The internet is in a constant state of evolution, paralleling shifts in user behavior. Recent releases of large language models, such as ChatGPT, Bard, and Claude, have ushered in a new era where direct conversations with search engines yield instant answers.

Historically, website owners employed various search engine optimization techniques to ensure their content ranked prominently on search engine results pages for specific user queries. However, we are on the verge of a new era, where users may no longer need to click external links; instead, they can receive responses directly from search engines integrated with large language models.

This impending transformation marks the emergence of a new field known as AI engine optimization. Furthermore, this development poses both opportunities and challenges, including complex queries, product recommendations, and multilingual support. While this shift may raise concerns for content creators, it presents exciting prospects for personalized AI-driven interactions and more efficient information retrieval.

Introduction

The internet is in a state of perpetual evolution, mirroring the ever-changing landscape of user behavior. Previously, people would visit websites to stay updated with the latest news, but this dynamic shifted rapidly. Thanks to continuous advancements in search engine technology, many straightforward queries—such as checking the local temperature, determining a mountain's height, finding the distance between two cities, or learning a country's population—no longer necessitate navigating websites. The advent of powerful search engines has eliminated the need for traditional keyword-based searches, allowing users to engage in direct conversations with these engines to obtain answers.

Generative AI

This recent advancement owes its existence to the availability of generative AI models [1], notably large language models (LLMs) [2, 3]. These LLMs possess the capacity to generate text that caters to a wide array of user requirements, including question answering and document summarization, among others. However, it's important to note that LLMs are trained on data available on the internet up to a specific date. Consequently, inquiries pertaining to information beyond this date may not yield entirely accurate results. Nonetheless, a notable development is the direct connection of large language models to the internet. This enables users to pose real-time questions about current events. For instance, consider the integration of large language models like ChatGPT with the internet [4].

The transformations we observe in search engines and user behaviors didn't happen overnight; rather, they unfolded over the course of more than two decades. We've witnessed a swift evolution from article excerpts to information boxes prominently displayed on search engine results pages. This shift is gradually shaping a user behavior where clicking outbound links in search results is becoming less necessary. Instead, users can access answers directly from the main site. This holistic approach, where questions and answers coexist seamlessly within a single platform, is progressively becoming the default behavior.

When large language models (LLMs) become fully integrated into search engines, it will open up intriguing possibilities for users. They will visit search engine websites and engage in conversations, not only to receive the latest news but also to obtain updates on news developments since their last interaction with the search engine. Users will be able to inquire about the most recent news in a city or country, track the progress of ongoing legislation or political issues, stay updated on sports events, learn about the individuals who supported a particular legislation, explore various perspectives on these topics, and even request news narratives from different political viewpoints, be it right, left, conservative, liberal, and more.

The questions posed to search engines could become even more intricate. Instead of merely inquiring about the distance and available travel options from Lyon to Paris, users might seek information on weather conditions during their journey and request eco-friendly travel solutions for this route.

Users will have the capability to request product reviews, outline their specific needs and budget, and seek recommendations for products that precisely align with their requirements and budget constraints. These interactions could yield quick or in-depth product reviews, comprehensive comparisons of multiple products, and even the option to make purchases directly within the conversation.

Both employers and employees will have the ability to inquire about a company's portfolio from various angles, such as profitability, employee satisfaction, the company's performance over recent years, and more.

Given the constant influx of scientific research publications, scientists will be able to request the latest developments in their respective research fields. They can seek concise summaries of specific scientific articles, inquire about the challenges these studies have addressed, as well as the remaining obstacles and future prospects in those areas.

Developers can specify their requirements and request search engines to create an application and submit it to the application store. The search engines will offer them a comprehensive plan for development and deployment, including server deployment.

These represent just a selection of the intriguing and diverse questions that AI engines will be capable of addressing when fully integrated with the internet. While some of these concepts might seem like science fiction, the past few months have showcased promising initial outcomes, albeit alongside certain issues and challenges.

From SEO to AI engine optimization

However, a significant challenge arises. Historically, humans authored answers on forums, blogs, websites, and other platforms. They addressed questions through blog posts, provided code snippets, and generated an extensive body of online content, which was subsequently consumed by other humans. Machines played a vital role in this process as well.

Due to the sheer volume of articles on any given topic, search engines adopted various methods to present relevant results for user queries. Simultaneously, website owners employed diverse techniques to enhance their website's visibility in search results. Practices like search engine optimization (SEO) offered different strategies for achieving a coveted position in the top 10 search results or on the first few pages.

However, LLM-based search engines no longer require providing links to the sources of all these websites. Instead, they can offer summaries or predictions of content directly to users based on these sources [5, 6].

What does this mean for companies? It signifies the emergence of a novel technique known as AI engine optimization [7, 8, 10, 11, 12] or LLM-based search engine optimization, which will take center stage in company strategies. While this concept shares some similarities with AI-powered SEO [13, 14], the primary objective shifts from improving search engine ranking (SERP ranking) to ensuring a company or its products appear in the predicted or summarized text generated by search engines for user queries. We are still in the early stages of this evolution, and in the near future, we will witness how advertisements are integrated into conversations with search engines.

Personal AI is on the brink of becoming a reality, where answers will be tailored to suit the user's preferences. However, this leads us to a looming challenge for the future. Will we continue generating content as we have in the past? Why would anyone invest in creating a website when users no longer need to visit them for answers?

The question arises: Who will provide the answers to these inquiries in the years to come? These concerns are a source of worry within the blogosphere [5, 6, 9]. At the core of the web lies its openness, but for those of us who hail from the blogging era, adapting to this transformative shift is undeniably challenging.

Last but certainly not least, another significant challenge lies in the limitations of these models when confronted with multilingual questions and answers [15, 16, 17]. These models have been trained on extensive portions of internet content, which tend to be predominantly in a select few languages. This raises questions about languages that are underrepresented on the internet. Does it imply that speakers of these languages will be unable to utilize these search engines? If content creation dwindles on the internet, it may indirectly impact these languages and the future training of large language models.

Conclusion

The emergence of Large Language Models (LLMs) and their integration with search engines have given rise to profound questions and implications. The generative capabilities of LLMs empower search engines integrated with them to respond directly to user queries, obviating the need for users to click on external websites. This paradigm shift raises concerns about the revenue streams of website owners and product creators, as traditional web traffic may decline. To address this, the concept of AI engine optimization is gaining traction, offering strategies for owners to ensure their website content becomes part of the responses generated by conversations with search engines.

Simultaneously, there are apprehensions that some bloggers may cease producing web content altogether due to this changing landscape. Such a trend could have indirect repercussions on less-represented languages on the internet, potentially affecting their integration with LLMs. In this evolving landscape, the intersection of AI, search engines, and content creation presents both opportunities and challenges, driving the need for creative solutions to ensure a harmonious coexistence between the digital world and content creators, while also promoting linguistic diversity and inclusivity in the era of AI-powered interactions.

References

  1. Generative artificial intelligence
  2. Large language model
  3. Bender, Emily M., et al. “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜” Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, Association for Computing Machinery, 2021, pp. 610–23. (PDF)
  4. OpenAI connects ChatGPT to the internet
  5. I Tried Google’s Generative Search. It Will Change Blogging Forever.
  6. Plagiarism Engine: Google’s Content-Swiping AI Could Break the Internet
  7. Forget SEO: Why 'AI Engine Optimization' may be the future
  8. AI Transforms the Marketing Funnel by Leading Human Buyers
  9. AI is killing the old web, and the new web struggles to be born
  10. AI Engine Optimization: The New SEO?
  11. Marketing in the Age of AI Engine Optimization
  12. AI Engine Optimization: The New Frontier Beyond SEO
  13. How to Supercharge Your SEO Strategy with AI SEO
  14. How AI will change the future of search engine optimization
  15. Introducing The World's Largest Open Multilingual Language Model: BLOOM
  16. Introducing LLaMA: A foundational, 65-billion-parameter large language model
  17. List of languages supported by ChatGPT