top of page
SubscribePopUp

Can AI perform multilingual SEO translation and localization?

Author: Adriana Stein

an image of author Adriana Stein, accompanied by various search-related iconography, including a calendar, scatter plot, and globe icon

While artificial intelligence has the potential to make you a more effective SEO or digital, using it for translation and localization in search engine optimization comes with business considerations that you need to take seriously.


Website owners carrying out international SEO campaigns have a responsibility to present linguistically and legally accurate, helpful information, regardless of what country they’re in or what language they speak.


And although Google has responded to how it perceives AI-generated content, AI is still limited when it comes to generating correct information. That means that when you use AI for direct translation, it’s absolutely necessary to have a human edit and fact check the output.


Otherwise, you could publish content that’s not only wrong, but that gets flagged by a search engine—not to mention it won’t engage and convert customers in your international target regions.


So, the question remains:


Where and how does AI fit into multilingual SEO localization?


As a multilingual SEO and marketing specialist myself (and someone who is currently learning my fourth language), I’ll share my perspective and concrete examples generated by AI.



Table of contents:


What AI is used for translation?


Nowadays, when people refer to AI for translation, they’re typically referring to generative AI technologies like ChatGPT.


“ChatGPT is fine-tuned from GPT-3.5, a language model trained to produce text. ChatGPT was optimized for dialogue by using Reinforcement Learning with Human Feedback (RLHF) – a method that uses human demonstrations and preference comparisons to guide the model toward desired behavior.” OpenAI, the creator of ChatGPT

Other generative AI tools are also available, and they work much in the same way (meaning that they are susceptible to the common pitfalls I’ll discuss later).


While Google Translate now includes some AI-powered features, it’s nowhere near as flexible as a tool like ChatGPT. Nevertheless, I’ll compare ChatGPT output with Google Translate to show you how far AI has (or hasn’t) come for multilingual SEO.


How accurate are AI translations?


This is how Mohammad Omar, CEO of AI data management firm LXT, describes AI’s value for translations:


a quote from Mohammad Omar that says “Grammar provides structure to language, and the free components that make up most of its meaning are its vocabulary. But there are translation and localization challenges that go beyond the structure of language itself. These include figures of speech, sarcasm and irony, compound verbs like ‘look up’ or ‘get over,’ which may not translate to other languages, as well as words with multiple meanings or no equivalent in another language. Each of these can be more easily solved with today’s AI technology.”

While he describes the nuances of languages perfectly, he’s missing a crucial point:


Language nuances are where a real person is needed the most because they are so highly specific. In many cases, AI tends to fail miserably because it doesn’t account for subject context and cultural relevance. If brands want to maintain linguistic and legal accuracy (the basics for successfully operating in international markets), the context, environment, tone, and intent simply can’t be copy/pasted from auto-generated text.


Language is an area that’s rather “subjective,” because people say things in many different ways. Even within the same language, dialects, regional differences, slang, and other components of speech can be drastically different.


As a consultant that’s translated thousands of texts from German to English (and who leads a team of translators that does the same in 30 languages), I can wholeheartedly confirm that true multilingual SEO success is never a copy/paste job.


Although AI translation can generate relatively decent outputs to help get you started (more on this below), it still 100% requires human oversight.


From an SEO perspective, a successful process requires that someone that is fluent in the target language and someone that understands the business review any content to ensure that the correct intent and information are carried across from language to language.

To put it simply: from a multilingual SEO and content perspective, AI tools can give you a head start, but you should never solely rely on them.


Can AI translation replace human translation?


As more AI platforms roll out, many of us digital marketers are wondering the same thing: Will AI replace human translators in the multilingual SEO process?


To bring us closer to an answer, there are a few elements to consider.


  • Translation, localization, and content adaptation processes are context-specific and must adhere to various legal regulations: The way that product names, titles, phrases, and concepts need to be stated for one website is likely entirely different for another. For example, there are vastly different regulations across different countries surrounding medical and healthcare messaging. If you copy/paste AI outputs here, you could end up a quick step away from a lawsuit.

  • The generalization and lack of personality behind basic, AI-generated translations just isn’t good enough right now: Global companies that target specific regions can’t use the same approach everywhere. If you do this, native-language SEO competitors will always beat you, because they will always engage the local audience better. Plus, this could get you in trouble with Google. Using 100% AI-generated content still poses a risk for manual penalties, largely due to the question of quality, accuracy, or potential search engine ranking manipulation. Churning out thousands of AI-generated blog articles daily—regardless of the language—will negatively pique Google’s interest.

  • Google expects content to demonstrate E-E-A-T: In the case of AI translations, it’s extremely important to maintain that first “E” (which stands for “experience”) in E-E-A-T by injecting actual human experience into the output. This consideration takes on another level of complexity when it involves a different language. When it comes to SEO content localization, this is always a must because one region’s audience is vastly different from another.


Let’s also consider this from the view of large language models (LLMs), the process used in ChatGTP for translation. Here’s an abstract from “Dissociating language and thought in large language models: a cognitive perspective,” an academic paper by researchers at the University of Texas at Austin, MIT, and UCLA:


“Today’s large language models (LLMs) routinely generate coherent, grammatical and seemingly meaningful paragraphs of text. This achievement has led to speculation that these networks are—or will soon become—‘thinking machines’, capable of performing tasks that require abstract knowledge and reasoning. Here, we review the capabilities of LLMs by considering their performance on two different aspects of language use: ‘formal linguistic competence’, which includes knowledge of rules and patterns of a given language, and ‘functional linguistic competence’, a host of cognitive abilities required for language understanding and use in the real world. Drawing on evidence from cognitive neuroscience, we show that formal competence in humans relies on specialized language processing mechanisms, whereas functional competence recruits multiple extralinguistic capacities that comprise human thought, such as formal reasoning, world knowledge, situation modeling, and social cognition. In line with this distinction, LLMs show impressive (although imperfect) performance on tasks requiring formal linguistic competence, but fail on many tests requiring functional competence.” — Mahowald et al. (2023)

What is highlighted here is precisely the limitations I described earlier:


AI doesn’t understand specific linguistic contexts like a human can.


With the concept of “formal competence,” we humans create stems from language processing mechanisms that are deep within our brain and built on years of experience. While AI can mimic and summarize our human experiences, it’s not the same as an actual person.


So, will AI replace human translators and multilingual SEO strategists?


Not even close.


Considerations for using AI tools in your multilingual SEO strategy


I’m not saying that you should avoid AI completely—it could be quite useful for certain types of tasks. A lot depends on the complexity of your business and your particular role within the company. In the end, it’s up to you to decide whether AI helps or hinders your efforts.


If you do want to try to use AI to help with your multilingual SEO strategy, you need to use it strategically. From my experience, some of the biggest takeaways include:


  • Fact-check everything: AI tools often summarize sources without verifying whether they’re actually trustworthy or accurate. Instead, do your own research and fact check any AI outputs, regardless of whether they’re translated or in a single language.

  • Steer away from “set it and forget it” mass translation: Solely relying on AI outputs and copy/pasting them across your website is bound to be harmful over the long run. Any content directly translated by AI needs to be manually reviewed by a human digital marketing strategist for brand messaging, product information, and legal accuracy, as well as reviewed by a fluent speaker for language accuracy.

  • Focus on SEO localization: SEO localization needs to happen during keyword research in order for multilingual SEO to be truly successful. AI can’t handle SEO localization, so translators and SEO strategists still need to be intimately involved in reviewing outputs.

  • Do your own audience and market research: Again, AI summarizes information—often without context. The majority of digital marketers recognize the value of talking to real customers and understanding their experiences—it’s a crucial element of business success.


Priel Manes, founder of Two Dots Consulting and business growth strategist for 350+ companies, emphasizes the importance of deeply understanding customer needs:


A LinkedIn post from Priel Names, founder of Two Dots Consulting and business growth strategist for 350+ companies, where she emphasizes the importance of deeply understanding customer needs.

So even if AI is in your tech stack, don’t let it fully replace your SEO strategists and translators—they remain essential for successful international SEO and digital marketing.


Examples of ChatGPT for multilingual SEO localization


To help you determine whether or not generative AI is useful for your situation, I’ve put together five examples demonstrating how ChatGPT could be used for multilingual SEO.


Market-specific buyer personas

Here, I wanted to see if ChatGPT could create market-specific buyer personas for a CRM (customer relationship management) company that’s based in the US, but wants to expand into Germany. I entered following prompt into the platform:


Summarize buyer personas for a CRM company based in the US that wants to expand into Germany. List customer pain points, buying obstacles, and topics of importance in the German market surrounding data, lead generation, and marketing automation


A screenshot of ChatGTP’s output for the prompt related to creating buyer personas for a CRM company from the US who wants to expand into Germany.
Click to enlarge.

The output is decently accurate, but I know this because of my own experience (I’m from the US and have been living in Germany for the better part of a decade). Many companies outside of Germany don’t realize the importance of GDPR or how it affects brand messaging on a website in Germany. It’s also absolutely true that businesses must offer German language sales and customer service to be successful in the German market, so at least ChatGTP caught that.


Nevertheless, the information is still basic and only a starting point. The potential I see here is that the output could be used to formulate appropriate customer research questions for a business that is entirely new to the market (but it definitely doesn’t fully replace doing your own in-depth customer research).


If I was using ChatGPT for a market I was unfamiliar with, I’d still 100% check with a local market representative to understand if what’s listed is correct. Never accept what ChatGPT gives you at face value, because the risk of misinformation is simply too high to ignore.


Keyword localization

Next, I wanted to see how ChatGTP handles SEO keyword localization for a heated tobacco company that’s based in the UK and wants to expand into Italy. I entered the following prompt:


Localize the following English keywords into Italian and list the search volume

(the English keywords were listed)


A screenshot of ChatGTP’s output after being prompted to localize keywords related to heated tobacco from English to Italian.

The results were a direct translation, which isn’t the same as keyword localization. While this can be helpful as a starting point, Google Translate has long provided the same level of usefulness.


For multilingual SEO to work, you need to identify search intent match keywords with search volume, which (in many cases) aren’t direct translations. ChatGPT isn’t capable of providing search volume either, so you’d still need to look that up manually, meaning that this use of AI is absolutely not a replacement for an SEO tool.


In this context, I don’t really find AI useful at all, as both a human SEO strategist and SEO tools are still required.


International SEO blog post

Next, I wanted to see how well ChatGPT could write an SEO blog article for a contract management software company that’s based in the US, but has target audiences in France and Germany. I used the following prompt:


Translate this blog article to French


A screenshot of ChatGTP’s output for the prompt to translate a English blog article into French for a contract management software company.
The full output isn’t shown for ease of readability.

My notes for this are similar to the above: it’s about as useful as Google Translate. ChatGPT removed internal links and missed some parts of the web page. So, in that regard, the output is actually even worse than doing your own copy/paste translation with Google Translate.


While it seems like this could be relatively useful for easy topics, the output should still be manually proofread by a native speaker SEO strategist that understands the particular target market, target audience, and product nuances. However, the original blog didn’t adhere to SEO best practices, so if the original content were of better quality, the translation output might also improve.


Meta description translation

Another SEO task I wanted to test was to create meta description translations for the same contract management software company that’s based in the US but has target audiences in France and Germany (referred to in the example above). I entered the following prompt into ChatGTP:


Translate the following German meta descriptions to English in the form of a table.


A screenshot of ChatGTP’s output for the prompt to translate meta descriptions from German to English for a contract management software company.

Yet again, it translates the same as Google Translate, so I’m not really sure that it saves any time from what automatic translators already do. The one benefit is that you can copy/paste plain text and AI adjusts the formatting into a table (when you prompt it to).

Regardless, if you’re doing true SEO localization, the meta description would need to include the new target keyword to support on-page optimization (as I explained above). So, this part will still require a manual proofread from a multilingual SEO strategist who has already identified the relevant keywords.


B2B homepage hero section translation

For my last attempt at using AI for multilingual translation, I tried translating a B2B homepage hero section for the same CRM company that’s based in the US and wants to expand into Germany and France. I entered the following prompt:


Translate to German with 5 variations


The World’s Most Loved CRM Platform​


With a commitment to creating customers for life, X Brand ranks as a Champion CRM platform two years in a row for the only people that matter—CRM users. ​


It’s time to let the platform do the work.


 A screenshot of ChatGTP’s output for the prompt to provide 5 variations of a CRM company's homepage hero section copy translated from English to German.

I found the translation very basic, but the variations could be useful for localization. Many homepages go through iterations with copy variations, so this could be beneficial for A/B testing.


However, as always, a native speaker marketing strategist (that understands the focus market, target audience, and product nuances) should proofread the output.


AI is a translation and localization tool, not a human replacement


Languages are one of the most subjective and abstract concepts in existence, so it’s hard to imagine AI being able to come anywhere close to human experiential understanding. With the various nuances that come from localizing one language to another, topped with the layer of SEO and a unique business strategy, humans will still be the ones driving your marketing (at least for the foreseeable future).


However, I predict that many companies will perceive AI knowledge as an efficient skillset (regardless of whether it’s actually used in practice). So, regardless of how you view it, learning how to work with AI is a must to remain in the game.


 

adriana stein

Originally from the US and now living in Germany, Adriana Stein is the CEO and founder of the marketing agency AS Marketing. She leads a team of multi-language SEO experts who develop holistic international marketing strategies for global companies.

Get the Searchlight newsletter to your inbox

* By submitting this form, you agree to the Wix Terms of Use

and acknowledge that Wix will treat your data in accordance

with Wix's Privacy Policy

Thank you for subscribing

bottom of page