Language Model Awareness Optimization (LMAO)
LLM Awareness is the new SEO. To succeed, business leaders must consider how their organization is represented and recalled by the world's AI models.
It’s happening.
The internet (and how we store and share human knowledge) is changing again. And with it, the fundamental nature of how we access information and understand truth.
AI (and especially LLMs) are rapidly changing how we engage with information and knowledge. And one of the most significant places this is evident is in Search.
A decade from now, your kids will laugh at your stories about having to read through pages of links to find what you needed instead of having AI do it for you.
Frankly, it’s the logical next step. LLMs increasingly contain much if not all of the world’s public (and non-public) knowledge/content. As we get better at compressing it, prompting it, and updating it, there will be little reason to browsing pages of web-crawler-produce search result links. Your personal AI will just know what you were looking for and how it fits into your current context.
The grand implications of this all are likely beyond our imagination, but there is a path forward for those flexible and focused enough to find the edge and surf the wave of coming change.
On the eve of OpenAI’s next big announcement anticipated to make splashes in search, voice, mobile and overall AI capabilities—here are my thoughts on the evolution from Search to AI as our source of truth, understanding and discovery. And its impact on business strategy.
Search Engine → AI Knowledge
We have quickly and (relatively) quietly begun our departure permanently away from the “10 blue links” search engine experience we all know and love/hate. Until now, it has dominated the modern internet era when it comes to information search and discovery—but this is changing.
It began with the release of ChatGPT when people began to realize these new AI models could answer, with reliability, the simple general knowledge questions that we’d all been asking Alexa, Siri, and Google—but now it came with a single clear answer, THE answer, rather than a list of links. Recipes, how-tos, medical questions, historical tidbits, etc. These kinds of needs began to go to AI over search.
Then came the release of Bing with GPT-4 which further blurred the line between Search functionality and AI. Most recently, updated knowledge cutoffs and improvements in model training have added clarity and relevance.
As people become more comfortable with LLMs we/they will expect them to have the “internet’s” knowledge, and as a result increasingly rely on AI as the primary means to find and verify information.
This has serious implications for businesses and individuals. For today, we’ll focus on business.
Language Model Awareness Optimization: Brand Success in the Age of AI-driven Discovery
You used to have to worry about showing up on page 1 of Google results, now you have to worry about showing up in ChatGPT’s outputs. Said differently, you used to have to optimize for Search Engines (SEO)—now you have to optimize for Language Model Awareness (LMAO). (Note: the irony of the acronym here is not lost on me).
[Definition] Language Model Awareness Optimization:
The practice of optimizing content, data and internet presence with the strategic purpose of increasing prominence of key brand related information in the neural networks of LLMs being trained on public internet data. The goal being immediate recall, recognition, and association of certain topics or prompts with brand concepts.
In other words, SEO but for maximizing presence in LLM outputs.
Why Language Model Awareness Optimization?
Who cares? Why do it? For those who can master it, there is a massive reward to be had in expanding business relationships through the newest means of information transfer—AI. For those in doubt, there are some risks that you should consider.
When AI delivers your information:
It doesn’t know what it hasn’t been trained on sufficiently
Problem: If your business doesn’t have a large enough (or any) content presence in a particular AI’s training materials, your business won’t be found by customers. You will not exist to them or anyone who uses that particular AI model.
It only knows what it has been trained on
Problem: If your company doesn’t have enough of the right presence in the AI’s training materials, you won’t be discovered by your target customers or represented to customers in a way that accurately reflects your brand. You might show up, but now how you’d like to be represented and not to the people you’d like to see your information.
It only provides one answer (THE answer), and it doesn’t tell you there’s more to know about the subject unless asked
Problem: If for whatever reason your company is not the most recognized solution/option (e.g. OpenAI starts a paid promotion partnership with you competitor, or you are a small player in a large field) you may never be presented to your customers, even those who specifically ask for what you do in your area.
This should be alarming to any business leader. And the full weight of these implications may never come to bear—maybe AI companies will ultimately decide to altruistically carve a path and build infrastructure to ensure fair and balanced information exchange.
But that has yet to be seen.
In the mean time, models are gobbling up internet data for training and in doing so shaping their understanding of your company, what it does, who it serves, what value it provides, and how it fits more broadly into its computed understanding of the world.
So what do you do?
The truth is we’re already moving toward a pay-to-play world. Soon AI providers will charge a fee to companies who want to be placed at the forefront of its outputs to users. It will likely be an “advertising (but not advertising because its LLMs)” model of some kind.
It’s important to note that some AI companies have opted for charging users for Pro and Premium accounts rather than monetize information results. Perplexity is the major search challenger doing this, but it still must prove it can sustain without being tempted to sell off search placements for the extra cash.
OpenAI, the biggest player, has yet to make a direct play in Search but they are already minting exclusive partnerships with content providers (FT and StackOverflow). Instead, their partnership with Microsoft’s Bing proved out the value of AI-aided search indirectly and there is speculation this week’s announcement will include a shift to “real-time-knowledge”—not a direct stab at Search Engines, but real time knowledge means essentially Search.
So how do you get started preparing (playing catch-up) to get your business ready for an AI-driven information world and optimize it’s relevance and placement for business success in the long term.
There are (at least) two timelines to consider right now:
Short term optimization opportunity
Long term planning and development imperative
The Short Term: Optimization Opportunity
While LLM providers are still in the early stages of developing their business models, and stabilizing basic services there is an opportunity to build your brands presence and carve out a corner of the neural network to call your own.
In short, the opportunity is this: maximize the chance your brand will be well known to AI neural networks as they continue to scan the web and update their training knowledge.
Consider how your writing effects brand presence in text models: “We do xyz…” may be less impactful than “[Company Name] doesn't xyz...”
Consider what is most universally important about your brand and how should that be communicated to the public—then repeat everywhere so the AI finds it frequently and consistently across various sources and content types.
Consider the comprehensive story your telling with content on your company website, social profiles, and public directory profiles—does it encompass all of the ways your customers may need to find your information? Can an AI logically connect you to the subjects where you most want to be found?
I could go on, but you get the picture. With Search Engines you could be relatively confident you would show up alongside the bundle of other relevant results—with AI-driven search it is the TOTALITY of how you show up online that may most impact how your company is represent to the AI’s users.
We’re in the early days of AI Search and this short term opportunity may close in the coming years. But while training is still broad and general, and models are still building infrastructure, there is an edge to be gained and risks to be considered.
Long Term: Time will tell
In the longer term timeline businesses need to be preparing for a variety of potential futures. All of them likely require developing new capabilities for your teams.
Based on my own experience as a technology team leader and my beliefs about where thing are headed there are a few areas you can start considering now to get ahead (or at least keep up with) the coming changes:
AI understanding—if your team doesn’t use AI tools regularly and understand their ins and outs as users, they won’t be able to grasp the implications of these technologies for your business and how best to maximize them. Dedicating time, even just a little, to developing and exploring AI tools can prepare the team to work with them when the time comes.
Content development (helpful with AI)—content feeds AI (for now, until synthetic data—another post for later). Your content is the key to your presence within the AI neural network and its outputs. If you aren’t creating publicly available content in multiple formats across multiple platforms or websites, you should. There are a lot of AI tools that can help write blog posts, create videos, and streamline podcasts. The more you create, the more robust your information cluster will be, the clearer AI will understand your company and its role.
Brand curation—a lot of brands have a hard time keeping all of the data points updated and consistent with the most recent product information, value propositions and even contact info. This inconsistence may damage a neural network’s confidence in the information and decrease the likelihood of it being surfaced. Ensuring tight and comprehensive curation of brand information across the internet will become vastly more important in the future. (I expect entire roles and companies dedicated to the maintenance of “true real time information” feeding models updates).
Maybe none of this will be relevant after tomorrow’s OpenAI announcement and maybe we’ll find ourselves in a blissful state of AI-led abundance, fairness and truth in the future.
But my belief is hope for the best, prepare for the worst. As a business leader today that means preparing for the inevitability of AI-driven information discovery and positioning your team and your brand to succeed in Language Model Awareness Optimization.
That’s a wrap. Thanks for reading. I’ve included an AI-generated comparison of Traditional Search Engines and AI-Search below if you care to read it.
Until next time.
Appendix
Comparing Traditional Search vs. AI Search
Traditional Search Engines (e.g., Google, Bing, Yahoo)
Pros:
Accuracy and Reliability: Traditional search engines have a long history of development, leading to mature and often more reliable search methodologies. They use extensive link analysis and ranking algorithms that are very effective in delivering accurate results for many queries.
Wide Information Access: They index a vast array of web pages and are excellent for broad searches that require access to multiple sources or when searching for specific websites or content.
Privacy: Some traditional search engines have clear privacy policies and options for users who are concerned about data security and privacy.
No Training Required: Users are generally familiar with the interface and mechanics of traditional search engines, requiring no additional learning curve.
Cons:
Keyword Dependency: They rely heavily on keyword matching, which can make them less effective in understanding the context or the intent behind user queries, especially if the keywords are vague or have multiple meanings.
User Effort: Users often need to sift through pages of links to find relevant information. This can be time-consuming and inefficient, especially when the query is complex.
Static Results: Traditional search engines provide a list of links that users must explore individually; they do not synthesize information or provide direct answers.
AI-Driven Search Engines (e.g., Perplexity, Updated Bing with GPT, Gemini)
Pros:
Contextual Understanding: By using large language models, AI search engines can understand the context and nuance of queries better, allowing for more accurate and relevant responses, especially in complex or nuanced inquiries.
Direct Answers: They can synthesize information from multiple sources to provide direct, concise answers instead of just links. This can save users time and effort in finding the information they need.
Conversational Interaction: AI-driven searches can handle follow-up questions and provide more interactive and engaging search experiences, resembling a human-like conversation.
Personalization: These engines can tailor responses based on the user’s past interactions and preferences, potentially increasing the relevance of the information provided.
Cons:
Dependence on Training Data: The performance of AI-driven search engines heavily depends on the quality and breadth of their training data. This can lead to biases or gaps in knowledge if the training data is not comprehensive.
Transparency Issues: AI algorithms can sometimes operate as "black boxes," making it difficult for users to understand how results were derived. This can be a concern for tasks requiring high transparency.
Privacy Concerns: AI-driven engines often require more personal data to tailor responses effectively, which can raise privacy issues unless managed with strict data protection measures.
Complexity and Cost: They are typically more complex and costly to develop and maintain than traditional search engines, potentially limiting their accessibility or leading to higher costs for end-users.