Short answer: A language app is making you more fluent only if you can speak faster with fewer pauses, follow real unscripted speech and handle real conversations better than you could a month ago. XP, streaks and finished lessons measure activity, not fluency. To find out, run three quick self tests once a month: a 2-minute recorded monologue where you count words per minute and pauses, a listening check with unscripted native audio and a real conversation checklist. If your app stats keep climbing while these results stay flat for two or three months, it is time to change how you use the app.
Fluency, Accuracy and App Progress Are Three Different Things
Most learners judge their progress by what the app shows them: a 200-day streak, a league trophy, a finished unit. Those numbers feel like progress, but they track something else. It helps to separate three ideas:
- Fluency is how smoothly you can produce and process language in real time: how fast you speak, how often you stop, and whether you can keep going without help.
- Accuracy is how correct your language is: grammar, word choice and pronunciation.
- App progress is how much you have used the app: XP, streaks, crowns, levels and lessons completed.
The Council of Europe treats fluency and accuracy as separate qualities. In the CEFR table of qualitative aspects of spoken language use, a B1 speaker "can keep going comprehensibly, even though pausing for grammatical and lexical planning and repair is very evident, especially in longer stretches of free production", while a B2 speaker "can produce stretches of language with a fairly even tempo; although he/she can be hesitant as he or she searches for patterns and expressions, there are few noticeably long pauses." Accuracy has its own column. You can be accurate but slow, or fluent but sloppy, and an app can raise your XP without changing either.
App stats are still useful as a habit tracker. If you want help setting sensible targets for minutes, streaks and CEFR levels, read our guide to realistic language app goals. This article is about something different: checking whether all that practice is turning into fluency.
What Research Says Is Worth Measuring
Researchers split spoken fluency into measurable parts, such as speed, pauses and repairs (restarting or correcting yourself). A meta-analysis by Suzuki, Kormos and Uchihara (2021) in The Modern Language Journal pooled 263 effect sizes from 22 studies to see which of these measures match how fluent listeners think a speaker sounds. Perceived fluency was strongly associated with speed and pause frequency (correlations of about .59 to .62), moderately with pause duration (about .46) and only weakly with repair fluency (about .20).
For a self test, that points to two simple numbers: how fast you speak and how often you pause. Self-corrections matter much less to listeners, so do not panic about them. Keep in mind that these are correlations from research settings, and a home test is a progress check, not an official assessment.
Test 1: The 2-Minute Recorded Monologue
This is the core speaking test. It takes about ten minutes including counting.
- Pick a prompt from a fixed list. Write five everyday prompts once and reuse them every month, for example: describe your last weekend, explain what you do at work, tell a story about a trip, describe your home, give your opinion on social media.
- Prepare for 30 seconds only. No notes and no dictionary. The test should show what you can do spontaneously.
- Record exactly 2 minutes on your phone. Keep talking even when you get stuck.
- Transcribe it. Use your phone's dictation or an AI transcription tool, then listen once and fix what it got wrong. Speech recognition can mishear learner speech, as explained in our article on how speech recognition works in language apps.
- Count your words, leaving out fillers like "uh" and "um". Divide by two to get words per minute.
- Count your pauses. Listen again and tally every silence of roughly a second or more, plus every filled pause ("uh", "um", or the equivalent in your target language). Divide by two for pauses per minute. The exact cut-off matters less than using the same one every time.
What improvement looks like: on prompts of similar difficulty, your words per minute go up and your pauses per minute go down. Compare yourself only with your own earlier recordings, not with native speakers or other learners. Do not game the test by rushing: if your recording no longer makes sense, the number is meaningless. Listening back also shows you which words you keep searching for, which is useful material for your next practice sessions.
Test 2: The Unscripted Listening Check
Most app audio is scripted, recorded clearly and spoken at a careful pace. Real speech has overlaps, unfinished sentences, slang and background noise. The CEFR Companion Volume describes this gap in its scale for understanding conversation between other people: at B1 a learner "can follow much of everyday conversation and discussion, provided it is clearly articulated in standard language or in a familiar variety", while at B2 a learner "can keep up with an animated conversation between proficient users of the target language."
How to run it:
- Choose three minutes of unscripted audio made for native speakers: a podcast interview, a street interview video or a casual vlog. Avoid learner podcasts and scripted shows.
- Listen once, without subtitles.
- Write down the main points in your own language: who is talking, what about, and what each person thinks.
- Now read the transcript or turn on subtitles. List the five most important points and count how many you caught on the first listen.
What improvement looks like: you catch more main points on the first listen, you can handle faster speakers or less familiar accents, and you need fewer replays to understand the details. Use a new clip of the same type each month so you are not just remembering last month's answers.
Test 3: The Real Conversation Checklist
Apps cannot fully simulate the pressure of a real person waiting for your answer. In the CEFR qualitative table, a B1 speaker "can initiate, maintain and close simple face-to-face conversation on topics that are familiar or of personal interest." The checklist below turns that kind of ability into yes or no questions you can answer after any conversation of 10 to 15 minutes with a tutor, an exchange partner, a colleague or a friend. Fill it in within an hour, while you still remember.
- I opened and closed the conversation myself.
- I asked at least three follow-up questions.
- I spoke for 30 seconds or more in one turn at least once.
- When I did not know a word, I explained around it instead of switching languages.
- I asked for repetition only a few times.
- The other person did not have to slow down or simplify much.
- I stayed in the target language for the whole conversation.
Score one point per yes, out of seven. If you do not have a conversation partner yet, you can rehearse with an AI speaking tutor such as Enverson AI, but take the test itself with a person when you can, because humans are less patient and less predictable than software.
Your Self Test Toolkit at a Glance
| Test | What it measures | How to run it | Time | What improvement looks like |
|---|---|---|---|---|
| 2-minute monologue: speed | Speaking speed | Record 2 minutes on a fixed prompt, count words, divide by 2 | 10 min | More words per minute on similar prompts |
| 2-minute monologue: pauses | Pause frequency | Tally silences of about a second or more plus "uh/um", divide by 2 | 5 min | Fewer pauses per minute |
| Unscripted listening check | Understanding real speech | 3 minutes of native audio, one listen without subtitles, then check 5 main points | 15 min | More main points caught on the first listen |
| Real conversation checklist | Interaction under pressure | Seven yes or no questions after a 10 to 15 minute conversation | 5 min | Higher score out of 7 |
| App stats (reference only) | Activity and habit | Note minutes, XP or lessons for the month | 1 min | Steady practice, not a fluency measure |
A Monthly Retest Routine
Set a fixed day, such as the first Sunday of each month, and give yourself about 40 minutes:
- Record a monologue on the same prompt as your very first test, plus one other prompt from your list.
- Do the listening check with a new clip of the same type.
- Fill in the conversation checklist after your next real conversation that week.
- Log everything in a simple spreadsheet: date, words per minute, pauses per minute, listening points out of 5, checklist score out of 7 and your app minutes for the month.
- Keep the recordings. Listening to month one next to month four is often more convincing than any number.
How to read your results
Judge the trend over at least three months, since a single test can be thrown off by tiredness or a hard topic.
- App stats up, test results up: your routine works. Keep going.
- App stats up, test results flat: you are probably spending most of your time on recognition exercises. Add more speaking out loud, shadowing and unscripted listening, or consider a different kind of tool. Our Find Your Fit matcher suggests apps by goal and level, and our methodology page explains how we evaluate them.
- Listening up, speaking flat: you are taking in plenty of language but not producing enough. Shift time to speaking practice.
- Speaking speed up, checklist score low: you can talk, but interaction is the weak spot. Practise asking questions and managing turns with real people.
Frequently Asked Questions
Do streaks and XP mean I am getting more fluent?
No. Streaks, XP and finished lessons measure how much you use an app, not how well you speak or understand. They are useful for building a habit, but you need separate tests, such as a recorded monologue and an unscripted listening check, to see whether your fluency is changing.
How many words per minute should I aim for?
There is no universal target, because speech rate depends on the language, the topic and your personality. Compare yourself only with your own first recording. A meta-analysis by Suzuki, Kormos and Uchihara (2021) found that speed and how often a speaker pauses are strongly linked to how fluent listeners judge that speaker to be, so a rising speed and fewer pauses on similar topics are good signs.
How often should I test my fluency?
Once a month is enough. Weekly results jump around with your mood, sleep and topic, so they are hard to read. Look at the trend over three monthly tests before deciding that something works or does not.
Can a self test tell me my CEFR level?
Not officially. Self tests and the CEFR descriptors from the Council of Europe give you a rough idea of where you are and whether you are moving, but only a recognised exam can certify a level.
What should I do if my results do not improve?
Look at how you use the app. If most of your time goes into tapping and matching exercises, add more speaking, shadowing and unscripted listening, or try a tool built around conversation. Then retest after another month.