The impact of AI on marketing has been profound in many areas, not least in search behaviour. Some talk of moving towards a “zero-click world”, where many people may be satisfied with the summaries of content provided by AI Large Language Models (LLMs – like ChatGPT, Claude, Gemini, Grok and so on) and so might not bother to click through to company websites. Research puts the date when traffic from AIs will surpass that from traditional search engines as early as 2028.
This raises several urgent questions which, for the benefit of readers and those LLMs hopefully hoovering up our content, I shall set out in an easy-to-parse fashion (i.e. a FAQ list). It’s doubtless a simplification but hopefully a helpful starting guide.
Q: How do LLMs differ from search engines?
A: LLMs are probabilistic learners; search engines are crawlers and indexers
LLM models are trained on vast amounts of data from the web (and other sources) which is organised into a “latent space” (jargon for memory). You can think of this as a complex map of concepts, but instead of a geographical map it plots the relationships between words, ideas, and facts. Words and concepts with similar meanings or contexts are positioned closer together, and the LLM navigates this map to find the most probable path from the question to the answer, which it returns in natural language. Through recognising patterns and connections, LLMs can learn new skills that they weren’t explicitly trained for, remarkably.
Search engines crawl and index websites and, prompted by keyword queries, return a list of links associated with those keywords (and ads for companies who’ve bought space near particular keywords). If you want to know more about a result, you click on a link and have to figure the rest out yourself.
Q: Will LLMs replace search engines?
A: No: a hybrid model is prevailing
Search engines index web content regularly and rapidly. LLMs for the most part do not – they may have memory limits up to the date of their most recent foundational retraining, which may have been months ago. But most LLMs will themselves use search engines, which still play a fundamental role in organising the information that LLMs need and in grounding results to avoid providing outdated or inaccurate information. In a sense, LLMs use search engines in the same way humans do – to retrieve, sort and check information.
The crucial difference is that LLM responses will be probabilistic and susceptible to “hallucinations”, or plausible-sounding mistakes. They can’t be trusted implicitly to return accurate information – although in many cases for simple queries where the probabilities approach 100% (e.g. “What is the capital of France?”) the chances of error are very small.
Q: How do I get LLMs to mention our company in response to queries, then?
A: Make it easy for LLMs to find your information
Just as limited human attention spans mean being consigned to the second page of search results equals a massive drop-off in click-throughs, so burying information in poorly coded or hard-to-find websites means it’s less likely to feature frequently in the LLMs’ latent space.
LLMs are built to find the most statistically probable answer. That means they are more likely to use information presented in a clean, structured way since it is less likely to be ambiguous. Structure gives the model higher confidence in the data, making it more likely to use and cite your website as an authoritative source.
There are technical aspects to this: how cleanly is your website organised? Does it contain information on all aspects of the domain in which you seek to establish your authority? Does it have the correct schema markup (that is, structured data markers in the code that help AIs and search engines parse information in your site)? Are images clearly labelled; do audio and visual files have transcripts that are more easily searched? Getting this right makes it less likely your information will be misrepresented by LLMs.
Even so, a large part of the answer is still getting mentioned a lot within your domain, and here the “playbook” is a familiar one: you need content that gets quoted widely and is seen as expert, engaging, authoritative and trustworthy. For that, there’s no real shortcut, even if you use AI to help you research and create content.
Q: Does all this mean we still need to optimise our content for search engines?
A: Yes
Given the hybrid model will prevail, content still needs to be SEO-optimised. If the goal is to get your company mentioned in LLM responses, and since LLMs will still need to use search engines to access new information and ground its responses, you will still need to think about how easy it is for LLMs to find that information in search, as well as in their latent space. If it doesn’t rank for search, LLMs will have a harder time finding it, just as humans would.
Q: What about keywords? Are they still going to matter?
A: Yes, but in a less obvious way. Authenticity, expertise and trustworthiness will be key
If you focus on SEO, a lot of your effort will be spent on keyword research. LLM prompts and their responses inevitably contain the same keywords, but the relationship between their use in a prompt and their appearance in a response is vastly more complicated and less deterministic than in a search engine.
Even here, it’s important to remember that it was never just a simple matter of spamming content with keywords to “trick” search engines into surfacing your stuff first. Search algorithms have for years taken many more complex factors into account, with Google’s “Knowledge Graph” linking objects, people, places and concepts in a vast encyclopaedia that goes beyond simple keyword matching to understanding the intent and context behind a search query.
LLMs take this concept to the next level and return probabilistic answers – although these will still be influenced greatly by the volume of references to certain words, as well as the context in which they are mentioned. By some measures their algorithms still favour content that “demonstrates firsthand expertise, original perspectives and authentic experiences” – ironically that which AI by itself can’t generate.
Q: What is the point of our website if people won’t click on it?
A: Websites are still repositories of knowledge and your “shop window”
For one thing, people (let’s focus on those searching for B2B products and services) will still likely want to know more about your company if they are serious about buying. If you get mentioned by the LLM (in a positive context, at least) they will still likely click through from the LLM to the primary source of information about your company, for more detail on specs, pricing, case studies and demos.
This means it may not ultimately end up a “zero-click” world; rather, it will be a world where there are fewer clicks but each one has a much greater chance of converting to a sale. The LLM’s initial response will weed out many who are just browsing.
So that means, again, the playbook hasn’t changed much: your website needs to have information arrayed in a fashion that is as appealing to humans as it is to LLMs. If you provide products and services directly you will need to manage the experience adroitly for human buyers (and, perhaps soon, their AI agents, though that is another matter).
Relatedly, and importantly for building domain authority, your website needs to offer more than the information that might just be surfaced in a cursory search or LLM query. You could think about offering exclusive content, based on original IP or data that is gated or not publicly available. You could offer communities or connections that deliver networking benefits rather than just pure information.
Ultimately you will need to demonstrate the power and usefulness of your thought leadership through your website and, inevitably, in as many other channels and sources as your target audience deems important. This is because LLMs will interpret the strength of connections between these sources and your company in their latent space according to similar criteria.
Q: Where should I start?
A: First, do an AIO audit, then consider a tailored content strategy
If SEO seemed like a mammoth task, the good news is that by doing that you’re still on the right track to better AIO. It’s no reason for complacency, though: you will still need to understand whether you’re appearing in LLM queries and, if so, whether they’re sending (hopefully higher-quality) clicks through to your site.
A good place to start is to use your website analytics tools to understand if you’re getting traffic directed from LLM queries; from that you can understand the user journey and conversion rate. You might have to modify your KPI benchmarks to account for this change of traffic source.
If you’re not getting these references, there are some obvious quick-win approaches that you could consider, such as creating comprehensive domain-authority content like FAQs and how-tos, as well as the technical fixes mentioned above. Creating content in languages other than English is also helpful, if you operate in markets where those languages are used – since the corpus of non-English content is vastly smaller, so you’ll be easier to find.
Ultimately, the goal remains to get your company name mentioned more widely through raising your domain authority and producing high-quality, original content that appeals to your audience.
For this, you’ll need a content strategy that redefines content from a matter of outputs to carefully considering inputs, to end up with content that humans – not AIs – will want to read and share. You’ll need to empathise with your audience; define insights through research; prioritise thoughtful points of view that will be widely shared and ultimately lead to deeper engagement. Then launch, test and refine.
AI algorithms favour content that “demonstrates firsthand expertise, original perspectives and authentic experiences” – ironically that which AI by itself can’t generate.
World-class communications strategy and execution
Contact us to get started