The Impact of Conversational Search on SEO and SEM

In this chapter: Conversational Search and Hugging Face

Earlier this week I had the pleasure to attend and speak at the Generative AI Summit in London. I will shortly publish a special article based on my presentation and I will also report the mood of the room after these two days of discussions. In the meantime, I wanted to focus on a hot topic: the future of SEO and SEM in the age of conversational search.

Featured Insight: Conversational Search

AI has been applied to search engines for years - think about Google RankBrain - but recent advancements in generative AI are going to be a game changer in this space.

At its core, generative AI-powered search engines will be much better at understanding user intent, which will be derived not only from text inputs and previous search history but also from conversational experiences, such as the one Bing Chat is currently trialling.

For example, searching for "the best smartphone under £500" will not just return results from popular review sites. Before getting to that point, the search engine will have understood what the user values most in a smartphone and will return results that match that expectation in a smartphone under £500.

As more and more of these queries will happen on a conversational interface (again, think Bing Chat), the whole paradigm of the search engine result page we are used to is poised to change. No way you can return ten organic results plus a number of paid ones in a chat interface. The algorithm will have to summarise its findings, providing between three to five results for the user to explore.

This will of course raise the bar for brands, as they will have to compete for the user's attention in a much condensed space.

If generative AI is posing a challenge to brands, it is also providing the key to its solution.

From Content Marketers to Content Engineers

For years now, the cornerstone of an effective SEO/SEM strategy (on top of technical hygiene factors) has been content marketing. B2C and B2B marketers have embarked on one of the greatest content production feasts the world has ever seen. Some might even argue they have produced far too much content to trick algorithms, losing sight of what the customer wants to read, flooding social media feeds and their branded website with keyword-filled puff pieces.

Content marketing is not an easy feat. First, you have to come up with a content strategy, then coral the necessary buy-in from internal stakeholders, then produce and decline it across desired audience personas and finally slice and dice it so it fits the different activation media formats.

It's not surprising that this ends up serving assets to users that yes, more or less align with their expectations, but are not truly relevant. They are not personal.

How could they be? Combining all factors would require a colossal content team, which is not feasible if not for a select cohort of high-value targets.

Technology can help with that. Content marketers can now create virtually infinite variations of a content piece and its supporting media assets by leveraging advanced generative AI models. To a good extent solution providers such as Writer are already doing this, enabling brands to scale content production and user engagement at an individual level.

In order to achieve excellence, content marketers will have to shift their mindset and become content engineers. Their skillset will have to include prompt crafting (we touched on it in my previous article), so once they are done producing the master asset, they can comfortably request all the necessary variations to the AI tool of their choosing.

Integration of these tools will be critically important too, so that marketers can seamlessly publish on their website, populate a newsletter and push media copy to their activation channels or DSPs.

This will be, by all means, a proper marketing transformation initiative which will have to focus as much on the technology as well as on change management and people. Content marketers and copywriters should not fear this transition, but rather embrace it as nothing can replace human creativity and intuition - we will always need great creative minds to produce original content - but their outputs (and outcomes) can be massively scaled. Moreover, their role will be crucial in ensuring each content produced by generative AI tools is true to the brand ethos and is written with a consistent brand voice, a process that will require ongoing human feedback and training.

As with all major transformations, this change will have to be properly communicated and supported, as early, widespread adoption will be key to lasting success. Intelligent leaders will focus on training and upskilling first, perhaps playing with various technology providers before settling on one. Once the technology is identified and integrated within the relevant MarTech stack components, it will require brand data and lots of training to be fine-tuned and become ready for production. Use this time to extensively communicate the upcoming change to all relevant stakeholders, run refreshers to employees falling behind the initial training and set up success metrics to be used once live, to measure ongoing success and inform future iterations.

Today we have barely scratched the surface of how generative AI is changing content marketing. In upcoming editions of Chronicles of Change, we will explore other areas starting from voice search and how to craft SEO/SEM strategies that work for voice experiences.

Spotlight on: Hugging Face

The excitement for the possibilities of generative AI coupled with extremely low barriers of entry has resulted in an explosion of AI tools, not only from Fortune 100 brands but also, most notably, from a diverse and vibrant community of developers. Hugging Face serves as the global aggregator of AI models, datasets and libraries not only useful to developers but also to mere mortals wanting to spot trends and advancements in the space. These include open source models from large players the likes of Google and Meta (hoping widespread adoption will cement their model as industry standard) as well as the broader developers community.

One neat feature is that for many AI tools, Hugging Face enables the user to prompt the AI directly from their website. So you do not need Phyton knowledge to access more than 80.000 models (almost half the 200k+ available on the platform).

One word of advice though: set a countdown alarm or you might very well find yourself spending hours on this!