Creative Linguistics: Infusing Global Content Strategy with AI in 2024

In the past, brands looked to expand further and further afield, but this is changing. Big brands, which already have brand awareness globally, are acting more cautiously. They’re working to identify where big pockets of volume are in main markets, while also working to make content in those markets better and more effective. 

More brands are directing budget towards brand campaigns and realising that it’s not just about the efficiency of performance. It’s showing the industry that the focus needs to be on higher quality content in big markets and other markets with pre-existing brand awareness. We have seen some big brands failing to check if their website and content is localised in markets where they already have some good brand volume. That’s the ‘low hanging fruit’ that brands need to address in the year ahead, to capitalise on what they already have instead of jumping into something new. 

New brands are beginning to understand that internationalisation, while an exciting opportunity, is more complex than it first appears. It’s better to have an attuned brand presence in just a few markets, rather than launching in multiple markets at once and seeing how it goes. 

The market has a growing preference for non-static assets. A new study showed that video accounts for 65% of all internet traffic with a 24% year on increase. The continued domination of platforms like YouTube and TikTok over linear TV, and a general shift towards social media, is creating a requirement to produce more moving assets at a faster speed. We’ve seen this in the way that budgets have shifted from just search and display to majority social. With moving assets, localisation reaches beyond text ad copy. We have seen huge growth in voiceover, although less so in dubbing, with someone speaking about products, addressing followers, and so forth - it’s a growing art. With this shift, brands need to focus on more asset adaptation as well, as they’re needed for many different platforms and in many different versions and formats. I think we will begin to see a lot of production at scale, high-quality tailored transcreation and voiceover happening due to this shift. 

Brands need to start approaching localisation differently and understand the need for thorough research first. Before localisation can take place, brands need to really think about what their consumers want and the best way to do this is through in-market insights and ethnographic research. While it’s important to identify whether a brand will need copywriting, transcreation or localisation services, it’s most important to identify what the target audience actually needs before deciding on content creation methodology.

The industry has a great process when it comes to creative and media pitches as well as RFPs (request for proposal) but the same can’t be said about language. Language remains an area that is understood and summed up as requiring either copywriting or transcreation or translation. It is seen as a reactive element of the work stream, not necessarily as something that is woven into the strategy process from end-to-end. 

The industry needs to understand that it can’t approach localisation by focusing on one single type of content; it must be addressed as a whole to get the best recommendations. With more visual assets coming through, you may need to look at the visual segments, you may have to create more creative content from scratch and to understand what the platforms are, what the mediums are. Once that’s established, smart decisions can be made about what, when, where, how and with whom. Brands can’t make those decisions, though, if they don’t understand what’s needed. So, the advice is not to rush, understand needs thoroughly and include language consideration earlier in, and throughout, the process. 

When we talk about AI in localisation, we’re talking about different types of technology. This includes technology that we have used for years, such as NMT (neural machine translation) systems that Google Translate is based on. Now we have LLMs (large language models) like ChatGPT and Bard. LLM translation is not going to replace the current process for a number of reasons. 

The first is trust. We already know that LLMs often produce hallucinations and as such cannot be trusted to always provide an accurate or even a truthful response. The other is consistency. Different people asking LLMs the same question may receive a different answer, with some being totally incorrect. If you are not an expert, how do you know which option to select for your particular requirements? Then you have what we may call 'false affirmations' – sometimes when challenging the LLM output, it will agree and support your criticism and adjust its previous response, even though the original was factually accurate and your concerns turn out to be unjustified.

What will continue to happen in the field of multilingual content production, then, is a hybrid approach. LLMs will give us more fluency as they are mostly linguistically and grammatically spot-on but terminology and factual accuracy can be a challenge. This is where NMT would come in. This approach can then be combined with client-specific TMs (translation memory), which is a database of client-approved bilingual content that provides 100% effective accuracy. A team of content specialists will smartly merge usage of LLM, MT, TM and human production. They will ensure that that final step is always carried out by a highly-specialised academic native expert who reviews the content and focuses on post-editing to produce something that works for the needs of the client and fulfils their specific creative idea and objective.

Artificial Intelligence, or more specifically a Large Language Model (LLM), excels in enhancing efficiency and providing rudimentary ideas and information. However, it falls short in generating content that effectively engages with a rapidly evolving target audience. It is commonly acknowledged that LLMs are, in essence, 'statistics on steroids' - a system trained on extensive portions of historical data, limited to a certain point in history, and lacking awareness of current events crucial for crafting campaigns for highly time-sensitive topics.

As culture and language rapidly and unpredictably evolve, AI lacks the capacity to comprehend how the audience in a specific market communicates or, more critically, how they feel. Human creation of new thoughts and phrases are not solely based on our past linguistic knowledge; they are influenced by experiences derived from tactile, visual, auditory, and olfactory sensations. These sensory experiences lead us to formulate words and transform them into rich, creative expressions, whether in writing or speech.

As it stands, AI cannot innovate based on the integration of these senses; it does not venture into the realm of novel content ideas. In contrast, a human, engaging all their senses in content creation, possesses the unique capability to produce something original and meaningful, something that truly resonates with and leaves a lasting impact on audiences, thereby crafting experiences and memories.