied Linguistics, 36(3), 306–325.

Shintani, N., Li, S., & Ellis, R. (2013). Comprehension-based versus production-based grammar instruction. Language Learning, 63(2), 296–329.

Swain, M. (1985). Communicative competence. In S. Gass & C. Madden (Eds.), Input in second language acquisition (pp. 235–253). Newbury House.

Uchihara, T., Karas, M., & Thomson, R. I. (2025). High variability phonetic training (HVPT): A meta-analysis of L2 perceptual training studies. Studies in Second Language Acquisition.

van Zeeland, H., & Schmitt, N. (2013). Lexical coverage in L1 and L2 listening comprehension. Applied Linguistics, 34(4), 457–479.

Webb, S. (2007). The effects of repetition on vocabulary knowledge. Applied Linguistics, 28(1), 46–65.

Webb, S., & Rodgers, M. P. H. (2009). The lexical coverage of movies. Applied Linguistics, 30(3), 407–427.

Zhang, S., & Zhang, X. (2022). The relationship between vocabulary knowledge and L2 reading/listening comprehension: A meta-analysis. Language Teaching Research, 26(4), 696–725.

How we help a person get better at a language, and why we do it this way.