Gliglish
Gliglish is an AI language teacher you speak to out loud in 38 spoken language variations: hold free conversations, roleplay real-life situations, and get instant grammar, pronunciation and translation feedback. Some 295,861 learners have used it.
What is Gliglish?
Gliglish is a web application for practising a foreign language out loud. You speak into your microphone, the AI answers by voice, and the exchange continues as a real conversation across 38 spoken language variations covering 29 languages, with an interface available in 34 locales. The premise comes from its creator: people learn to speak by speaking, the way children do, rather than through the text-heavy habits of classroom courses. The free version is deliberately open. It runs with no sign-up at all and gives 10 minutes of speaking per day, conversations capped at 50 messages and reduced availability at peak hours. Gliglish Plus lifts those ceilings: unlimited speaking time, unlimited conversation length, 50 topics, mini-lessons, a beginner mode and priority access at busy times. Inside a session the tool does more than answer back. It corrects grammar as you go without cutting the thread, explains any word in context, translates on request, shows transliterations for Chinese, Japanese, Korean, Thai and Greek, suggests what to say next, and lets you slow the AI down. Every word is clickable to be heard. Pronunciation feedback is scored, but it is labelled beta and covers American English only. Multilingual speech recognition lets a beginner ask a question in their own language. Free satellite products sit alongside it: Gliglish Music, Gliglish Comics, Minimal Pairs, a phrasebook and Gliglish Translate. Two organizational offers exist. Education addresses schools and universities and Business addresses companies; both add an admin dashboard, usage tracking, account management and report exports, with custom scenarios for large deployments. The engine is described as running on the same technology behind ChatGPT, with OpenAI, Microsoft and Groq named as subprocessors for the language model, speech-to-text and text-to-speech. The publisher claims 295,861 users, 231 countries and territories, 4.6 million conversations and 10 million minutes of speech, plus six independent academic studies run in classrooms between 2023 and 2025. It is published by Click & Speak Pte. Ltd. in Singapore and built by one person, Fabien Snauwaert, who has been making English learning tools since 2007 and launched Gliglish in 2023. Access is by browser, with a native iPhone and iPad app; on Android, only an Add to Home Screen web app.
What it does
- Hold a free-flowing spoken conversation with an AI teacher in 38 language variations
- Roleplay real-life situations such as ordering at a restaurant, a job interview or a hotel check-in
- Correct grammar in real time without breaking the thread of the conversation
- Explain any word in context and translate a word or a sentence on demand
- Score the pronunciation of a sound, a word or a sentence (beta, American English only)
- Suggest what to say next when the learner runs dry, and adjust the AI's speaking speed
- Run themed mini-lessons on greetings, telling the time, numbers and prices, or personal information
When to use Gliglish / When not to
A quick filter to help you decide if Gliglish is the right fit.
When to use Gliglish
- Adult learners who can already read and write a language but freeze when they have to speak it
- School and university students practising conversation, the setting covered by the six independent academic studies published on the site
- Language teachers and institutions that need an admin dashboard with usage tracking and activity exports, through the Education plan
- Companies and international teams rehearsing work situations, with usage analytics on the HR side, through the Business plan
- Anyone preparing for a job interview or a trip abroad who wants to rehearse set scenes such as ordering at a restaurant or a hotel check-in
When not to use Gliglish
- Developers and product teams hoping to embed speaking practice: there is no public API and no developer documentation
- Learners of any language other than American English who expect scored pronunciation feedback, a feature still in beta and limited to US English
- Anyone looking for a structured grammar curriculum, a diploma or a graded assessment: the terms state the service is provided for entertainment purposes only
- Android users who want a native application, since the help pages only document an Add to Home Screen web app
- Professionals who would need to discuss confidential material, as the publisher explicitly asks users not to share sensitive or confidential information in conversations
How to use Gliglish
A typical end-to-end flow, from setup to results.
- Open the free page on gliglish.com in any browser: no account and no card are needed to start
- Check your microphone first if in doubt, using the site's mic test page and its microphone help article
- Pick the language and the regional variation you want to practise from the grid
- Choose a scenario: free conversation, one of 50 conversation starters, a real-life roleplay or a mini-lesson
- Press the Record button and speak, ideally in a quiet place with no background noise or music
- Use the suggestions whenever you run out of things to say
- Click any word to hear it, have it explained in context or have it translated
- Adjust the AI's speaking speed if it goes too fast or too slow for you
- Ask your question in your own language thanks to multilingual speech recognition, which helps beginners
- Install the iOS app from the App Store, or on Android use Add to Home Screen from the main menu; subscribe to Gliglish Plus from the pricing section and manage or cancel it from the billing page
Pros & Cons
Pros
- A genuine test with no sign-up and no card: the free page is usable immediately
- Six independent academic studies published with figures, including a 23% fluency gain over eight weeks and a 75% improvement in overall score over six weeks
- Broad language coverage: 38 spoken variations and an interface in 34 locales
- Complete, versioned legal documentation: terms, privacy policy and data processing agreement, with archives going back to 23 December 2022
- Five subprocessors published with their role and country, which is rare at this size, and primary data hosted in the United Kingdom and Ireland
- Far cheaper than one-to-one tutoring, the site claiming twelve times less, with institutional pricing from USD 4.97 per user per month
- Nothing to install beyond a browser, plus a native iOS app, and the subscription can be cancelled in a few clicks from the billing page
Cons
- No refunds of any kind: all purchases are stated to be final and non-refundable, and buying waives the right to a refund
- The publisher reserves the right to have humans review conversations and voice recordings, and explicitly asks users not to share sensitive or confidential information
- The stance on model training is never stated, neither in the privacy policy nor in the data processing agreement, and no opt-out is documented
- The free plan is heavily constrained: 10 minutes a day, 50 messages per conversation and reduced availability at peak hours
- Pronunciation feedback, one of the headline features, is limited to American English and still labelled beta
- No Article 27 EU representative and no data protection officer are named despite the GDPR claim, no security certification such as SOC 2 or ISO 27001 is held, and all five subprocessors are US companies while primary data sits in the United Kingdom and Ireland
- A one-person operation, with no public API and no named third-party integration, and terms that allow suspension or termination without notice or explanation
Pricing & Plans
A permanent free plan is available and requires no account: it allows 10 minutes of speaking per day, caps each conversation at 50 messages and offers reduced availability at peak hours. The cheapest paid entry point observed is EUR 16 per month on the Gliglish Plus annual plan, billed EUR 190 per year; the monthly plan stands at EUR 29 per month. Pricing is regional, so the currency and the amount displayed depend on the visitor's country, the press kit announcing a starting point of $29 per month with regional pricing. There is no time-limited free trial, and all purchases are final and non-refundable.
- 10 minutes of speaking per day
- conversations capped at 50 messages
- reduced availability at peak hours
- no sign-up required
- EUR 29 per month
- that is EUR 348 per year
- with no commitment
- EUR 16 per month billed annually
- that is EUR 190 per year
- a stated saving of EUR 158 or 45%
- Both Gliglish Plus plans include unlimited speaking time
- unlimited conversation length
- 50 topics
- mini-lessons
- beginner mode
- multilingual speech recognition
- new features as they ship and priority access at peak hours
- and can be cancelled at any time
- USD 9.97 per user per month on a 3-month commitment
- USD 5.97 on 12 months
- USD 4.97 on 24 months
- USD 9.97 per user per month on a 3-month commitment
- USD 5.97 on 12 months
- USD 4.97 on 24 months
- Organizations can download a formal PDF quote covering name
- contact
- VAT or company registration number and billing address
- pricing is regional
- so amounts and currency vary by country
Data, GDPR & hosting
A consolidated view of how Gliglish handles your data.
GDPR overview
Compliance is claimed explicitly and documented concretely. The privacy policy states compliance with the GDPR and the UK Data Protection Act 2018, and FERPA compliance is claimed for US institutions. A full data processing agreement is public and readable before purchase, and takes effect automatically when an organizational subscription is bought. It incorporates the EU standard contractual clauses (Decision (EU) 2021/914, Module Two, controller to processor), names Irish law and the Irish Data Protection Commission for Clause 17, and relies on the IDTA or UK Addendum with the ICO for UK transfers. Breach notification is capped at 72 hours, five subprocessors are listed with their role and country, and data subject rights can be exercised in the interface or by email. Two gaps remain: no Article 27 EU representative and no data protection officer are named, and no security certification is claimed.
Who owns the data?
Click & Speak Pte. Ltd. (Singapore, UEN 201735018E) controls the data collected from individual accounts: name, email address, voice recordings and their transcriptions, the AI's replies, reference texts used for pronunciation scoring, translations, conversations, IP address, usage times and device identifier. The terms transfer no intellectual property over what learners produce, but the privacy policy reserves the right to review conversations and voice recordings, human review included, to improve the service. It is shared with essential service providers, five of which are named, and where law or service integrity requires it. Under an Education or Business subscription the organization becomes controller and Click & Speak its processor.
Reuse rights
Nothing in the terms claims ownership of, or a licence over, what a learner produces: there is no assignment clause and no restriction on reusing your own conversations, transcripts or translations as you see fit, including for teaching. What is framed is the use of the service itself. Accounts are strictly personal, sharing one with students or colleagues is forbidden, and holding several accounts is not allowed. The content policy bans hate, harassment, violence, self-harm, sexual content, political influence and campaigning, spam, deception and malware, and a breach can end the account without compensation. Press-kit assets are released explicitly: all images may be used freely in articles and videos about Gliglish. Learners can view, manage and delete their own data in the interface, or by email on request.
Data retention & training
Hosting summary
Primary data is stored on servers located in the United Kingdom and Ireland. The publisher, Click & Speak Pte. Ltd., is established in Singapore and acts as data importer under the standard contractual clauses. Five subprocessors are named, all US-based: Microsoft Corporation (database, web servers, asset and backup storage, speech recognition, language model, text-to-speech), OpenAI Inc. (language model, speech-to-text, text-to-speech), Amazon Web Services Inc. (assets, backups), Heroku Inc., a Salesforce subsidiary (database, cloud processing, backups), and Groq Cloud (language model, speech-to-text, text-to-speech). Transfers out of the EEA rely on the EU standard contractual clauses (Decision (EU) 2021/914, Module Two), with Irish law for Clause 17 and the Irish Data Protection Commission as competent authority; UK transfers rely on the IDTA or UK Addendum, under the law of England and Wales with the ICO. Stated security measures include encryption in transit and at rest, access controls, pseudonymization, HTTPS/SSL, regular backups, logging and regular security assessments. No hosting or security certification is claimed, and no broad hosting region is named: only those two countries.
Things to keep in mind
Risks and trade-offs to weigh before adopting Gliglish.
- Voice recordings and transcriptions are kept and may be reviewed by humans to improve the service; the publisher itself tells users not to share sensitive or confidential information in conversations
- The stance on model training is never stated and no opt-out is documented, so you cannot know whether your speech feeds future models
- The terms warn that the AI may occasionally produce incorrect or misleading information and produce offensive or biased content, and position the service as being for entertainment purposes only: do not lean on it for anything critical
- Educational risk: replacing human practice entirely with an obliging AI can fossilize mistakes it never flags, and can flatter a fluency that does not transfer to a real conversation
- Purchases are final and non-refundable, and the terms allow suspension or termination without notice or explanation, with loss of access to the content attached to the account
- The AI subprocessors are US companies (OpenAI, Microsoft, Amazon Web Services, Heroku, Groq) even though primary data sits in the United Kingdom and Ireland, and continuity of service depends on a single-person publisher
- Minors: no minimum age is set, parental permission is required under 13 in the United States, and under an organizational subscription the institution sees named usage metrics such as time spent and frequency
Setup & Integrations
Technical difficulty
Very low. Nothing is installed: the site states that no installation is required and that any web browser works (Chrome, Firefox, Safari, Edge), and the free version opens without an account or a card. You need an internet connection, a computer, tablet or phone, and a microphone. The only real friction is granting the browser microphone permission, for which dedicated mic test and microphone help pages exist, plus finding a quiet place. iOS users install the App Store app; Android users add the site to their home screen. Institutions need no LMS integration: a browser link is enough.
Deployment
Apps stores
Supported languages
Behind Gliglish
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Gliglish.
Frequently asked questions
Do I need an account to try Gliglish?
How many languages does Gliglish cover?
Is there a free plan, and what are its limits?
How much does the paid version cost?
Is there a mobile app?
Can I get a refund?
Where is my data hosted, and can anyone listen to my conversations?
Is there an API?
Is there an offer for schools and companies?
Is there any evidence that it works?
Should you pick Gliglish?
Gliglish does one thing and says so plainly: it makes you speak. It is not a full curriculum, it offers no certification and no graded assessment, and the terms describe the service as provided for entertainment purposes only. Judged on that narrow promise, it holds up unusually well. Six independent academic studies run in classrooms between 2023 and 2025 give it evidence that products of this size almost never publish, and the free page lets anyone test the claim in minutes, with no account, no card and no installation. Legal transparency is above average too: a public data processing agreement, five named subprocessors with their role and country, versioned archives of every document since December 2022, and primary hosting in the United Kingdom and Ireland. The reservations are real and worth weighing before paying. Nothing is refundable, and the terms say so in as many words. The publisher reserves the right to have humans review conversations and voice recordings, asks users not to share sensitive information, and never states whether that material trains models; no opt-out is documented. No Article 27 EU representative and no data protection officer are named despite the GDPR claim, and no security certification is held. Pronunciation scoring, one of the headline features, covers American English only and remains in beta. Behind it all stands a single person, Fabien Snauwaert, through Click & Speak Pte. Ltd. in Singapore, which is both the charm of the product and its main continuity risk. For an adult who reads a language but freezes when speaking it, or for a school looking for cheap conversation volume, the honest move is simple: spend ten minutes on the free version before committing anything.
- Choosing a selection results in a full page refresh.
- Opens in a new window.