For non-native English speakers who work in English

AI now writes most of the English you read at work. Learn to read it fast.

A short daily training program for professionals whose second language is English. Learn how AI-generated text is built, read it faster, and catch the mistakes it hides.

Measure your AI-English reading speed For teams →

Free · 3 minutes · Sign in with an email link · Compared with readers who share your first language

Works with text from ChatGPT, Claude, Gemini, Copilot, and every AI writing tool your colleagues use.

Grade 16.6

Reading level of typical ChatGPT text (Flesch-Kincaid). Non-native human writers in the same study: grade 12.1. [21]

14×

How much longer an AI assistant’s answers were than the developers’ questions, by median token length, in a 2025 study of 82,845 conversations. [24]

10 min/day

The whole program. Practice on text that looks like your job.

You did not ask for this reading load. It arrived anyway.

A pull request description written by Copilot. A ticket summary from the helpdesk AI. A sales call recap from the CRM. Month-end commentary from the spreadsheet assistant. A candidate summary from the recruiting system. A contract summary from the legal AI. A 900-word answer from ChatGPT to a 20-word question.

No person wrote any of it. All of it is in English. All of it is waiting for you before 10 a.m.

You can ask the AI to make it simpler. That is one more step and one more wait — and every rewrite can drop the one detail that mattered. You cannot skip it either, because that detail is somewhere in the middle. So you read all of it.

One document in English is fine. Thirty a day is exhausting. And when a passage is not clear, you switch languages in your head to work it out, then switch back. Every switch costs time and energy. By the afternoon, you read slower than you did in the morning.

The volume of text keeps growing. Your energy does not.

AI English is not normal English. It is harder — and harder for you than for a native speaker.

Researchers have measured this. Compared with human writing, text from large language models has:

Longer sentences with more subordinate clauses.: One idea, three “which” and “although” clauses. [21]
Rarer, more formal vocabulary.: “Leverage.” “Robust.” “Nuanced.” “Delve.” Words a native speaker skips and you have to decode. [21]
Padding.: It restates your question. It adds “it is worth noting that.” It ends with caveats you did not need.
One confident tone for everything.: Correct facts and wrong facts sound identical. The error does not look like an error.

The result: you read slower than you need to, and you trust it more than you should.

Three things you have probably tried

Translation extensions

Immersive Translate, DeepL, Google Translate

They work — today. Tomorrow you read exactly as slowly as before. And a translated hedge is still a hedge: the padding survives translation.

Speed-reading apps

They flash one word at a time and forbid re-reading. An AI sentence with three clauses needs re-reading. These apps were built for novels, not for tickets.

“Just ask the AI to write simpler English”

It works, at a cost. Every document becomes two: the original and the rewrite. You wait, then you read the rewrite — and a rewrite can quietly drop the number, the name, or the condition that mattered.

Four skills. Ten minutes a day. Practice text that looks like your job.

Skim the shape

AI text has a predictable structure: it restates the question, adds headings, delivers the payload, then adds caveats. Learn the shape, and much of the text becomes optional.

Learn AI English as vocabulary

Language models overuse a few hundred words and phrases. You will learn to recognize them instantly, the way you recognize your own name — no decoding, no translating.

Chunk, and stay in English

Read in groups of three to five words instead of one word at a time, and stay in one language while you do it. When a passage is not clear, help is next to the text in your first language. As your speed rises, that layer fades.

Slow down where it matters

Numbers. Names. Claims. Code. Anything you will act on. Practice texts contain planted errors. You learn to find them. Your catch rate is tracked next to your speed.

Practice material built for your work

Choose your field: engineering, product, finance, support, operations, sales. Every session generates fresh text in that domain — tickets, PR descriptions, summaries, reports, chat answers. You never see the same passage twice. Comprehension questions are asked in your first language, so we measure your reading, not your English writing.

One number that means something: effective reading speed

Most speed tests count words per minute and ignore whether you understood anything. We measure words per minute at 80% comprehension or better, on AI-generated work text. That is the only speed that counts.

You also track:

Catch rate: the share of planted errors you noticed
Independence: how much bilingual support you still use
Vocabulary: AI-English words and phrases you now read on sight

Your first benchmark takes 3 minutes. Retest weekly.

Effective reading speed · one learner, ten texts

Filled points cleared the comprehension gate; hollow ones did not. Checkpoints are read with no help at all. The dashed line is what is typical after about twenty texts — not a goal we set for you.

See the eleven readings as a table

Effective reading speed for one learner: a benchmark and ten texts that follow it.
Reading	Words per minute	Comprehension gate	Conditions
Benchmark	128	Passed	Benchmark, no support
Text 1	131	Passed	Practice
Text 2	129	Missed	Practice
Text 3	138	Passed	Practice
Text 4	131	Passed	Checkpoint, no support
Text 5	144	Passed	Practice
Text 6	151	Passed	Practice
Text 7	149	Passed	Practice
Text 8	158	Passed	Practice
Text 9	152	Passed	Checkpoint, no support
Text 10	171	Passed	Practice

The dashed reference line sits at 176 words per minute.

Built for people who read English all day and speak something else at home

Developers: AI-written PR descriptions, review comments, and incident summaries
Support and operations: AI ticket summaries, suggested replies, and knowledge-base answers
Sales: AI call recaps, pipeline summaries, and drafted outreach
Analysts and finance: AI-generated reports, variance commentary, and forecasts
HR and recruiting: AI candidate summaries, job descriptions, and policy drafts
Legal and compliance: AI contract summaries, clause comparisons, and regulatory digests
Product managers: AI meeting notes, spec drafts, and research digests

Recommended level: B1 or above. You can already work in English; you just read it slowly. dub.mom is not an English course. It assumes you know the language and teaches you to read one specific kind of it fast.

Business English for the AI era

Your team’s English training was designed for emails and meetings. Their reading load is now machine-written: longer, denser, and generated faster than anyone can read it.

Domain packs matched to your teams’ work
An admin dashboard showing effective reading speed per team — not vanity WPM
Weekly progress reports for L&D
SSO and per-seat pricing
A twelve-week pilot with a report at the end

Talk to us about a pilot Pilots start at 20 seats.

Start free. Pay when you want the whole program.

Your first ten texts are free, and your chart is yours whether you pay or not.

Free

Your first ten texts, and your record forever.

The 3-minute benchmark, unlimited retests
Ten full sessions
Two easy texts a day after that

Start free

Pro

$12

/month, or $99/year

The whole program, and the tools that come with it.

Unlimited sessions and domains
The fading bilingual scaffold
Planted-error drills and catch-rate tracking

Choose Pro

Program pass

$149

once

For people who would rather finish a program than hold a subscription.

The full program, no renewal
A year of maintenance mode and all tools
The library stays open afterwards

Buy the pass

Teams

Per seat

from 20 seats

Everything in Pro for every seat, plus what procurement asks for.

Domain packs matched to your teams
Admin dashboard: effective speed per team, never vanity WPM
Weekly progress reports for L&D

Talk about a pilot

Full pricing, refunds and regional prices →

Questions people ask first

Is this a speed-reading app?

No. Speed-reading apps train your eyes to move faster over any text. We teach you how one specific kind of text — AI-generated work English — is built, so you can skip what is padding and slow down for what is not. Re-reading is allowed. We measure comprehension, not just speed.

Why not just use a translation extension?

Use one if you need to. But translation makes each page faster without making you faster. dub.mom starts with a bilingual layer and removes it as you improve. The goal is to not need it.

Which first languages do you support?

Spanish, Portuguese, German, French, Japanese, Korean, Chinese and Hindi. The reading skills are the same in every language. What changes is the bilingual layer and the comprehension questions.

What level of English do I need?

B1 or higher. If you can do your job in English but reading it tires you, you are the person we built this for.

Is the practice text real?

It is generated by the same kind of models that produce the text you read at work, in your chosen domain, fresh for every session. Some passages contain deliberate errors. That is the point.

Does it work for text from Claude, Gemini, Copilot, and others — not only ChatGPT?

Yes. The structure and vocabulary patterns are shared across models. We update the training material as the models change.

How long until I notice a difference?

We will publish this from beta data rather than invent a number. What we can say: the benchmark is repeatable, so you will see your own curve from the first week.

Can my company pay for it?

Yes. Most L&D teams treat it as Business English training. See the teams page, or send them this one.

Billing questions · Questions from L&D and procurement

Why the name

Every non-native reader knows the moment. The English sentence is too long, so you dub it into your mother tongue in your head, understand it, and switch back. It works. It is also slow, and by the afternoon it is exhausting.

We named the product after that moment, because that is the moment we want to remove. dub.mom teaches you to read AI English at the speed of your mother tongue — without the dubbing.

Find out how fast you read AI English right now

Three minutes, one text, ten questions. No card — just an email link, so your result has somewhere to live. You will see your effective reading speed and how it compares with readers who share your first language.

Measure your speed