Write one plain English master, translate it with an AI tool using a glossary and placeholders for prices and dates, have a fluent reader of each language mark up the result, then lay it out in Canva and verify every number against one source sheet. AI does the first draft; people own accuracy.
The order matters more than the tool. Schools that translate first and fix later end up with four inconsistent documents and no record of which fee was right. Schools that fix the English, build a glossary and keep numbers out of the translated text can update all four versions in an afternoon next term. A brochure also carries a promise, so a mistranslated course level or a wrong price is not a typo; it is a claim to a prospective student.
Fix the English master before anything is translated
Translation tools magnify whatever you give them. An English master full of idioms ("hit the ground running", "a home from home") will produce translations that are either literal and odd or quietly different in meaning. Rewrite the master so that each sentence has one clear job, avoids phrasal verbs where a plain verb exists, and uses the same word for the same thing throughout.
Take a small school with a one-page flyer and a brochure of eight panels. The brochure has a welcome panel, courses and levels, a sample week, accommodation, activities, fees and dates, how to apply, and contact details. Before translating, I would work through the English with three questions: does every sentence say one thing, is every number or date a placeholder, and could a student with intermediate English follow it? That third question is a useful test because your readers will often compare the translation against the English.
A quick before and after, illustrative:
Before: "Our friendly teachers will help you hit the ground running from day one."
After: "Our teachers help you start speaking in your first lesson."
The "after" is shorter, free of an idiom, and makes a promise you can actually keep. Do this to every panel, and keep a changelog so you can show a checker what you altered from the previous brochure.

The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowBuild a glossary so course names do not drift
Your course names, level labels and brand terms must appear identically across every language, or the translation will treat them as ordinary words. Write a small table before translating anything: the English term, what it means, whether it stays in English or has an approved translation, and a note.
| English term | Treatment | Note |
|---|---|---|
| General English 20 | Keep in English | Product name, as on the website |
| Intensive course | Translate | Agree one term with the checker |
| Levels A1 to C2 | Keep as written | Widely recognised level labels |
| Homestay | Translate | Check it means a host family, not a hotel |
| Exam preparation | Translate, keep exam names in English | Exam brand names never translated |
| Registration fee | Translate | One term, used in every price table |
This table goes into every translation prompt, and later to your native checker. It is also the cheapest way of keeping brochure wording aligned with the website. For the wider habit of keeping terminology consistent, see how translators use AI to keep terminology consistent.
Choosing the four languages from your own enquiries
Pick languages from evidence you already hold: the language each enquirer writes in, or the one they pick on a form. Count a quarter's enquiries and bookings by language and choose the languages that generate both volume and conversion, not just the ones you hope to attract. If you lack data, start with two languages, learn from what the checkers say, and add the others later.
Add one practical test. For each candidate language, can you name a person who can check the translation within your deadline? If not, either delay that language or do not print it. A printed brochure in a language nobody on your side can read is a risk you cannot assess.
Translating with a prompt that protects numbers and names
You can use a general chat assistant or a dedicated translation tool such as DeepL; both are fine for a first draft. Check the vendor's current terms before you paste any student information, and for a brochure you should not need to, because it is public marketing copy. The prompt below is the one I would use, one language at a time.
Translate the brochure text below from English into [language] for adult students aged 18 to 35 choosing a language school.
Tone: friendly but professional. Use the polite "you" form that a school would use with a new student.
Rules:
1. Keep every placeholder in curly braces, such as {FEE_GENERAL_20} or {START_DATE_1}, exactly as written.
2. Use my glossary below. Terms marked "keep in English" stay in English.
3. Do not add, remove or soften any claim. If a sentence is ambiguous, translate it literally and add [CHECK: reason] after it.
4. Keep headings short enough to fit a brochure panel.
Glossary: [paste table]
Text: [paste one panel at a time]
If you would rather use DeepL's own glossary feature than paste the table into a prompt, check the plan first. DeepL's plans page lists one glossary with five entries per language pair on Free and Individual, and 1,000 entries on Team, so the six-row glossary above already outgrows the cheaper tiers. Free also caps text translation at 50,000 characters a month. A quick sum: a 1,200-word brochure is roughly 7,000 characters, so four languages use about 28,000, and a second round of drafts after the English changes would take you past the cap.
Translating panel by panel, rather than the whole document, helps in two ways: you see exactly which panel has a problem, and the model is less likely to drift in tone halfway through. The [CHECK] rule gives the native reader a list of the places the AI itself found doubtful.
An illustrative piece of output for the welcome panel, in Spanish, with the original beside it:
EN: "Learn English in small classes of no more than {MAX_CLASS_SIZE} students."
ES: "Aprende inglés en clases reducidas de un máximo de {MAX_CLASS_SIZE} estudiantes."
That is accurate, but a native reader may note that the informal "Aprende" sets a very casual register, where some audiences expect "Aprenda" or a third-person form. This is exactly the kind of decision no tool can make for you, because it depends on who you are trying to reach. Record the decision in the glossary and apply it to every panel.
What goes wrong when the AI translates unaided
Four failure types recur. The first is the false friend: a word that looks like an English one but means something else in the target language, so "course" or "college" lands as a different kind of institution. The second is register, as above. The third is the silently changed claim: "up to 15 students" becoming "15 students", or "approved by" becoming "accredited by". The fourth is the dropped or reordered number, particularly in tables and sentences with dates.
The fix for the third and fourth is mechanical: after translation, search each output for every placeholder and count them against the master. If the master has 14 placeholders and the translation has 13, a number has gone missing. A short second prompt can do this for you.
Compare the English master and the [language] translation. List every placeholder in curly braces that appears in one but not the other, and any sentence where a quantity, limit or condition ("up to", "at least", "from") has changed. Do not rewrite anything.
A back-translation, asking a second tool to translate the result back to English, is a cheap triage. It tells you where meaning has moved, but it cannot tell you whether the text reads naturally, which is why it supplements the native check and never replaces it. For wider guidance on this stage, read how to check AI translations before customers see them.
The native-speaker check: who, what and how much
Your checker should be a fluent reader of the target language who understands your audience. A former student who now works in the language industry is ideal. A member of staff can work, with the caveat that they are marking something their colleague produced. A freelance proofreader gives you independence and a record.
Give them a package: the English master, the translation, the glossary and a one-page brief. The brief tells them to mark five things, using a colour or a comment tag for each category, so you can triage the replies quickly.
- Meaning: anything that says something different from the English.
- Register: too formal, too casual, or inconsistent between panels.
- Naturalness: reads like a translation rather than a school brochure.
- Terms: deviates from the glossary or from what students would call it.
- Risk: anything that could read as offensive, as a legal promise, or as misleading.
Illustrative effort: a 1,200-word brochure may take a fluent reader about an hour and a half to review thoroughly, and checkers differ on rates, so ask for a quote for a fixed scope instead of an open hourly arrangement. Send all four languages out in the same week and ask for replies by a set date. Give each checker permission to say "rewrite this panel", and agree that their wording wins on questions of naturalness while the English master wins on questions of fact.
When comments return, apply them in the translated file, not in the AI chat, and update the glossary with every term they changed. That glossary becomes your second-edition brochure's first input. If you are weighing the cost of this step against full professional translation, AI translation against human translators: what documents really cost breaks down when each makes sense.
Laying it out in Canva without breaking the text
Build the English version first as the template, then duplicate the pages for each language so the grid, images and styles are identical. Paste the checked translations into those copies. The translation itself does not have to happen inside Canva, and for a brochure it usually should not, because Canva's tool cannot take your glossary or your placeholder rules.
Canva does have its own Translate tool, and it is worth knowing how it behaves before anyone on the team presses the button. Canva's pricing page lists Translate among the AI tools in Pro ($144 a year for one person, as of November 2026), where it does not draw on the monthly AI allowance, subject to fair use; the Free plan gets only "limited access" to Canva's AI tools. Per Canva's help article, Translate covers 100+ languages, works one language at a time, and handles up to 2,000 characters per run. Two settings are on by default: "Duplicate page when translating", which keeps your English page intact, and "Reduce font size to fit", which silently shrinks text that runs long. For a brochure, turn the second one off, or the headings in your longest language end up a point smaller than the rest and nobody notices until print. A third setting, "Mirror page to match text direction", exists for right-to-left languages. If you do use Canva Translate for a quick internal draft, run the placeholder count and the native check on its output exactly as you would on any other machine translation.
Plan for four layout problems.
- Text expansion: many languages run longer than English, so leave 20 to 30 per cent spare room in every text box. Test with the longest language first.
- Fonts: a font that looks right in English may lack accents or whole scripts. Preview every character set, including quotation marks, before duplicating pages.
- Right-to-left scripts: if one of your four languages reads right to left, the layout has to mirror, not just the text. Ask the checker to review the final layout, not only the words.
- Line breaks: no single word should sit alone on a line in a heading, and long compounds can overflow a column. Adjust the box, not the translation.
For another worked example of building print material from checked copy, the printable and QR menu tutorial shows the Canva steps in a different small business, and Canva AI for small business lists which AI features save real time.
Price and date verification: a separate pass, by a different person
Numbers are the most expensive thing to get wrong, so treat them as a separate job. Keep one sheet of truth with every figure in the brochure: fees, deposits, term dates, registration dates, class sizes, hours, accommodation costs, discounts and exam dates. Each row gets a source (your fee schedule, your calendar), a name of who confirmed it and a date.
| Placeholder | Value | Source | Checked by |
|---|---|---|---|
| {FEE_GENERAL_20} | $320 a week | Fee schedule, version 14 | Office manager |
| {REG_FEE} | $60 once | Fee schedule, version 14 | Office manager |
| {START_DATE_1} | 6 January | Term calendar | Director of studies |
| {MAX_CLASS_SIZE} | 12 | Academic policy | Director of studies |
| {HOMESTAY_WEEK} | $210 a week | Host contract | Accommodation officer |
All figures are illustrative. When a brochure is built from this sheet, merge or paste values only after translation, and format them for the language: decimal commas, date order and currency symbols differ, and the checker should confirm the convention for each. A brochure that lists a price must say what it covers, so add a footnote that lists what the fee includes and excludes, translated and checked like the rest.
The final pass is a read-aloud against the sheet: one person reads each figure from the proof, and another ticks it against the source. It feels slow and finds more errors than any other check in the routine.
Two realistic mistakes and how they show up
Both are illustrations of patterns that recur, not reports from a particular school. The first: a school translates its fees panel with the numbers inside the text. In one language the model swaps "$320 a week" to a different currency symbol with the figure unchanged, so the brochure appears to quote a much lower price. Nobody notices in the layout stage because the page looks tidy. It surfaces only when the person doing the read-aloud against the sheet stops at that line. The fix is structural: numbers enter only through placeholders, after translation.
The second: a checker returns the Portuguese text with 40 comments, mostly about register, and the school's director disagrees with half of them. The way out is a rule agreed in advance. Facts follow the English master; wording and naturalness follow the native reader; where two readers disagree, ask a third, or choose the version that matches the website. Without the rule, the comments become an argument and the print date slips.
What a useful checker comment looks like
Ask checkers to give the problem, the suggested wording and a one-line reason, so you learn something each time. An illustrative comment: "Panel 3, heading: reads as a literal translation of 'Course overview'. Suggest 'Our courses'; students would say this. Reason: the original sounds like a university prospectus." Comments in that shape can be applied in minutes and added to the glossary. A bare "sounds wrong" costs you another round.
Print file, PDF file and the page they point to
The two outputs have different jobs. The print file needs bleed, usually a few millimetres of extra background beyond the trim, and a high-resolution export. Ask your printer for their exact specification, including bleed, colour profile and whether they want crop marks, rather than guessing. Order a physical proof for at least one language, because screen colours and text fit differ on paper.
The PDF for email, the website and agent partners should be small enough to open on a phone, with real text rather than flattened images so search and screen readers can use it, and links that work. Name each file with the language and a version, for example brochure-2027-es-v3.pdf, and add the version to the bottom of the last panel so anybody holding a copy can tell whether it is current.
Finally, point every language at a landing page in that language through a short link or QR code. If the page does not yet exist in the language, do not print a link to it. The companion tutorial on marketing to agents and students in several languages covers the landing pages and the emails that carry the brochure.
A worked timeline for a four-language run
An illustrative schedule for a school with one person doing marketing alongside other duties:
- Day 1: rewrite the English master, make the glossary and the number sheet.
- Day 2: translate four languages panel by panel, run the placeholder-comparison prompt, fix gaps.
- Days 3 to 7: native checks return; apply comments and update the glossary.
- Day 8: build the layout in the longest language first, then duplicate.
- Day 9: insert verified numbers, run the read-aloud against the sheet.
- Days 10 to 12: print proof for one language, final PDF for all four, version numbers added.
The slowest part is almost always the checkers, so brief them before you start translating and agree return dates.
What the routine costs, in hours and dollars
Illustrative sums for a four-language, 1,200-word brochure. Your own time: about eight hours across the master, glossary, translation prompts, layout and number checks. Native checks: four reviews of roughly an hour and a half each, so six hours of outside time; if a checker quotes $40 an hour, that is $240, and a quote for a fixed scope may differ. A chat assistant or a month of a paid DeepL plan is a small fee next to that. Canva Free can hold the layout if you paste in the checked text yourself; you need Pro only if you want Canva's own Translate tool or its premium templates.
Set that against a full professional translation of the same four versions, which you may still want for legal terms or visa-related wording. The saving is real but modest on the checking, and large on the speed of the first draft. Next term, with the glossary and number sheet already in place, the same run drops to a few hours because only the changed panels go back to the checkers.
Checking the finished brochure worked
Measure the brochure the way you measure any marketing asset: by what it brought in. Give each language its own QR code or short link so that landing page visits and enquiries can be counted per language. After a term, compare enquiries per language with the quarter before, and read what enquirers ask, because repeated questions often mean a panel is unclear. Keep the checkers' comments file: it is a free training set for the next edition.
If you want help deciding which parts of this routine to automate, and what to keep manual, that is what my AI implementation consultation is for. The sensible first step is usually the glossary and the number sheet, because they make every later edition faster.
Brochure questions language schools ask
Can AI translate a whole brochure well enough to print?
It can produce a strong first draft, often good enough to understand and edit, but not safe to print unseen. Register, idioms and course terms are where it slips. Have a fluent reader of each target language check the whole text before it goes to print or onto your website.
Who should do the native-speaker check?
Someone who reads the target language fluently and ideally knows how prospective students speak: a former student, a teacher on your staff or a freelance proofreader. Give them the English master beside the translation and a short checklist, not just the translation on its own.
How long does a four-language brochure take?
For a short brochure of about 1,200 words, plan roughly a day of your own time for the master, drafts and layout, plus a few days for outside checkers to return their comments. Printing adds the proof round. Build the schedule around the slowest checker.
Should prices be translated by the AI?
No. Keep prices, dates and codes out of the text sent for translation, using placeholders, and insert them from one verified sheet afterwards. A model can quietly change a number or a currency symbol, and a wrong fee in print is expensive to correct.
Do I need a separate print file and PDF?
Usually yes. Print needs bleed, a high-resolution export and often a different colour setup, while the PDF for email and the website should be smaller, keep real selectable text and carry live links. Keep the same text and the same version number across both.
Further reads
- AI Receptionist for a Hostel: Night Calls in Several Languages — Another small business serving guests in several languages.
- How to Offer Multilingual Customer Support With AI Translation — Answer enquiries in the same languages once the brochure brings them in.
- AI for Tour Guides: Multilingual Enquiry Replies and Review Responses — Another small business handling multilingual replies safely.
- Canva for Education vs Pro: Which Plan Fits a Tutor or Nursery? — Check which Canva plan fits a small school.
- Is It Safe to Put Client Documents Into AI Translation Tools? — Decide what may be pasted into an AI translation tool.
- How Small Translation Agencies Build an AI-Assisted Workflow — See how professional translators fold AI drafts into their work.
- Tour Descriptions in Several Languages: An AI Workflow — A repeatable workflow for tour descriptions in several languages: one master, a facts table, AI rewrites, three translation checks and channel versions.
- How Are Language Schools Using AI? Eight Admin Jobs and the Risks — Eight back-office jobs a language school can give to AI, from enquiry replies to homestay matching, with the specific risk and check for each.
- How to Start a Language School: Accreditation, Agents and AI — What a new language school costs in year one, how accreditation and agents fit together, and where AI saves admin time from the first term.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Canva pricing page and Canva's help article on translating designs, DeepL Pro plans page (all checked 4 November 2026).