The arithmetic
There are roughly four hundred million first-language English speakers and somewhere between one and one and a half billion people who use English as a second, third or working language. For anything published on the open internet (and certainly for anything published from or about South Asia, Africa, or most of Europe) the majority of your English-reading audience learned it in a classroom.
They are not less intelligent readers and frequently they have better formal grammar than native speakers. What differs is specific and predictable: idiom, phrasal verbs, low-frequency vocabulary, and culture-bound reference.
Phrasal verbs are the single biggest cost
A phrasal verb is a verb plus a particle whose meaning cannot be derived from the parts: put off, get by, take on, run into, come across, work out, turn down, hold up.
English has thousands of them, they are extremely common in casual writing, and they are individually unguessable. Turn down means reject, reduce, or fold, depending. A reader who knows turn and knows down has no route to any of these. Worse, they are not obviously hard: the words are all short and common, so the sentence looks simple and will not parse.
The fix is not to avoid them entirely, which produces stilted text. It is to prefer the single-word verb where the phrasal one is doing the load-bearing work of the sentence:
- “we had to turn down the order” → “we had to refuse the order”
- “the machine gave out” → “the machine failed”
- “work out the cost” → “calculate the cost”
Four more things that cost disproportionately
- Stacked negation. “Not unlike”, “it is not uncommon for this to fail to work”. Each negation must be resolved, and they compound. Say the positive.
- Culture-bound reference. Cricket, baseball, a 1970s advertisement, a politician, a sitcom. These are invisible to the writer because they feel like general knowledge. If a reference is load-bearing, make the sentence work without it.
- Elegant variation. Native writers are taught not to repeat a word, so a piece calls the same object a lathe, a machine, a tool and the unit. To a second-language reader this reads as four things. Use the same word for the same thing, every time. It is worse prose and much better communication.
- Numbers written as words. “Two thousand four hundred” is harder than “2,400” for almost everyone and much harder across languages with different grouping conventions. Use digits. Also state currencies explicitly:
₹,£,$alone is ambiguous across several countries.
What does not help
Readability formulas. Flesch–Kincaid and its relatives count syllables per word and words per sentence. That is all they count. They therefore reward chopping a clear sentence into three fragments and punish using the correct technical term because it has four syllables. A text can score as “easy” while being incomprehensible, and a well-structured explanatory paragraph can score as difficult while being the clearest thing on the page.
Use them as a rough smoke alarm for drafts that have drifted into 60-word sentences. Do not use them as a target, because optimising a proxy is how you end up with text that is measurably simple and functionally useless.
Simplifying the ideas. This is the thing people do instead, and it is the wrong lever. The difficulty is in the language, not the content. A second-language reader can follow a complex argument perfectly well; they cannot follow come up against.
Translation, machine and otherwise
If you are publishing across languages, machine translation is now good enough to be useful and not good enough to be trusted with anything that matters. Its characteristic failures are exactly the categories above (idiom, ambiguity, and terms with a specific meaning in your field) plus one more: it produces fluent, confident text when it is wrong, so errors do not look like errors.
The free, practical path: machine-translate, then have one fluent speaker read it for sense, not for grammar. That one pass catches the reversals of meaning, which are the failures that actually damage you. Keep a short list of your key terms with their agreed translations, so the same thing does not get three names across a site.
The limitation
Everything here optimises for comprehension, and comprehension is not the only thing writing does. Idiom, rhythm and reference are also how a piece establishes that it comes from somewhere and belongs to someone. Text stripped of all of it reads like an instruction manual, and an instruction manual persuades nobody.
The judgement is about what each piece is for. Safety information, instructions, prices, anything where being misunderstood has a cost: strip it. A personal essay, a piece whose value is voice, work aimed at a community that shares the references: keep the idiom and accept that some readers will work harder. What you should not do is make that choice by accident, which is what happens when you never think about who is reading.
Every lesson here is open. Your progress is recorded once you pass the Making Things with AI exam.
Check your understanding
Why does calling the same object a “lathe”, then “the machine”, then “the unit” cost a second-language reader more than repeating one word would?
Pick the one you would defend. Your answer stays private.
Stuck on something?Ask Meher, your AI tutor