Generative Engines Are Rewriting Search Results

If you’re constantly looking things up on ChatGPT, Perplexity, Bing Copilot, or just reading AI-Overviews on Google, congrats, you’ve moved from search engines to Generative Engines. You are no longer content with ten blue links. You want answers in plain language, with clear recommendations, and ideally, the sources for when you feel more adventurous.

Generative Engines (a.k.a GEs) blend parts of search engines and AI language models. Instead of just showing you a list of websites based on heuristic ranking, they search on your behalf, summarize the key information, re-rank it using their own cognitive analysis, and respond in a way that feels human.

Under the hood, GE is an amalgamation of several systems working together. While we can’t know the exact proprietary architecture of each GE, we can logically break it down into a few general building blocks. Unlike conventional search engines that return relevant listings or standalone Language Models that generate responses based on their training, Generative Engines perform a well-coordinated orchestration of several underlying subsystems. Here’s a simplified look at the typical subsystems within a Generative Engine -

Query Reformulation: GE begins by breaking down complex queries into simpler sub-queries that are easier for search engines to process.

Search Engine Integration: These sub-queries are then sent to search engines to retrieve documents from the internet or other databases.

Summarization: Retrieved documents are processed and summarized to extract the most relevant information.

Response Generation: A final model synthesises information from multiple sources to generate a comprehensive, coherent response.

Fig-1: How GE process user Query

This architecture enables GEs to handle more complex queries, provide more nuanced responses, and maintain a higher contextual accuracy than traditional search engines and a higher factual accuracy than standalone Language Models.

Here’s how Generative Engines stack up against what we’ve used before.

This shift brings both new challenges and opportunities for anyone creating content online. Traditionally, visibility meant getting your page to rank on the first page of Google. However, with Generative Engines answering user questions directly, users might not always click on links. Yet, being cited remains incredibly valuable. It helps build strong brand awareness, and positions your content as authoritative. Most importantly, it drives high-intent traffic, which can significantly increase your conversion rates.

This means the definition of success is changing. Instead of optimising your content just to rank higher, you now want it to be cited within a GE’s response. In this new landscape, content that gets referenced by the AI is the content that wins attention. This is giving rise to a new practice, Generative Engine Optimization ( a.k.a GEO). It’s still early, but we’re already seeing some patterns. Content that’s clearly structured, backed by real data, and written with authority is more likely to be picked up. That includes quoting credible sources, citing statistics, and staying focused on substance over fluff. GEO also varies across domains.

Techniques that work well for health or finance content may not work the same way for education, entertainment, or B2B marketing. But some fundamentals remain consistent - use of statistics, a confident and informative tone, transparent sourcing, and quotable writing all improve your chances of being cited. This is a significant shift in how content visibility is achieved. We’re moving from optimising for clicks to ensuring we’re the source behind meaningful answers.

Generative Engines are more than just the next phase of search. They are changing how we discover, trust, and engage with information online. For users, it means faster and clearer answers, and for content creators, a newer kind of challenge. It is no longer just about being ranked; it is about being visible. This shift asks us to think differently. If your content is accurate, factual, well-structured, and easy to quote, you are already one step ahead in this new landscape.

On a related note, I’m working on a visibility tracking platform to help businesses to navigate this evolving environment. It will go beyond the current metrics used by tools like Otterly, AmIOnAI, etc. Curious to see what’s coming? If you’d like early access to a closed beta (when launched) or just want to chat, feel free to reach out. Would love to connect!

This blog post includes insights from the research paper “GEO: Generative Engine Optimization” by Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan, and Ameet Deshpande (KDD ’24, August 25–29, 2024, Barcelona, Spain). You can read the full paper at: https://generative-engines.com/GEO/