I tested ChatGPT Translate against Google Translate — here's which one wins
Looking for the best AI polyglot
OpenAI's quiet release of a standalone ChatGPT Translate tool for the web lays down a gauntlet against one of the most popular Google tools ever. But can ChatGPT's models perform as well as the much more mature Google Translate?
ChatGPT Translate offers a familiar dual‑pane interface, with users entering a source language on the left and receiving a rendered translation on the right in more than 50 languages. Like Google, it can detect what language you are starting with, and it will even offer AI‑infused flourishes such as style adjustments and tone presets.
On the other hand, Google Translate can handle text, images, documents and website links, strengths built up over nearly two decades. But when it comes to their core abilities, the stage is set for an undoubtedly fierce face‑off between the two tools.
I set up my own, informal head-to-head contest to see how they did with some of the commonly called upon translation needs, aided by a couple of multi-lingual friends for verification.
Everyday conversation
The first task involved a breezy bit of everyday English, using both informal spelling and a couple of idioms to represent how people commonly speak when not being formal for translation purposes. I wrote, “I’m gonna bounce once I grab some grub,” into the two translators and asked to have it rendered in Spanish.
Google Translate wrote back, "Me voy en cuanto coma algo," which gets the meaning across as it literally means "I'll leave as soon as I eat something." Accurate, but without any of the informal or hasty implications of "gonna grab" and "grub."
Meanwhile, ChatGPT went for the subtly different "Me voy a ir en cuanto agarre algo de comer," literally "I'm going to leave as soon as I get something to eat."
Sign up for breaking news, reviews, opinion, top tech deals, and more.
According to the native speaker I asked, it could also mean "grab" instead of "get." That sounds a lot closer to the underlying meaning of the original, though both are accurate enough for casual use.
Technical jargon
The second test moved far away from casual slang and into the thicket of corporate jargon. German seemed appropriate for this test. I asked for a translation of a barrage of buzzwords: “We need to align stakeholder expectations during the initial sprint planning to mitigate downstream scope creep.”
It is also the kind of writing that demands clarity from a translation tool, although clarity in the original language is maybe a little spotty, too…
ChatGPT Translate wrote, "Wir müssen die Erwartungen der Stakeholder während der initialen Sprint-Planung abstimmen, um spätere Umfangsänderungen zu vermeiden." According to reverse translation and a handy German speaker, that means, "We need to align stakeholder expectations during the initial sprint planning to avoid scope changes later on."
No mention of creep or that the size will go up. Both 'stakeholder' and 'sprint planning' are basically the same words in both languages according to ChatGPT, though.
Google Translate came back with "Um eine Ausweitung des Projektumfangs im weiteren Verlauf zu vermeiden, müssen wir die Erwartungen der Stakeholder bereits in der ersten Sprintplanungsphase aufeinander abstimmen," It's a longer sentence than ChatGPT wrote, but translates back to English almost the same: "To avoid an expansion of the project scope later on, we need to align the stakeholders' expectations during the initial sprint planning phase."
Along with a different way of writing sprint planning, the main difference is that Google got the 'expansion' element right.
French literary stylings
I came up with a lyrical little paragraph about Paris to test with French. I wrote, "The first time I visited Paris, it rained nonstop. The narrow streets shimmered with reflections, and the scent of damp stone clung to the air. But I didn’t mind. Something about the quiet made me feel like I had stepped out of time, like the whole city was pausing just long enough for me to notice it."
ChatGPT Translate rewrote that as, "La première fois que j’ai visité Paris, il a plu sans arrêt. Les rues étroites scintillaient de reflets, et l’odeur de pierre humide flottait dans l’air. Mais cela ne me dérangeait pas. Il y avait quelque chose dans ce silence qui me donnait l’impression d’avoir quitté le temps, comme si toute la ville faisait une pause juste assez longtemps pour que je la remarque."
A very direct translation would be "The first time I visited Paris, it rained non-stop. The narrow streets shimmered with reflections, and the smell of damp stone hung in the air. But I didn't mind. There was something about the silence that made me feel like I had stepped outside of time, as if the whole city had paused just long enough for me to notice it."
Barring some word choices about scent and smell and clung versus hung, the only notable change is that the French language requires a somewhat more formal sentence construction. Otherwise, it's basically identical, though you could quibble over how much those little differences matter.
Google Translate's answer looked quite different in French. It rewrote my paragraph as, "La première fois que j'ai visité Paris, il a plu sans interruption. Les rues étroites scintillaient de reflets, et l'odeur de la pierre mouillée flottait dans l'air. Mais cela ne me dérangeait pas. Ce calme ambiant me donnait l'impression d'être hors du temps, comme si la ville entière s'était arrêtée un instant, juste assez longtemps pour que je puisse l'admirer."
Reverse translated by humans and machines, this says, "The first time I visited Paris, it rained without interruption. The narrow streets shimmered with reflections, and the smell of wet stone hung in the air. But I didn't mind. The surrounding calm gave me the feeling of being outside of time, as if the entire city had paused for a moment, just long enough for me to admire it."
Again, it seems 'clung' and 'scent' are English terms that don't survive the journey into French and back, at least through these services. Google also chose wet over damp. Interestingly, Google chose "admire" instead of "notice" in the last sentence for some reason. And while ChatGPT rewrote "quiet" as "silence," Google went for "calm," a much more emotion-laden word.
All of the tests demonstrated that there's no major reason to choose ChatGPT Translate or Google Translate barring personal preference. Both can clearly handle day-to-day language, even with some idioms flattened or odd word choices shading the meaning a little bit.
ChatGPT could be a little more literal in some verb choices, but I'd argue that Google Translate was better at being both precise and briefer in its translations. Google Translate was also perhaps a second faster, not so much as to be really noticeable, but it stands out when doing direct comparisons.
ChatGPT Translate is sure to improve as it sees more use, but if I'm traveling to places where I don't speak the language, I think I'll stick to Google to help me navigate safely.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.

Eric Hal Schwartz is a freelance writer for TechRadar with more than 15 years of experience covering the intersection of the world and technology. For the last five years, he served as head writer for Voicebot.ai and was on the leading edge of reporting on generative AI and large language models. He's since become an expert on the products of generative AI models, such as OpenAI’s ChatGPT, Anthropic’s Claude, Google Gemini, and every other synthetic media tool. His experience runs the gamut of media, including print, digital, broadcast, and live events. Now, he's continuing to tell the stories people want and need to hear about the rapidly evolving AI space and its impact on their lives. Eric is based in New York City.
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.