Can an AI-generated translation be used in official proceedings?

No, not on its own. Official institutions require a translation bearing a signature and stamp, with a clearly identifiable responsible party. AI output can only serve as raw material in a process where a sworn translator reviews it from start to finish and assumes responsibility with their own signature; the final document must in all cases carry the translator's declaration.

Are sworn translation and notarized translation the same thing?

No. A sworn translation is one whose accuracy is declared by a translator who has taken an oath before a notary public, with their signature and stamp. In a notarized translation, the notary's certification is added on top; the notary certifies the translator's signature and authority, not the content of the translation. The institution requesting the document determines which one is required.

Can a translation office issue an apostille certificate?

No. In Turkey, apostille certificates are issued by provincial governorships and district governorships for administrative documents, and by the competent judicial authorities for judicial documents. A translation office cannot issue an apostille; however, it facilitates the process by preparing the translation and notarization steps in the correct order for the apostille chain.

What advantage does the e-Apostille offer?

The e-Apostille allows the apostille certificate to be created electronically and verified online, reducing the circulation of physical paperwork. According to official statements, these documents are recognised as valid in more than 130 countries party to the Hague Convention. Since submitting the document merely as a printout or photocopy may weaken its verification features, it is important to preserve the digital copy.

With AI performing simultaneous translation, why are human interpreters still needed at conferences?

While instant translation models are practical for everyday communication, they cannot assume responsibility for managing context, terminological consistency, cultural nuance or immediate error correction. In court hearings, notarial transactions, arbitration and critical commercial negotiations, the translation has legal consequences, so a professional interpreter who can bear that responsibility remains essential.

What determines the total cost of the translation and certification process?

The language pair, the document's length and terminological density, whether notarization is required, the number of copies and the urgency of delivery are the main factors. Third-party items such as notary and apostille fees are subject to official tariffs and are separate from the translation service. For precise information tailored to your document, you can reach us via our quote form.

The Goncourt AI Confession and Trust in Official Translation

In the first week of November 2026, a wave of AI news from the worlds of literature, technology and media put the question of 'textual integrity' back on the agenda for translation and official document processes.

The first week of November 2026 was unusually eventful for anyone following the relationship between artificial intelligence and the written word. The winner of the Prix Goncourt, France's most prestigious literary award, revealed the day after the prize was announced that he had used AI while writing his novel. In the same week, it emerged that Microsoft had filed a patent application for an AI system that alters the words spoken in a live broadcast depending on the viewer and redraws the speaker's lip movements to match. Meanwhile, a new simultaneous interpretation model was announced in China that begins translating before the sentence is even finished.

At first glance these three stories seem unrelated, yet they share a single question: When we read a text, how do we know who actually touched those words? For literary prizes, this may be an ethical debate; but when it comes to official documents such as diplomas, court rulings, powers of attorney or medical reports, it becomes a matter of trust with direct legal consequences. In this article, we summarise the week's developments and examine what they mean for sworn translation, notarization and apostille procedures.

What happened this week? Three stories, one shared question

The first story came from the literary world. Philippe Jaenada, the 62-year-old author who received the Prix Goncourt on 3 November, explained the following day in response to questions on France Inter how many times and for what purposes he had used AI while writing his novel. And this was not the season's only AI controversy: The author of another novel, Orélien, had initially denied using artificial intelligence; on 21 October, the Canadian publisher Éditions du Boréal confirmed that AI had been used in the book, parted ways with the author and announced that the novel would not be reprinted. The Académie Goncourt had already removed the book from its first longlist on 25 September.

The second story came from the technology front: Microsoft's patent application describes an AI that changes the words spoken in live video and redraws the speaker's lips; the application was published on 29 October 2026, and no patent has yet been granted. The third story came from translation technology itself: The Hy-SimulMT model, developed by Tencent's AI unit Hunyuan, can perform bidirectional, real-time voice translation between Chinese and English, and can begin translating before the speaker has finished talking. At the intersection of these three stories lies a landscape in which the distance between a text's source and the version that reaches the reader is growing increasingly blurred.

The Goncourt confession: Labour and disclosure in literary texts

The real significance of the Goncourt affair lies not so much in whether AI was used, but in the question of disclosure. In the case of Orélien, what led the publisher to withdraw the book was not merely the use of AI, but the fact that it had first been denied. For readers, prize juries and publishers alike, the problem is the same: being misled about whose labour lies behind the text. Literary institutions are therefore announcing increasingly explicit AI policies and expecting transparency from authors.

This debate maps directly onto the world of translation. In the translation of a novel, style, rhythm, cultural references and wordplay are conveyed not by skimming over machine output, but by living with the text for an extended period. It would be no surprise if publishers in literary and artistic translation projects began demanding process transparency from translators as well: Which tool was used at which stage, and who is responsible for the final text? The crisis of trust experienced by awarding institutions is pushing the 'declare your source' rule into every link of the publishing chain.

If words change in a live broadcast: The patent application and textual integrity

Microsoft's application is, for now, only a patent filing; there is no product on the market. But the architecture it describes is striking: An AI agent listens to the audio of a live broadcast, and based on a set of parameters and the viewer's profile, flags the words to be replaced. In other words, the same speech can reach different viewers with different words. An invisible layer of editing inserts itself between what leaves the speaker's mouth and what the listener hears.

For the world of official translation, this is not an abstract dystopia; it is a timely reminder of why the concept of 'document integrity' is protected by such strict rules. In sworn translation, the target text carries the translator's wet-ink signature and stamp together with a declaration of fidelity to the source document; who intervened and what was changed is always traceable. In an era when content can be automatically tailored to the individual, a signature that can say 'this text corresponds exactly to its source, and I am personally responsible for it' is not losing value — it is gaining it.

Translation before the sentence ends: Where is simultaneous interpreting headed?

Tencent's model focuses on the latency problem in simultaneous translation technology. The model also features voice cloning, making it easier for listeners to distinguish who is speaking in meetings and interviews with multiple participants, and the system can translate both from English into Chinese and from Chinese into English. For everyday communication, tourism and quick internal company meetings, the practical value of such tools is beyond dispute.

On the other hand, professional interpreting is not merely the transfer of words. In a court hearing, a notarial transaction, an arbitration session or a sensitive commercial negotiation, the interpreter manages context, tone of voice, legal terminology and the parties' intentions all at once, requests clarification when necessary, and resolves ambiguity on the spot. This is why, at corporate events and in official proceedings, consecutive and simultaneous interpreting continues to be delivered under a model that uses technology as a support tool but keeps responsibility with a human being. AI has entered the booth; but responsibility for the words that leave the booth still rests with a professional.

Turkey's agenda: The BİLGE model and real-time translation in the public sector

In the same week, Turkey's AI and language technology agenda was equally busy. Minister of Industry and Technology Mehmet Fatih Kacır announced the goals of Turkey's 2026-2030 AI Action Plan at TRT World Forum 2026; alongside the target of bringing AI literacy to 5 million citizens, he announced that the first version of BİLGE, the Turkish large language model, will soon be made available to developers. The spread of a large language model focused on Turkish could also influence the quality of Turkish-based translation tools in the years ahead.

Another notable development on the public side came from media regulation: It was announced that the Via+ infrastructure of RTÜK (Turkey's Radio and Television Supreme Council) can continuously monitor more than 800 channels from six satellites simultaneously; still in testing, the system stands out for its capacity to analyse 19,200 hours of speech per day and provide real-time translation in 12 languages. The fact that even state institutions are using AI to scan multilingual content streams shows that translation technology is no longer a niche tool but an infrastructure component. Yet the purpose of these systems is scanning and detection; the official translation of a document with legal consequences is subject to an entirely different regime of responsibility.

Why is the chain of trust in official documents still built on humans?

If we translate the week's news into the world of official documents, the picture is this: technology can generate text, alter it, translate it instantly — but it cannot assume responsibility. A translation chain with official validity in Turkey works roughly as follows: first, a translator who has taken an oath before a notary public renders the document into the target language and signs and stamps it as a sworn translation. If the receiving institution also requires it, this translation is certified by a notary public, becoming a notarized translation. These two procedures are distinct: the sworn translator is responsible for the accuracy of the translation, while the notary certifies the translator's signature and authority.

If the document is to be used abroad, an apostille is added to the chain. The critical point here is this: the apostille certificate is issued not by a translation office or a notary, but by the competent state authorities. According to a statement by Minister of Transport and Infrastructure Abdulkadir Uraloğlu, the e-Apostille system is a digital service that enables official documents bearing an apostille issued by the competent authorities to be created, approved, transmitted and verified electronically, and documents created through it are recognised as valid in more than 130 countries party to the Hague Convention. For administrative documents, provincial governorships and district governorships are competent; for judicial documents, the judicial authorities. To build the translation and certification legs of this chain in the correct order, the apostille translation process should be planned from the outset; an incorrect sequence can lead to the document being rejected in the destination country.

If institutions start demanding 'AI declarations': Recommendations for document holders

A likely institutional ripple effect of the Goncourt debate is the spread of transparency declarations in content and translation workflows. Some foreign universities and publishers are already requiring the disclosure of AI use in academic texts; it would be no surprise if a similar expectation seeped into the translation supply chain. For document holders, this means it will become even more important that who produced the translation and who signed it can be documented.

Before commissioning an official document translation, we recommend the following checklist:

Find out the receiving institution's requirement: Is only a sworn translator's stamp required, or notarization, or an apostille? These are three different procedures, and an unnecessary step means both time and expense.
Clarify the sequence: Whether the apostille is to be affixed to the source document, the notarized translation, or both varies by destination country and institution.
Check the consistency of names and dates: An exact match between the spelling in the passport and in the translation eliminates one of the most common grounds for rejection from the outset.
Preserve digital verifiability: Submitting an e-apostilled or e-signed document merely as a printout or photocopy may render its verification features useless.
Ask who is responsible: If an error appears in the translation, who will correct it, and within what timeframe? Unsigned, unstamped printouts are invalid in official proceedings.

The table below summarises the key differences between unofficial AI translation and the types of official translation:

Feature	Unofficial AI translation	Sworn translation	Notarized translation
Responsible party	None (user bears the risk)	Sworn translator	Translator + notarial certification
Signature and stamp	None	Translator's signature and stamp	Translator's signature + notary seal
Official validity	None	Accepted by many institutions	Mandatory where notarization is required
Apostille eligibility	Not eligible	May not suffice on its own	A core link in the apostille chain
Typical use	Everyday comprehension, drafts	Schools, institutions, application files	Courts, land registry, powers of attorney, procedures abroad

Factors that affect the cost

The cost of these processes is not a single line item and varies from document to document. On the translation side, the main factors determining the amount are the language pair (there is a difference between common and rare languages), the length of the document and its page/character volume, the density of terminology (legal and medical texts require additional expertise), whether notarization is needed, and the urgency of delivery. If the same document is to be used in more than one country, this also affects the planning of copies and certifications.

On the certification side, notary fees are subject to the official tariff, while apostille and e-Apostille procedures follow the current practice of the relevant public institutions; these items are independent of the translation service, and current amounts should be confirmed with the relevant authority. For a precise quote tailored to your document, you can submit it via our free quote form; you will receive personalised information based on language, volume and certification needs. In last-minute situations, the process can be planned on a compressed timeline with our urgent translation option.

Conclusion: Speed belongs to the machine, trust belongs to the signature

The first week of November showed all three faces of AI in the world of text at once: generation (the Goncourt debate), intervention (the word-swapping patent application) and transmission (instant translation models). On all three fronts, technology offers impressive speed; yet on all three fronts, the same gap remains: accountability. Literary prizes close this gap with mandatory disclosure, while the world of official documents closes it with the chain of signature, stamp, notarization and apostille.

As of the publication date of this article, legislation and institutional practice are moving in this direction; however, since both technology and administrative practice change rapidly, we recommend confirming the current requirements with the relevant institution before any critical transaction. At Ziya Translation, we continue to provide sworn, notarized and apostille-ready translation services in a wide range of languages, led by English, German, French, Arabic and Russian, with an approach that treats technology as a tool and human responsibility as the foundation.

From the Goncourt Confession to the Word-Swapping Patent: What AI Week Means for Official Translation