Spectacular demos are popping up everywhere: a speaker talks in English, and a few seconds later a synthetic voice repeats it in French. So the question is on everyone’s lips: are interpreters going to disappear? I prefer to look at the facts rather than at promotional videos.
What we call “machine interpreting”
These systems chain three building blocks: speech recognition (turning speech into text), machine translation, then speech synthesis or subtitling. Progress is real and fast. National and international institutions are already studying their use, as noted by the University of Mainz laboratory specialising in these technologies (University of Mainz).
And these tools have useful applications: making a meeting accessible in a rare language for which no interpreter is available, live subtitling for people who are hard of hearing, following a video in an unknown language.
What recent research shows
Researchers have taken up the subject. In 2026, entire issues of scientific journals are devoted to the sustainability of the interpreting profession in the age of AI (IJLTIC, 2026).
A study published in January 2026 compared human interpreting and AI interpreting of a press conference on climate. The method: 56 journalists, split into two groups, followed by a comprehension test (full study). Other studies compare models such as GPT-5 with human interpreters on political speeches, to find out whether the machine also conveys the relational functions of discourse, and not just the information (IATIS).
What this research questions is less “does the machine translate the words?” than “does the audience understand and remember the message?”. And that is exactly where the value of interpreting lies.
Where machines still stumble
- Implicit meaning: irony, humour, diplomatic innuendo, cultural metaphors.
- Imperfect speech: unfinished sentences, hesitations, strong accents, several people talking at once.
- Handling the unexpected: a speaker who skips a section, an ambiguous term, an off-topic question.
- The human relationship: in a negotiation, a training course or a sensitive interview, the presence of a trusted professional matters.
- Accountability and confidentiality: who answers for an error? Where does the data from your confidential meeting end up?
Towards an augmented profession
For the profession, AI is also a tool. Software already helps interpreters in the booth by displaying figures, proper names and technical terms live, which improves accuracy (LANS-TTS). I use these tools to prepare my glossaries faster, and I keep a close eye on those arriving in the booth.
So, do you have to choose?
It all depends on what’s at stake. To follow an informative webinar in a rare language, an automatic tool may be enough. For a negotiation, a training course, a general meeting, or an event where your image and your relationships are at stake, a human interpreter remains the guarantee of a message that is understood, nuanced and faithful.
AI will not replace interpreters. But interpreters who understand AI will be better able to advise their clients. That is what I strive to do.
Wondering whether machine interpreting is right for your event? Let’s talk, no taboos: I’ll tell you honestly what works and what doesn’t.
