FYI, I think LLMs are here on they Hype Cycle. If you’re in a rush and want to find out why head to this heading “Where do I think we are on the Hype Cycle and why?”
Not in a rush and want a good longish read for lunch, continue below 🙂
LLMs in 2024
The hype around artificial intelligence is nothing new. Neither is our fear and mistrust. Greek mythology provides us with one of our earliest images of a robot in the myth of Talos, the sentient automaton that patrolled Crete.
Since then we haven’t been able to shake the fascination and apprehension that surrounds artificial intelligence.
But fast forward roughly 2,700 years and you’ll find society less worried about giant bronze security guards and more intrigued by machines that can realistically reproduce text and answers that are indistinguishable from an actual human.
This specific form of AI I’m talking about is the Large Language Model (LLMs), the engine that powers ChatGPT and its ever multiplying rogues-gallery of competitors.
2023 was a crazy year for the application of AI, and specifically LLMs, in search. ChatGPT became so ubiquitous in the public consciousness that the BBC were running news stories around the potential dangers of AI and even the tech-bro golden bollock boy himself, Elon Musk was warning of the dangers AI could pose on society.
When Bing integrated ChatGPT into its search engine in May of 2023 it caused a further escalation in the LLM arms race that had been brewing between companies such as Google and Microsoft.
For me personally, the real sign that LLMs had broken through the barrier between early tech adopters and the general public was the fact my mum, yes my technophobe mum, was trying out ChatGPT to see what it could do.
Since those heady days of May through to August in 2023 the general hype around LLMs has died down somewhat. An interesting data point that emerged from the end of 2023 was that despite integrating ChatGPT into their search engine first, Bing didn’t see any growth in their usages by November. Their market share remained flat at 3%.
This to me demonstrates a general apathy from search engine users when it comes to using LLMs*. The excitement is over and we’re now sitting back and thinking about how we can actually use tools like ChatGPT in our everyday lives.
The novelty of getting LLMs to write poems in the style of famous film actors (or infamous presidents, see below) is over. We no longer want it to write crap comedy sketches for us. We need to seriously think about how we’re going to use this new technology.
For me, the best way to chart how LLMs will impact search is through the Gartner Hype Cycle. In this article I want to pin-point where we’re at in the cycle and what evidence there is to support that. I’ll also explain where I see LLMs heading in the next 12 months and how SEOs can harness these developments to get ahead of their competition in 2024.
* This apathy is just one of several factors that could explain why Bing didn’t gain any market share when it incorporated ChatGPT. The other main reason could be Google brand loyalty from consumers and just Google’s general dominance of the search market.
What is the Gartner Hype Cycle?
If you’re reading this article and you’re unsure what the Gartner Hype Cycle is i’ll quickly summarise below. The Gartner Hype Cycle is a graphical representation of the lifecycle stages that emerging technologies typically go through. It consists of five key phases:
Innovation Trigger
Peak of Inflated Expectations
Trough of Disillusionment
Slope of Enlightenment
Plateau of Productivity.
The model helps organization’s understand the maturity, adoption, and potential risks associated with new technologies over time. You can see an example of the hype cycle below.
So, where are LLMs on the Gartner Hype Cycle at the start of 2024 and what actions can we as SEOs take from this to help to plan strategies for the year ahead.
Where do I think we are on the Hype Cycle and why?
Before we place LLMs on the Hype Cycle I think it’d be wise to list the most recent developments in the space so we can better contextualise the current state of LLMs in search.
For me, I think there are 3 big things we should keep in mind before we place LLMs for search on the Hype Cycle:
1) At the tail end of 2023 Google announced the launch of Gemini it’s most advanced AI model that will power its LLM. Gemini is important as it can understand a variety of media, not just text. Gemini can watch videos, listen to podcasts and understand excel spreadsheets.
This is important because it opens up a variety of new data sources a ‘traditional’ LLM might not be able to understand. This gives it a larger pool of information to learn from and will therefore make it a better AI tool.
2) Currently available in America but heading to the UK and Europe in 2024 is Google Perspectives.
Perspectives is a new search feature designed to help users find tips, advice, and stories from sources other than websites. You can access Perspectives via a filter in the search results and it serves users with content shared by real people and content creators. This includes short and long form videos, images, and written content posted on social media platforms and online forums.
Again, this is powered by Gemini and does a great job of highlighting how advanced LLMs and an AI powered search will become in 2024.
3) Last but not least, it’s worth mentioning again that Bings integration of ChatGPT did not lead to an increased market share for the browser. This, to me, indicates that the general internet users isn’t massively moved to action by these tools. They’ll use them when they’re presented with them but won’t go out there way to research and understand them.
After reviewing all these points I believe that end consumers won’t lead the adoption of LLMs. That’ll fall to big businesses and search engines. LLMs will power the tools created by brands to captivate, convert and assist customers. Outside of some specific industries no ones going to be using ChatGPT on a day-to-day basis.
So, with all that out the way, where are we?
Personally I think we’re just about to come out of the trough of disillusionment. Although I do feel in this instance ‘disillusionment’ isn’t really the right term. I think ‘rate of adoption by the general public’ would probably be better.
If we think of the ‘expectation peak’ as those crazy days around April 2023 when everyone and their mum was fiddling with ChatGPT I feel the hype around LLM tools has largely subsided outside of marketing and technical circles.
I do however feel this trough will be very short lived as I’ve seen multiple instances of brands and Google itself developing tools and products for use by consumers which integrate LLM using APIs.
Once brands know how to harness this technology to benefit customers we’ll see a mass adoption of this type of AI by the general consumer who probably won’t realise they’re being served information and result by the same technology that is used by ChatGPT.
The most compelling example I’ve found so far that highlights this point comes from a recent The Verge article. One of the new features that will be added to Search Generative Experience will allow searchers to get AI to create an image of an item of clothing they’d like to buy using LLM prompts and then Google will find the closest product match on the web and serve that to users. You can see that process in action via the gif below.
As 2024 slogs on I predict we’ll see more and more brands, as well as internet browsers, finding unique, innovative and captivating ways to deploy LLMs in ways that greatly enhance user experiences. But how could SEOs use LLMs and what impact will they have on search? I’ll share my 2 cents below.
How can SEOs harness this knowledge in 2024?
We all know Search Generative Experience (SGE) is coming. This isn’t news to anyone working in digital marketing or SEO anymore but it’s worth mentioning here. SGE is probably the biggest change to the SERP in the past decade and is likely going to cause site traffic to drop off a cliff for a lot of brands.
Especially if your site gets a lot of its visitors for informational content you’re likely going to see a sharp drop in visitors as they get their queries answered in the SERP.
That being said, revenue should remain consistent as searchers can’t yet complete their purchase journey on the SERP page. They still need to go to your site (let’s hope nothing changes there!).
One idea I’ve seen floating around recently is the idea of setting up a static landing page on your site which acts like a LLM answer box tool but is only trained on information on your site.
This means brands can still have all the content they need on their website for SEO purposes but site visitors can navigate to a static page with an embedded chat box which can serve them the information they need without them having to trawl through your site to find the answers they need.
This functionality can be added to an actual popup, lower third chatbot too if that works better for your site.
The tool could also help people along their buyer journey by recommending specific products or landing pages once a user’s query has been answered.
This could change site architecture to funnel visitors towards an LLM tool before using it to send them down product funnels or to sales representatives. A chat box could be embedded on a site’s homepage but the technical considerations in regards to site speed would need to be weighed up before you went ahead with that.
All in all there are infinite possibilities when it comes to incorporating LLMs and AI into your SEO strategy. The SERP landscape is changing rapidly and that will have a knock on effect in regards to how people search. Getting ahead of this trend and integrating LLMs into your site will give users the best experience possible and set LLMs well on their way to the plateau of productivity.
