Local translators, (who are vital for transcreation,) may not have the best grasp of the nuances of Paid Search, such as how different elements are deployed or character limits. Each round of changes adds to the administrative burden and likelihood for errors, so maintaining a high level of consistency across multiple markets can be a tricky task at best. 

With advances in generative AI this is all set to change. 

Large language models (LLM) offer a revolutionary leap forward in this area. Now it is possible to translate quickly, easily, cheaply, and crucially accurately across several languages simultaneously at the touch of a button. At Merkle, we have created a proprietary AI-based tool called d.Scriptor, that can review the content of a landing page, be given the context of any particular translation, be it a headline, site link or description, fit a brand’s tone of voice and asked to remain under character limit(s) whilst providing the equivalent in local language nearly instantly. 

 A centralised PPC team need only provide a base version of all text assets whilst the LLM will fill in the ‘blanks’ in a wide variety of languages with series of calls to the LLM API. Although this could of course be achieved with a manual translation, the speed and efficiency of this new AI-enabled process is leagues ahead. This process outputs a convenient format for review and correction by a native speaker, which can then be uploaded directly into the platform by your PPC team. This AI-based approach cuts down on administrative back and forth and hence provides a big efficiency for teams and translation costs over time. We would however always recommend that a native speaker reviews all outputs prior to use, as although LLMs have been improving fast there is no substitute for a human input and review.

As the landscape of copy and creative in search continues to evolve, the rise of conversational search and Performance Max means that we will require more frequent updates and changes to effectively optimise campaigns. This is crucial because agencies and advertisers have direct control over this lever of optimisation. As we move forwards the platforms will be increasingly pushing automated creation of assets, again necessitating faster turnarounds on localisation than is achievable via traditional methods.   

What does this mean for how we should structure teams going forwards? 

 Increasingly this means that an experienced, centralised team supported by powerful AI tools and enterprise level management solutions is going to be best placed to support international brands efficiently and effectively. Maximising economies of scale and minimising the additional administrative costs of localised support from individual markets and siloed teams.   

Centralisation is equally important between Paid and Organic search. Breaking down the silos between these teams will become increasingly important as the lines between the two approaches blur. Thanks to conversational search (the growth of searches with multiple interactions and inputs on the SERP) and the exponential growth of key-wordless campaigns such as performance max  only offers keyword exclusions (and not targeting) resulting in even more unique searches and a richer picture of the user. Logically we should view Search Engine Marketing (SEM) holistically, with SEO practitioners also in a key position to benefit from the increased speed and accuracy of translations via a Large Language Model. This will enable readability, and efficiency for content activation. At Merkle we have developed our Total Search team and proposition with this in mind. 

Early adopters in the multi-market campaign space are going to be in pole position to benefit from usage of LLM (Large Language Model), enjoying savings and efficiencies to succeed with a leaner translation requirement and flexible teams based in a global or regional hub. These teams will be in a better position to experiment and optimise via ad copy creation, with a faster feedback loop and more rapidly actualised gains from operating at scale.