Legal translation — why India has no sworn translator, and what makes a translation acceptable instead

What a legal translation has to survive

An ordinary translation is judged by whether it reads well. A legal translation is judged by something else entirely: whether somebody who cannot read the original will act on it correctly.

That reader is a registrar deciding whether to register a deed, a visa officer deciding whether a marriage is real, an admissions committee deciding whether a degree is what it claims to be, a court deciding what a foreign judgment said. None of them can check the original. All of them will treat the translation as the document.

Three consequences follow, and they are the opposite of what good translation usually means.

It must be literal. An idiomatic rendering that captures the sense is a defect here. If the original says the applicant is the son of a named person, the translation says exactly that, not “whose father is”. Where a term has no counterpart, it is rendered as closely as the language allows and the original term is kept in brackets.

It must be complete. Everything on the page: the heading, the preprinted form text, the marginal notes, the seal, the stamp, the endorsement on the back, the fee stamp, the signature block. Not the parts that seem relevant. A translation that omits a departmental stamp has removed the thing that made the paper official.

Somebody must be answerable for it. The certificate of accuracy is what converts an anonymous document into one with a named person standing behind it. That, and not the quality of the prose, is what the receiving authority is actually relying on.

Five things people mean by “certified”

The word does most of the damage in this subject, because five different requirements travel under it and people quote it to each other without checking which one is meant.

Reading down that table answers most of the questions people arrive with. If a consulate asks for a certified translation, the second and third rows are what it will get and will almost always accept. If a foreign registry asks for a sworn translation, the fifth row cannot be supplied from India, and the conversation has to be about what it will accept instead.

India has no sworn translator

In a number of civil-law countries a translator can be appointed or sworn before a court and given a seal. That translator’s signature on a translation carries official status by operation of law — the German vereidigter Übersetzer, the French traducteur assermenté, the Spanish traductor jurado are all examples. The receiving authority does not check the translation; it checks that the translator was sworn.

India has no equivalent. There is no statute creating a category of sworn or court-appointed translators, no official register, and no seal that carries that meaning. Translators here may be highly qualified and may belong to professional bodies, but no qualification converts a signature into an official act in the way those systems do.

This is not a gap to apologise for; it is a fact to plan around. When a foreign university, registry or consulate asks for a sworn translation, there are only three real answers, and the right one has to be found before any work is commissioned.

• It will accept the Indian equivalent — a certificate of accuracy, notarised, and apostilled or legalised. This is the commonest outcome, and most authorities say so when asked directly.
• It requires a sworn translation done in its own country, by a translator sworn there. Then the Indian document is apostilled here and translated there, and translating in India first is money spent twice.
• It requires a translation from its own consulate’s empanelled list. Then only that list matters, whatever the quality of anybody else’s work.

All three are ordinary. What is not ordinary is discovering which one applies after the translation has been paid for, which is what happens whenever the question is not asked at the start.

The certificate of accuracy

Because nothing in Indian law prescribes its form, the certificate of accuracy is a document whose entire value lies in what it happens to say. A vague one is the commonest technical reason a file is returned, and it takes five minutes to get right.

Two habits make it more useful. Attach the certificate physically to the translation and to a copy of the original, and page-number the set as a whole, so that nobody can later wonder which original a translation belongs to. And keep the translation’s layout close to the original’s, so that a reader comparing them side by side can see that they correspond without reading either.

Where the translation is going into a court file or a registration, the person who signed should be contactable and willing to depose to it if it is questioned. A certificate signed by somebody who cannot be produced is worth very little at the moment it is challenged.

What a notary adds, and what it does not

Section 8 of the Notaries Act, 1952 lists the functions a notary may perform, and among them is to translate, and verify the translation of, any document from one language into another. So a notary may in terms verify a translation.

In practice what happens is narrower and it is worth being precise about it, because people rely on the wrong thing. The usual act is that the notary attests the translator’s signature on the certificate of accuracy, having satisfied himself of the identity of the person signing. That is an authentication of who signed. It is not, and does not purport to be, an examination of whether the Malayalam or the Mandarin was rendered correctly.

Understanding that changes how the step is used. Notarisation is not quality control, and a notarised translation is not a checked translation. What notarisation does is give the document a signature that the next link in the chain recognises — which matters enormously, because the apostille and the embassy legalisation both operate on the notary’s signature and not on anything else.

Two practical points follow. The notary’s details must be legible and his register entry number present, because that is what is verified later. And notarisation should be the last step before the onward chain, not an afterthought added after a correction — a translation corrected after notarisation is a fresh document and needs the step again. Our notary attestation guide deals with the domestic side in detail.

Where the apostille fits

For a document going to a country party to the 1961 Hague Convention, the apostille replaces the older chain of consular legalisation. The essential point for this page is a narrow one, and it is constantly misunderstood: an apostille certifies the authenticity of the signature, the capacity in which the signatory acted and, where appropriate, the seal or stamp — not the content of the document beneath it.

So an apostilled translation is not an officially verified translation. It is a translation whose notary has been verified. Nobody in the chain has read the two languages against each other. That is precisely why the certificate of accuracy carries the weight it does, and why an authority that genuinely wants the language checked will insist on its own panel instead.

For countries outside the Convention the chain is longer — authentication and then legalisation at that country’s mission — and the mission’s own requirements about who may translate become decisive. Our apostille page sets out both routes, which categories of document are covered, and the State-level step that comes before the national one; there is no purpose repeating it here.

What belongs here is the interaction between the two processes, and that is the next section, because the order in which they are done decides whether the work has to be done twice.

Translate first, or attest first?

Both orders are used, both are correct somewhere, and choosing without asking is how files come back.

The common pattern — attest, then translate. The original document is authenticated and apostilled in India, and the translation is then made of the document together with the apostille and every endorsement on it. The recipient gets a single set in which everything visible is rendered. This is usually right where the destination will accept an Indian translation at all, because it leaves nothing on the page that the reader cannot read.

The reverse — translate, then attest. The translation is made and certified first, the notary attests it, and the notarised translation is itself apostilled. This is what some authorities want, and it produces a translation that carries its own chain of verification rather than riding on the original’s.

The third pattern — attest here, translate there. The Indian document is apostilled in India and sent abroad untranslated, to be translated in the destination country by a translator that country recognises. Where a sworn translation is genuinely required, this is the only route that works, and doing an Indian translation first achieves nothing.

There is no way to reason a way to the right answer from first principles, because it is not a question of logic but of what a particular office has decided. The only reliable method is to ask, and the next section is about how.

Ask the destination before anybody starts

This section is short because the advice is simple, and it is here because ignoring it accounts for more wasted money in this area than every other mistake combined.

Before commissioning any translation for official use, get the requirement in writing from the office that will receive it — the consulate’s document checklist, the university’s admissions page, the registry’s instruction sheet, or an email reply. Four questions answer almost everything.

• Will you accept a translation done in India with a certificate of accuracy, or do you require a sworn translation done in your country?
• Do you accept only translators on your own panel, and if so, where is the list?
• Do you want the apostille before the translation or after?
• Do you need the translation bound with the original, and do you want the original or a copy?

Where a consulate maintains a panel, that fact is usually published, and it is absolute. A translation from outside the panel is refused without being read, and no amount of certification changes it. This is not unreasonable of them — it is the only way an office that cannot read every language can exercise any control over accuracy — but it has to be known in advance.

We ask these questions as the first step of any translation instruction, and where the answer is that only their own panel will do, we say so and decline the work rather than take a fee for something that will be rejected.

Registration: Section 19 and the true copy

For documents used inside India, the clearest statutory requirement is in the Registration Act.

Three features of that section repay attention. The test is conjunctive — the officer must not understand the language and it must not be commonly used in the district — so a document in a language current locally does not attract it. The consequence is refusal, not a defect to be cured later. And what has to accompany the document is two things: a true translation and also a true copy. People routinely bring the first and not the second.

Section 62 then deals with what follows: the translation is transcribed into the register-book, and the translation and the copy are filed in the registration office, with endorsements made on the original recording that the document has been registered together with a copy and a translation. The practical effect is that the translation becomes part of the permanent record, which is a reason to be careful about its quality even where nobody at the counter is going to check it.

The same logic extends beyond registration. Wherever a document in one language is being handed to an office that works in another — a municipal body, a school, a bank — a translation plus a copy is the form that gets accepted, and it costs nothing to prepare both.

The language of the courts

The practical consequence is straightforward. A document in a regional language, or in a foreign one, that is to go before a High Court or the Supreme Court has to be accompanied by an English translation. Court rules set out the form, and they differ between High Courts, but the requirement itself follows from the Article.

Subordinate courts are a different matter, because the State prescribes the language in which they work, and it is frequently the regional language. So a document may need translating into the State language for the trial court and into English for an appeal from it — and the two translations need to correspond, because a discrepancy between them becomes an argument on appeal.

Article 348(3) deals with statutes rather than proceedings: where a State legislature has prescribed a language other than English for its Bills, Acts, Ordinances, orders, rules or bye-laws, a translation of the same into English published under the Governor’s authority in the Official Gazette is deemed to be the authoritative text in English. The corresponding provision for Central Acts is Section 5 of the Official Languages Act, 1963.

Hindi judgments, and the English translation that must follow

The proviso to Article 348(2) keeps judgments in English even where a Governor has authorised Hindi for proceedings. Section 7 of the Official Languages Act, 1963 is the route by which that changes, and it carries a condition that is the point of this section.

So even in the States where this has been authorised, an English translation follows the judgment — and crucially, it is a translation issued under the authority of the High Court, not one a party arranges. That distinction matters when a judgment has to travel: for an appeal, for enforcement elsewhere, or for production before a foreign authority, the court’s own translation is the one to obtain.

Section 3 of the same Act is the broader backdrop, providing for the continued use of English in addition to Hindi for the official purposes of the Union and for business in Parliament. It is the reason the whole system runs bilingually rather than having settled on one language, and the reason so many Indian documents exist in two versions whose correspondence nobody has ever checked.

Which raises the question the next section deals with: when two versions exist and they differ, what happens?

Proving a foreign-language document

A document in a language the court does not work in is not self-proving merely because a translation is attached. In substance the original is produced, a translation accompanies it, and the accuracy of the translation must be capable of being established — which in a contested matter means the person who made it may be required to depose to it.

Because the detailed requirements are set by the rules of the particular court and differ between them, the safe approach is procedural rather than clever. Have the translation done by somebody who can be produced. Have the certificate of accuracy signed by a named individual and not merely stamped by an office. Keep the original available, because a translation of a photocopy of an original nobody can find is a weak thing to build on. And prepare the translation in a form capable of being sworn to in an affidavit if the other side disputes it.

Where a translation is likely to be contested — a disputed contract term, a foreign judgment whose scope is in issue, a statement whose exact words matter — two further habits help. Render contested terms literally and put the original term in brackets beside the rendering, so the court can see the word being argued about. And avoid consolidating repeated formulae, because a party will eventually point to the place where the original said something slightly different.

None of this makes a translation unchallengeable. It makes it defensible, which is all a document can be.

Transliteration — the mistake that costs the most

If this page achieves one thing, it should be this. More translations are rejected over the spelling of a name than over every substantive error put together.

A name is transliterated, not translated. It is carried across scripts, never rendered by meaning. A name that means “lotus” does not become Lotus. A patronymic is not converted into a surname. A title or honorific attached to a name is reproduced, not dropped because it seems decorative.

And the governing rule for how it is spelled is not phonetic accuracy. It is this:

The same discipline applies to every other proper noun on the document: place names as the official record spells them, the name of the issuing institution as it writes its own name in English, department names as they appear on their own letterhead. A district whose spelling changed a decade ago should be rendered as the document has it, with the current form in brackets if clarity demands.

Where a script does not map cleanly — Devanagari, Gurmukhi, Urdu and Tamil all raise this in different ways — the translator should be told the passport spelling at the outset rather than asked to choose. It is the single piece of information that most improves the outcome, and it is almost never volunteered.

When the documents themselves disagree

Sometimes the problem is not the translation. The birth certificate says one name, the school leaving certificate says another, the passport a third, and the translation of each is faithful to what it says. Now there are three names and one person.

The rule here is absolute, and it is the opposite of helpful instinct. The translation must not harmonise them. Each translation reproduces exactly what its own original says, including the discrepancy. A translator who quietly corrects a name to match the passport has altered a document — which is both a false rendering and, where the file is going to a government office, a considerably more serious problem than an inconsistency.

The gap is bridged outside the translation, by a document whose whole purpose is to bridge it. Where the variants are the same person recorded differently, an affidavit that both names refer to one and the same person is the standard route, and it should list each document and the exact form the name takes in it. Where the name was actually changed, the name change record is what establishes the link, and in many files the gazette publication is what the receiving authority is really looking for.

Two practical notes. Prepare the bridging document before the file is submitted, not after the query arrives, because a query answered weeks later has usually already cost a slot. And check dates of birth with the same care as names — a date written in one convention and read in another is the second-commonest mismatch, and it is invisible until somebody notices that a person was apparently born in the thirteenth month.

Birth, death and marriage certificates

These are the documents we translate most often, and they are deceptively simple — short, but unforgiving, because every field on them is checked against something else.

A civil registration certificate is a form. The translation should reproduce the form: the preprinted field labels as well as the entries, the register number, the date of registration as distinct from the date of the event, the name of the registering authority, and the signature and seal block. Where a field is blank on the original, the translation shows it as blank rather than omitting the row — an absent entry is itself information.

Three fields cause most of the trouble. Names, for the reasons set out above. Dates, which should be rendered unambiguously — spelling the month out removes the entire class of day-month confusion at no cost. And place names, which should follow the official spelling and, where a village or district has been renamed or reorganised, keep the original with a bracketed note rather than silently updating.

For visa and immigration files these certificates are usually needed with the full chain: the original certificate, its apostille or legalisation, and the translation. Whether the translation covers the apostille depends on the order chosen, which is why the order question is settled first. Where a certificate is old, handwritten or faded, a fresh certified extract from the registering authority is often quicker to obtain than an argument about an illegible entry, and it makes the translation straightforward.

Degrees, marksheets and transcripts

Education documents carry a rule of their own, and it is a rule about restraint.

A translation of a degree, marksheet or transcript reproduces what the document says: the qualification as named, the institution as it names itself, the subjects, the marks or grades as printed, the grading scale where the document prints one, the dates, and the seals and signatures. It renders the words. It does not do arithmetic.

What is legitimate, and useful, is rendering the explanatory text the institution itself prints — the key to its own grading scale, the note on its own reverse explaining what a division means. That is part of the document, and leaving it out is an omission.

Two practical points. Transcripts and consolidated marksheets are frequently issued by the university directly to the evaluator in a sealed envelope, and where that is required, a translation arranged separately by the student does not substitute for it — check which the destination wants before commissioning anything. And where a degree is in a regional language but the marksheets are already in English, translate only what is not, rather than producing a redundant set that invites questions about why the English original was re-rendered.

Company and commercial documents

Business documents crossing a border have their own pattern, and the risk sits in a different place — not in the language but in whether the translated document still proves what it was meant to prove.

A board resolution translated for a foreign bank has to show, in the target language, that the meeting was held, that a quorum was present, that a named person was authorised, and for what. If the translation renders the operative resolution but summarises the preamble, the bank cannot see the authority. Our board resolution page deals with what the resolution itself must contain; the translation’s job is to carry all of it across.

The same applies to a power of attorney going abroad, where the scope of the authority is the whole document and a compressed rendering narrows or widens it; to incorporation and good-standing documents, where the exact corporate name including its suffix matters and must never be adapted to the target country’s conventions; and to financial statements, where the notes carry the qualifications and translating only the primary statements produces a misleading set.

Three habits are worth building in. Keep a single glossary for a client so that the same corporate term is rendered the same way in every document in the file, because inconsistency across a bundle reads as carelessness. Never convert currencies or units — render the figure as printed, with a bracketed note of the unit if the symbol is ambiguous. And keep the document’s own numbering and clause structure, because the recipient will cross-refer to clause numbers.

Bilingual contracts and the language clause

Where a contract exists in two languages, a question arises that no amount of good translation prevents: if the two texts differ, which one governs?

The answer should be in the contract. A language clause names the governing version and says that in the event of any conflict or inconsistency that version prevails. Without such a clause, both texts have an equal claim to be the agreement, and a dispute about what was promised becomes a dispute about translation — fought with expert evidence, at length, over a sentence.

Two refinements are worth knowing. Where one party genuinely cannot read the governing version, a clause saying so and recording that the other version was provided for convenience only is honest and usually accepted; a party that later says it did not understand what it signed has that clause put to it. And where a contract is to be performed and enforced in different countries, the governing-language clause should sit alongside the governing-law clause and the dispute-resolution clause and be consistent with them — an English-governing text with a dispute forum that works in another language guarantees a translation exercise at the worst moment.

For statutes rather than contracts the position is already settled and needs no clause: Article 348(3) makes the English translation published under the Governor’s authority the authoritative text of a State enactment, and Section 5 of the Official Languages Act, 1963 does the same for Central Acts. Where a party cites a regional-language version of a provision, that is the text to go back to.

Medical and technical documents

Medical records, hospital summaries, laboratory reports and technical certificates are translated for visa files, insurance claims, compensation matters and court records. They carry two risks that ordinary documents do not.

The first is terminology. A clinical term rendered approximately can change what the document appears to say, and the safe practice is to keep the original term in brackets beside the rendering wherever there is any doubt, so that a reader who knows the field can see what was written. Drug names, dosages, units and reference ranges are reproduced exactly and never converted.

The second is the boundary of the job, and it is worth stating plainly. A translation says what the document says. It does not say what the document means. We do not interpret a finding, do not comment on whether a report supports a claim, and do not summarise a record into a conclusion. Whether a result means what somebody hopes it means is a question for a doctor, and in a claim or a case it is a question for an expert whose opinion is given as evidence.

For handwritten clinical notes — still common, and often barely legible — the honest approach is to render what can be read, mark what cannot as illegible, and say so in the certificate. A translator who guesses at a handwritten dosage has created a risk out of nothing. Where the illegible part matters, a typed transcript or a fresh summary from the hospital is the right fix, and it is usually obtainable.

How to check a translation you cannot read

Most people commissioning a translation cannot verify it, and they assume that means they can only trust or not trust. That is not quite true — four checks work without knowing a word of the other language, and between them they catch most of what gets files rejected.

• Layout. Hold them side by side. A one-page certificate should produce roughly one page. A document with a table should produce a table with the same number of rows. Large structural differences mean something was summarised.
• Numbers. Check every digit — dates, registration numbers, roll numbers, marks, amounts, page numbers. Numbers survive translation unchanged, so any difference is an error, and this check needs no language at all.
• Seals and signatures. Count them on the original and find each one accounted for in the translation, whether rendered or described in brackets. A missing seal is the most common serious omission.
• Your own name. Letter by letter against the passport, including initials and their order.

A fifth check is available where the stakes justify it: have a different translator produce a back-translation into the original language, and compare it with the original. It will not match word for word and is not meant to — what it reveals is anything that has changed in substance, a dropped clause, a reversed negative, a confused party. For a contract or a judgment that is going to be relied on, it is cheap insurance.

If any of the four checks fails, the fix is a corrected translation and a fresh certificate, and where the document had already been notarised, a fresh notarisation. That is a reason to run the checks before the attestation chain starts rather than after.

Where translations go wrong

• Not asking the destination first, and discovering the panel requirement after paying.
• Translating when a sworn translation was required, which India cannot produce.
• The wrong order — translating before an apostille that then has to be translated too.
• A name spelled better than the passport spells it.
• Harmonising a discrepancy between documents inside the translation instead of bridging it outside.
• Summarising instead of translating, especially preprinted form text and preambles.
• Dropping a seal, a stamp or a reverse endorsement.
• Converting currencies, units, dates or grades.
• Claiming equivalence for a qualification, which is not the translator’s function.
• A vague certificate of accuracy that does not identify the document.
• Correcting after notarisation and sending the corrected version with the old attestation.
• Only the translation filed for registration, without the true copy Section 19 also requires.
• No language clause in a bilingual contract.

What we do, and what it costs

Translation starts at ₹900 and ordinarily takes 1 – 3 days, with the timeline set by the attestation chain rather than by the translation where a document is going abroad.

The first step is not translating. It is establishing what the receiving office actually requires, because that determines the form, the order and sometimes whether the work should be done in India at all. Where the answer is that only their own panel will be accepted, we tell you and do not take the instruction.

Government, notarial and consulate charges are at actuals. Nothing is payable in advance. We do not interpret medical or technical findings, and we do not assess or claim equivalence for qualifications.

If you can only answer one of these, make it the first. Everything else on this page follows from what the destination says it wants.