The Best Chinese Learning Apps in 2026: An Honest Comparison (From Someone Who Built One)

Every few months someone asks me which Chinese app they should use, and every few months I give the same slightly annoying answer: it depends what you mean by "app", because the seven tools everyone names are not really competing with each other. Most of them are not even doing the same job.

So this is my attempt at the version of this list I wish existed. Not a ranking, because ranking a dictionary above a course is like ranking a hammer above a Tuesday. Just what each one is genuinely for, what it does better than anything else, and the exact point where it stops taking you anywhere.

Cards on the table before we start, because it matters here: I build one of the apps on this list. Hanzi Express is mine. I have tried to write the rest of this as if it were not, and I have been deliberately hard on my own tool in its own section, but you should weigh everything below knowing who is typing it. My other credentials are less flattering: a 500-day Duolingo streak that never turned into reading, about a year of building my own Anki decks, and three years living between Shanghai and Hong Kong finding out the hard way which of this stuff survives contact with an actual menu.

Why most "best Chinese app" lists are useless

They rank tools that do unrelated jobs against each other.

There are really five separate jobs in learning Chinese, and almost no app does more than two of them well. Looking things up. Learning characters so they stay learned. Learning grammar and how sentences go together. Reading real text. Producing the language, out loud or by hand.

Once you see the jobs, the arguments stop making sense. "Pleco or Duolingo" is not a question. "Anki or HelloChinese" is not a question. The real question is which two or three tools cover the jobs you currently need, and that changes depending on where you are.

Here is each one, by job.

Pleco: the one everybody should have installed by tonight

Job: looking things up.

Pleco is the Chinese dictionary. Not a Chinese dictionary. It has been the answer for well over a decade and nothing has come close to displacing it.

The core app is free, and the free version alone is better than anything else in the category. Look up by pinyin, by character, by handwriting if you draw it with your finger, or by pointing your camera at a sign and letting the OCR read it. That camera feature is not a gimmick. Standing in front of a menu you cannot read and getting an instant breakdown of every character on it is the single most useful thing any of these apps has ever done for me.

The paid add-ons are one-off purchases rather than a subscription, which in 2026 feels almost radical. You buy the better dictionaries once and they are yours. Budget somewhere in the tens of dollars if you want the good ones, and check the current prices in the app rather than trusting a number in a blog post.

Where it stops: Pleco is a reference, not a teacher. It has a flashcard add-on and it is a perfectly decent one, but it is a flashcard system bolted onto a dictionary. It has no opinion about what you should learn or in what order, which is the same limitation Anki has, for the same reason.

Verdict: install it. There is no argument to have here. It is free and every other tool on this list works better with it sitting next to them.

Duolingo: a habit machine that is not really teaching you Chinese

Job: showing up every day.

I wrote a whole piece on this after my 500-day streak, so I will keep it short. Duolingo is superbly engineered to get you to open it again tomorrow, and that matters more than purists like to admit, because the people who end up reading Chinese are the ones who did a little every day for a couple of years.

The problem is the format. You spend most of your time tapping word tiles out of a bank into the right order, which is recognition wearing a costume. For most languages you get away with that. For Chinese the gap between spotting a character in a lineup and pulling it out of nothing is enormous, and Duolingo lives almost entirely on the easy side of it. It also has no model of how characters are built from reusable parts, so they just pile up undifferentiated.

Free, with the caveat that the old hearts became an energy meter, so the free tier now has a tap limit before it nudges you to wait, watch an ad, or pay.

Where it stops: the day you want to read something real with no tiles underneath it.

Verdict: fine as a starter, fine as a side habit. Not a plan.

HelloChinese: the beginner course Duolingo is not

Job: grammar and sentences, for beginners.

If what you actually want is a structured beginner course in one app, this is the better pick and it is not particularly close. HelloChinese was built for Mandarin specifically rather than being a general language platform with a Chinese module bolted on, and it shows in the parts that matter. Pinyin gets taught properly instead of being assumed. There are real tone drills with speech recognition that will tell you when your third tone is wrong. Grammar gets explained in actual notes. Character writing practice is built in.

It is freemium, with a premium tier that runs somewhere around 70 US dollars a year at the time of writing.

Where it stops: it is a course, and courses have an end. It takes you through roughly the beginner and lower intermediate stretch and then you are out the other side. Character retention along the way is also still largely incidental, in the sense that the app is teaching you sentences and hoping the characters come along for the ride. For a lot of learners they partly do and partly do not.

Verdict: the best single-app starting point for a beginner who wants to be taught rather than to build their own system.

Anki: a superb engine with no car around it

Job: remembering things you have already decided to learn.

I have written the long version of this one. The short version is that Anki is one of the best pieces of learning software ever made and it is the wrong first tool for almost every beginner.

It is a spaced repetition engine and nothing else. It hands you a blank, very powerful machine and then leaves the room. What you learn, the order you learn it in, how the card is even built, all of that is your problem. And a good Chinese card is not a word and a translation, it is the character, the pinyin with tone marks, the meaning, audio, and ideally stroke order on top. Building a few thousand of those by hand is how most people quit.

Free everywhere except iPhone, where a one-off payment of around 25 US dollars bankrolls the entire project.

Where it stops: at the blank screen on day one.

Verdict: outstanding as a second tool once you are intermediate, especially for mining words out of shows and books that no course was ever going to teach you. Rough as a first one.

Skritter: the only one that takes handwriting seriously

Job: writing characters by hand, from memory.

Skritter is narrow and best in class at the narrow thing. You write the character stroke by stroke on the screen and it corrects you as you go, including when your stroke order is wrong rather than only when the finished shape is wrong. That distinction is the entire point. Tracing a shape teaches you to recognize it. Producing it from nothing, in the right order, is a different skill, and this is the tool for it. It carries HSK 3.0 decks across all nine levels.

It is a subscription, in the region of 15 US dollars a month or about 100 a year.

Where it stops: it does one job. It is not going to teach you grammar and it is not a reading tool.

Verdict: essential if you genuinely want to write by hand. The harder question, which most people avoid, is whether you actually do. Handwriting is a real time cost and for plenty of learners typing is enough. Answer that honestly before you subscribe.

Du Chinese: the step almost everyone skips

Job: reading real text at a level you can survive.

There is an ugly gap between finishing flashcards and reading anything a Chinese person actually wrote, and graded readers are the bridge across it. Du Chinese is the most polished one: levelled stories with native audio, tap any word for an instant definition, and a difficulty ladder running from absolute beginner up to genuinely hard.

Freemium, with a subscription for the full library.

This is the tool people add about two years too late. Reading is where characters stop being flashcards and start being language, and you can start far earlier than feels reasonable, as long as the text is pitched right. If you want to check whether something you found elsewhere is at your level, we have a free text difficulty checker that will tell you.

Where it stops: it assumes you already know some characters. It is a reading tool, not a teaching one.

Verdict: add it earlier than feels comfortable.

Hanzi Express: mine

Job: learning characters so they stay learned.

This is the one I build.

Hanzi Express is an all-in-one Chinese character learning system. It brings spaced repetition, stroke order practice, speed review, and more together in one place, so a character gets reinforced from several angles instead of one flashcard flipped over and over. Answers are checked by AI, so a correct synonym gets accepted instead of marked wrong for not matching one deck author's exact phrasing.

You can start for free.

It is not everything, and it is not trying to be. It is not a dictionary, so keep Pleco next to it. It is not a grammar course, so if you cannot yet build a sentence, HelloChinese or an actual textbook is doing a job this does not do. It is not speaking practice, and it will not fix your tones in conversation.

Verdict: if characters keep evaporating no matter how many hours you put in, this is what I built to fix that specific problem.

What about just using ChatGPT

This comes up constantly now, so it deserves a straight answer.

The AI chatbots are genuinely excellent at one thing: explaining whatever confused you, the moment it confused you, at whatever length you need, without judgement. Why does this sentence use 了 twice. What is the difference between 二 and 两. Rewrite this paragraph three levels simpler. That is real value, and it used to require a tutor.

They make a bad curriculum, though, and it is worth being precise about why rather than hand-waving. A chatbot is reactive. It does not know you have confused second and third tone eleven sessions running, because it is not tracking you at all. It will not push you, will not correct a mistake you did not ask about, and will not decide what you should study on a tired Tuesday. It will occasionally produce Chinese that is subtly off with complete confidence. And if your main pronunciation model is synthesised speech, your accent drifts toward the synthesiser rather than toward a person.

Verdict: a brilliant on-demand explainer sitting next to a real system. Not the system.

If you are actually taking the HSK

None of the apps above will tell you whether you would pass. They teach, they do not assess, and that gap is real. Sitting one full mock under exam conditions is worth more than another week of vocabulary, because it tells you which section is quietly costing you the most.

Same disclosure as before, this one is also mine: NexaProva runs full HSK mock exams graded by AI across every section including writing. If you are still working out which level to aim at, start with what actually changed in HSK 3.0, then the HSK 1 and HSK 2 guides.

The stacks that actually make sense

Forget rankings. Pick a stack.

Complete beginner who wants to be taught: Pleco, plus HelloChinese, plus Duolingo only if the streak is genuinely the thing keeping you sitting down.

Beginner whose real goal is reading: Pleco, plus Hanzi Express for the characters, plus Du Chinese the moment the easiest stories are survivable.

Sitting an HSK exam: the level guide for your target, plus whatever is teaching you characters, plus one real mock exam before you book anything.

Wants to write by hand: add Skritter, and accept that it is a separate skill with a separate time cost.

Intermediate and stalled: this is where Anki finally earns its place, mining the specific words that keep tripping you up in real material.

One rule across all of them: no more than three at once. I have watched more people fail from running six apps badly than from picking the wrong one.

The honest bottom line

There is no best Chinese app, which I realise is a deeply unsatisfying thing to hear from someone who makes one.

What there is, is a set of tools that each do one job properly. Pleco is free and non-negotiable. Something has to teach you sentences. Something has to make characters stick, in an order that respects how they are built. Something has to get you reading real text. If all four boxes are ticked and you are still showing up most days, you are going to get there.

The thing that actually kills people is not choosing wrong. It is the two-year cycle of trying an app for three weeks, hitting the plateau, blaming the app, and switching. Every single tool here works better in month six than in week three. Pick the stack that covers your jobs, and then, and this is the whole trick, stay put.