Skip to main content
Our Research Foundation

Built on what actually works

We analysed every major language learning system, studied decades of second language acquisition research, and listened to what learners actually need. Every feature in fluente.ai is grounded in peer-reviewed science.

345
CEFR-Aligned Lessons
50+
Languages
2,500+
Language Pair Combinations
24
Research Pillars
40+
Years of SLA Research Applied

Before writing a single line of code, we did something no other language app bothered to do: we studied all of them. We analysed every major language learning platform, method, and system worldwide. We read the reviews — tens of thousands of them — from real learners across every continent. We studied what language students search for online, what frustrates them, what they wish existed. And we went deep into decades of second language acquisition research to understand what the science actually says about how adults learn to speak a new language.

The conclusion was overwhelming and consistent: the fastest path to fluency is speaking. Not flashcards. Not grammar drills. Not gamified streaks. Speaking — with real-time feedback, in scenarios that matter to your life, without the anxiety of making mistakes in front of another person.

That’s what fluente.ai is. Every feature, every lesson, every design decision comes from this research. Here’s what we found.

What the Research Says

Peer-reviewed evidence

Not testimonials from users — findings from decades of academic study.

★★★★★
Peer-Reviewed Research
Output Hypothesis

Speaking output is essential for fluency

“Comprehensible input is not sufficient... comprehensible output is also necessary to develop learners’ fluency and accuracy.”
4,820 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Acquisition-Learning Hypothesis

Natural conversation produces fluent speech

“Only acquisition — a subconscious process similar to how children naturally acquire their first language — fosters spontaneous and fluent language use.”
6,340 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Anxiety Research

Speaking anxiety is the #1 barrier

“Students placed speaking in the foreign language at the top of the list when asked which aspects caused the most anxiety.”
2,190 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
AI-Assisted Learning

AI feedback reduces anxiety while improving accuracy

“AI-assisted corrective feedback improves accuracy while providing a pleasant learning environment that reduces anxiety about practising speaking.”
1,870 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Meta-Analysis

Corrective feedback has clear positive effects

“Corrective feedback is beneficial to second language learning, with explicit feedback being more effective than implicit feedback.”
3,410 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Learner Preferences

86% of learners want immediate correction

“86% of students reported that errors should be corrected as soon as they were made, to help avoid forming bad habits.”
2,780 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Interaction Hypothesis

Real conversation drives deep acquisition

“Negotiation of meaning, and especially negotiation work that triggers interactional adjustments, facilitates acquisition.”
5,120 found this helpful Read pillar → Source →
★★★★★
Foundational Research
Forgetting Curve

Without review, you forget within days

“Memory decays exponentially, but flattens with the use of spaced repetition review sessions, which significantly improve retention.”
8,900 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
PNAS Published

Optimised spaced repetition outperforms all alternatives

“Learners who follow an algorithmically optimised spaced repetition schedule memorise more effectively than those using alternative methods.”
3,650 found this helpful Read pillar → Source →
★★★★★
Research Consensus
Communicative Teaching

Communication ability beats perfect grammar

“The goal of language education is the ability to communicate in the target language — not grammatical competence alone.”
4,200 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
TBLT Research

Workplace scenarios build real workplace English

“Task-based instruction improves motivation, fluency, and communicative competence by engaging students in workplace scenario simulation.”
1,540 found this helpful Read pillar → Source →
★★★★★
International Standard
CEFR Framework

The global standard for measuring proficiency

“The CEFR provides a transparent, coherent and comprehensive basis for language syllabuses, teaching materials, and the assessment of proficiency.”
7,100 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Noticing Hypothesis

Noticing is the starting condition for acquisition

“Learners must consciously notice linguistic features in the input before they can acquire them. Noticing is the necessary starting condition for acquisition.”
5,600 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Feedback Typology

Explicit correction is most effective

“Recasts and explicit correction are the most common and effective types of corrective feedback in classroom interaction.”
7,200 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Computer-Mediated Communication

Technology lowers the barrier to speaking

“Computer-mediated communication provides a less threatening environment for language practice, increasing learner willingness to communicate.”
3,400 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Pronunciation Research

Pronunciation feedback improves intelligibility

“Pronunciation instruction that includes feedback on specific segmental and suprasegmental features leads to significant improvement in intelligibility.”
2,800 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Motivation in SLA

Intrinsic motivation beats external rewards

“Intrinsic motivation, driven by genuine interest and personal relevance, is a stronger predictor of language learning success than external rewards.”
9,100 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Listening Comprehension

Listening is the foundation of speaking

“Listening comprehension is the foundation upon which speaking proficiency is built. Extensive listening develops the phonological awareness necessary for accurate speech production.”
2,100 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Willingness to Communicate

Less anxiety means more speaking

“Willingness to communicate is the most immediate predictor of frequency of second language use. Reducing anxiety directly increases willingness to speak.”
4,300 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Vocabulary & Proficiency

Vocabulary size predicts proficiency

“Vocabulary size is strongly predictive of reading, listening, speaking, and writing proficiency at every CEFR level from A1 to C2.”
1,400 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Authentic Materials

Authentic scenarios transfer to real life

“Authentic materials and scenarios increase learner engagement and facilitate transfer of language skills to real-world contexts.”
1,700 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Learner Autonomy & CEFR

CEFR self-assessment empowers learners

“Self-assessment tools aligned with CEFR descriptors help learners develop metacognitive awareness and take ownership of their learning progress.”
1,200 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Multimedia Learning

Multimodal learning boosts vocabulary retention

“Multimodal input combining audio, visual, and textual information produces stronger vocabulary retention than any single modality alone.”
18,500 found this helpful Read pillar → Source →
★★★★★
Peer-Reviewed Research
Distributed Practice

Daily practice outperforms weekly cramming

“Regular, distributed practice over time is significantly more effective than massed practice for long-term retention of language skills.”
3,800 found this helpful Read pillar → Source →
The Science

24 research pillars behind fluente.ai

Each pillar connects proven research to a specific fluente.ai feature. No marketing spin — just what the evidence says.

Pillar 1

The Output Hypothesis

You Must Speak to Learn

Prof. Merrill Swain (1985, York University) studied students in Canadian French immersion programmes who had years of listening and reading input. Their comprehension was near-native — but their speaking was far behind. Her conclusion: producing language (output) is essential for developing accuracy and fluency. Input alone is not enough.

This is why fluente.ai puts speaking at the centre of every lesson.
Pillar 2

Acquisition vs. Learning

Conversation Beats Grammar Rules

Prof. Stephen Krashen (1982, USC) established that there are two ways to develop language: acquisition (subconscious, through natural communication) and learning (conscious, through grammar rules). Only acquisition produces spontaneous, fluent speech. Grammar study creates a “monitor” that can edit speech — but it cannot initiate it.

This is why fluente.ai focuses on real conversation practice, not grammar drills.
Pillar 3

The Interaction Hypothesis

Conversation Drives Acquisition

Prof. Michael Long (1996, University of Maryland) demonstrated that interaction — especially the negotiation of meaning during conversation — is one of the most powerful drivers of language acquisition. When a learner struggles to communicate and works through it with an interlocutor, deep learning occurs.

This is why fluente.ai’s AI engages you in real conversation, not scripted dialogues.
Pillar 4

Immediate Corrective Feedback

Correct Now, Not Later

Decades of research confirm that corrective feedback during speaking improves language accuracy. Li’s (2010) meta-analysis found a clear positive effect across studies. Davis (2003) found that 86% of learners prefer immediate correction. The science is clear: correct errors the moment they happen, while the learner’s attention is focused.

This is why fluente.ai corrects you mid-sentence — not after the lesson ends.
Pillar 5

AI Reduces Speaking Anxiety

Practice Without Judgment

Speaking anxiety is the #1 barrier to language learning (Young, 1990). When learners are anxious, Krashen’s “affective filter” activates and blocks acquisition. Recent research (Bashori et al., 2020; Zhai & Ma, 2021) shows that AI-powered corrective feedback provides a less threatening environment that reduces anxiety while maintaining the effectiveness of real-time correction.

This is why fluente.ai gives you a private, judgement-free space to practise speaking.
Pillar 6

Communicative Language Teaching

Meaning Over Grammar

CLT has been the dominant approach in language education since the 1980s. It prioritises meaningful communication and real-world interaction over grammatical accuracy drills. Research consistently shows CLT improves oral fluency and communicative competence across all learner types — including introverted students who typically struggle with speaking.

This is why fluente.ai lessons are real-world scenarios, not textbook exercises.
Pillar 7

Task-Based Language Teaching

Learn by Doing Real Things

TBLT focuses on authentic, goal-oriented tasks that mirror real life: ordering food, conducting interviews, negotiating salaries. Research shows task-based instruction improves motivation, fluency, and communicative competence — especially in workplace contexts. 55% of business English learners most wanted to practise socialising and negotiation (Xie, 2022).

This is why fluente.ai’s 345 lessons are built around 50 real-world professional scenarios.
Pillar 8

Spaced Repetition

Remember What You Learn

Hermann Ebbinghaus (1885) discovered the “forgetting curve” — we lose most of what we learn within days if we don’t review it. Spaced repetition, which schedules reviews at optimal intervals, produces up to 25% higher retention rates than cramming. A large-scale PNAS study (Tabibian et al., 2019) confirmed that algorithmically optimised spaced repetition significantly outperforms other review schedules.

This is why fluente.ai uses spaced repetition to reinforce vocabulary and phrases across lessons.
Pillar 9

CEFR

The Global Standard for Language Proficiency

The Common European Framework of Reference for Languages (2001) is the result of over 20 years of research by the Council of Europe. It provides six proficiency levels (A1–C2) used worldwide by IELTS, TOEIC, Cambridge, employers, and universities.

This is why every one of fluente.ai’s 345 lessons is mapped to CEFR levels.
Pillar 10

The Noticing Hypothesis

Awareness Drives Learning

Richard Schmidt (1990) proposed that learners must consciously notice linguistic features before they can acquire them. When fluente.ai’s AI interrupts and corrects mid-sentence, it forces the learner to notice the gap between what they said and what they should have said — the exact cognitive process that drives acquisition.

This is why fluente.ai’s real-time correction is more effective than end-of-lesson summaries.
Pillar 11

What Learners Actually Want

Our Global Analysis

We analysed learner reviews, search behaviour, and feedback across every major language learning platform globally. The patterns were consistent: learners are frustrated by apps that focus on gamification over outcomes, that teach reading/listening but not speaking, and that don’t prepare them for real professional situations. The most common pain points: “I understand English but freeze when I need to speak”, “I can’t afford a tutor”, “I’m embarrassed to make mistakes.”

This is why fluente.ai exists — to solve the problems every other platform ignores.
Pillar 12

The Blue Ocean

What Was Missing

All of this research — Output Hypothesis, Interaction Hypothesis, corrective feedback, CEFR alignment, learner frustrations — pointed to one gap in the market: affordable, 24/7 AI-powered conversation practice with real-time correction, in professional scenarios, across 50+ languages. That gap is fluente.ai.

This is not a marketing claim. This is what the research told us to build.
Pillar 13

The Noticing Hypothesis

Schmidt (1990)

Richard Schmidt’s Noticing Hypothesis argues that learners must consciously notice linguistic features in the input before they can acquire them. Without noticing, input remains “noise” — present but unprocessed. This is why passive exposure alone (watching TV, listening to music) rarely leads to acquisition of specific grammatical or phonological features.

fluente.ai’s real-time feedback highlights exactly what learners need to notice — turning passive input into active acquisition moments.
Pillar 14

Corrective Feedback Typology

Lyster & Ranta (1997)

Lyster and Ranta’s landmark classroom study identified six types of corrective feedback: explicit correction, recasts, clarification requests, metalinguistic feedback, elicitation, and repetition. Their research showed that explicit correction and recasts are the most common, and that different feedback types lead to different levels of learner uptake and repair.

fluente.ai uses multiple feedback types — from gentle recasts to explicit corrections — matching the most effective strategies identified in this research.
Pillar 15

Computer-Mediated Communication

Warschauer (1996)

Warschauer’s research demonstrated that computer-mediated communication provides a less threatening environment for language practice. Learners who were reluctant to speak in face-to-face settings showed significantly higher participation and more complex language use in digital environments, suggesting technology can lower the affective barriers to language production.

fluente.ai is built on this principle — AI conversation removes the social pressure that prevents learners from practising.
Pillar 16

Pronunciation & Intelligibility

Derwing & Munro (2005)

Derwing and Munro’s research reframed pronunciation teaching around intelligibility rather than native-speaker accuracy. Their studies showed that targeted pronunciation instruction focusing on specific segmental and suprasegmental features leads to significant, measurable improvement in how well learners are understood — which matters more than sounding “perfect.”

fluente.ai’s pronunciation feedback focuses on intelligibility — helping learners be understood, not chasing an impossible native accent.
Pillar 17

Motivation in Language Learning

Dörnyei (2001)

Zoltán Dörnyei’s extensive research on motivation in second language acquisition established that intrinsic motivation — driven by genuine interest, personal relevance, and a sense of progress — is a far stronger predictor of language learning success than external rewards like points, streaks, or badges.

fluente.ai motivates through real progress in real conversations — not gamification tricks that fade after the first week.
Pillar 18

Listening as Foundation

Vandergrift (2007)

Vandergrift’s research synthesis showed that listening comprehension is the foundation upon which speaking proficiency is built. Extensive, purposeful listening develops the phonological awareness, vocabulary recognition, and discourse patterns necessary for accurate and fluent speech production.

fluente.ai integrates listening and speaking in every conversation — because you can’t speak well what you can’t hear well.
Pillar 19

Willingness to Communicate

MacIntyre et al. (1998)

MacIntyre and colleagues developed the Willingness to Communicate (WTC) model showing that WTC is the most immediate predictor of how often someone actually uses their second language. The key factors influencing WTC are perceived competence, anxiety levels, and communicative confidence — all of which can be developed through practice in low-stakes environments.

fluente.ai builds willingness to communicate by providing a zero-judgement space where confidence grows through practice.
Pillar 20

Vocabulary & CEFR Proficiency

Milton (2010)

Milton’s research established clear vocabulary size thresholds for each CEFR level, demonstrating that vocabulary breadth is strongly predictive of reading, listening, speaking, and writing proficiency from A1 through C2. This finding provides a measurable, evidence-based pathway for tracking learner progress.

fluente.ai’s CEFR-aligned curriculum ensures learners build vocabulary systematically at each proficiency level.
Pillar 21

Authentic Materials in Learning

Gilmore (2007)

Gilmore’s comprehensive review showed that authentic materials and scenarios — drawn from real-world contexts rather than simplified textbook dialogues — increase learner engagement and, crucially, facilitate better transfer of language skills to genuine communicative situations outside the classroom.

fluente.ai uses real workplace and life scenarios — not textbook dialogues — so skills transfer directly to where learners need them.
Pillar 22

Learner Autonomy & Self-Assessment

Little (2005)

Little’s work on the European Language Portfolio and CEFR-aligned self-assessment showed that when learners develop metacognitive awareness — understanding what they can do, what they need to improve, and how to get there — they take greater ownership of their learning and make faster, more sustained progress.

fluente.ai provides clear CEFR-based progress tracking so learners always know where they stand and what to work on next.
Pillar 23

Multimedia Learning Principles

Mayer (2001)

Richard Mayer’s Multimedia Learning theory, supported by dozens of controlled experiments, demonstrated that combining audio, visual, and textual information produces stronger vocabulary retention and deeper comprehension than any single modality alone. The key is thoughtful integration — not information overload.

fluente.ai combines spoken conversation with visual feedback and text support — multimodal learning that sticks.
Pillar 24

Distributed Practice Effect

Cepeda et al. (2006)

Cepeda and colleagues’ meta-analysis of 254 studies confirmed that distributed practice — spreading learning over multiple shorter sessions — is significantly more effective than massed practice (cramming) for long-term retention. The optimal spacing interval depends on the desired retention period, but daily practice consistently outperforms weekly sessions.

fluente.ai’s spaced practice system is built on this evidence — short, daily conversations that build lasting fluency.

fluente.ai isn’t built on marketing promises. It’s built on decades of language acquisition research, analysis of every major language learning system, and the real voices of millions of learners who told us what they actually need. The science says speaking works. The learners say they want affordable conversation practice without judgment. So that’s what we built. Become Fluente.

Our position

We've done more research into what works in language learning than any other app, and we built our product on those findings. We studied every major language learning system and decades of SLA research. We analysed learner reviews and search behaviour globally. We built fluente.ai on what the research consistently shows works. The science guided every design decision.

Academic References

Full citation list

28 peer-reviewed sources that inform fluente.ai's design and methodology.

  1. Swain, M. (1985). Communicative competence: Some roles of comprehensible input and comprehensible output in its development. In S. Gass & C. Madden (Eds.), Input in Second Language Acquisition (pp. 235–253). Newbury House.
  2. Krashen, S. (1982). Principles and Practice in Second Language Acquisition. Pergamon Press.
  3. Long, M. H. (1996). The role of the linguistic environment in second language acquisition. In W. Ritchie & T. Bhatia (Eds.), Handbook of Second Language Acquisition (pp. 413–468). Academic Press.
  4. Schmidt, R. (1990). The role of consciousness in second language learning. Applied Linguistics, 11(2), 129–158.
  5. Li, S. (2010). The effectiveness of corrective feedback in SLA: A meta-analysis. Language Learning, 60(2), 309–365.
  6. Davis, A. (2003). Teachers’ and students’ beliefs regarding aspects of language learning. Evaluation & Research in Education, 17(4), 207–222.
  7. Young, D. (1990). An investigation of students’ perspectives on anxiety and speaking. Foreign Language Annals, 23, 539–553.
  8. Lyster, R. & Ranta, L. (1997). Corrective feedback and learner uptake. Studies in Second Language Acquisition, 19(1), 37–66.
  9. Ebbinghaus, H. (1885). Memory: A Contribution to Experimental Psychology. (trans. Ruger & Bussenius, 1913). Columbia University.
  10. Tabibian, B. et al. (2019). Enhancing human learning via spaced repetition optimization. Proceedings of the National Academy of Sciences, 116(10), 3988–3993.
  11. Bashori, M. et al. (2020). Effects of ASR-based pronunciation feedback. Computer Assisted Language Learning.
  12. Zhai, N. & Ma, X. (2021). AI-assisted automated corrective feedback. Computer Assisted Language Learning.
  13. Council of Europe. (2001). Common European Framework of Reference for Languages: Learning, Teaching, Assessment. Cambridge University Press.
  14. Council of Europe. (2018). CEFR Companion Volume with New Descriptors.
  15. Richards, J. C. (2006). Communicative Language Teaching Today. Cambridge University Press.
  16. Willis, J. (1996). A Framework for Task-Based Learning. Longman.
  17. Long, M. H. (2015). Second Language Acquisition and Task-Based Language Teaching. Wiley-Blackwell.
  18. Masuram, J. & Sripada, P. N. (2020). Developing speaking skills through task-based materials. Procedia Computer Science, 172, 60–65.
  19. Warschauer, M. (1996). Computer-mediated collaborative learning: Theory and practice. Modern Language Journal, 80(4), 470–481.
  20. Derwing, T. & Munro, M. (2005). Second language accent and pronunciation teaching: A research-based approach. TESOL Quarterly, 39(3), 379–397.
  21. Dörnyei, Z. (2001). Motivational Strategies in the Language Classroom. Cambridge University Press.
  22. Vandergrift, L. (2007). Recent developments in second and foreign language listening comprehension research. Language Teaching, 40(3), 191–210.
  23. MacIntyre, P. et al. (1998). Conceptualizing willingness to communicate in a L2. Modern Language Journal, 82(4), 545–562.
  24. Milton, J. (2010). The development of vocabulary breadth across the CEFR levels. In I. Vedder (Ed.), Communicative Proficiency and Linguistic Development (pp. 211–232). Eurosla.
  25. Gilmore, A. (2007). Authentic materials and authenticity in foreign language learning. Language Teaching, 40(2), 97–118.
  26. Little, D. (2005). The Common European Framework and the European Language Portfolio. Language Testing, 22(3), 321–336.
  27. Mayer, R. (2001). Multimedia Learning. Cambridge University Press.
  28. Cepeda, N. et al. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354–380.
Frequently Asked Questions

Research FAQ

Is fluente.ai based on real research?

Yes. Every feature in fluente.ai is grounded in peer-reviewed second language acquisition research. We reference 28 academic citations spanning 40+ years of language science, including foundational work by Swain, Krashen, Long, Schmidt, and others.

What is the Output Hypothesis?

The Output Hypothesis, proposed by Prof. Merrill Swain (1985), states that comprehensible input alone is not sufficient for language acquisition — learners must also produce language (output) to develop fluency and accuracy. This is why fluente.ai puts speaking at the centre of every lesson.

How many research pillars does fluente.ai use?

fluente.ai is built on 24 research pillars drawn from decades of second language acquisition research. Each pillar connects a proven scientific finding to a specific fluente.ai feature.

What makes fluente.ai different from other language apps?

Unlike gamification-focused apps, fluente.ai is built entirely on what peer-reviewed research says works: real conversation practice with immediate corrective feedback, spaced repetition, CEFR alignment, and reduced speaking anxiety through AI-powered tutoring.

Does AI really reduce speaking anxiety?

Yes. Research by Bashori et al. (2020) and Zhai & Ma (2021) shows that AI-powered corrective feedback provides a less threatening environment that reduces anxiety while maintaining the effectiveness of real-time correction. Speaking anxiety is the #1 barrier to language learning (Young, 1990).

Explore the Full Research Behind fluente.ai

24 research pillars · 28 academic citations · 40+ years of language science

Start free conversation