Introduction to AI Search
The way Google ranks content is based on understanding search queries and web pages, plus a number of external ranking signals. With AI Mode, that’s just the starting point for ranking websites. Even keywords are starting to go away, replaced by increasingly complex queries and even images.
What are Latent Questions
Latent questions are a profound change to SEO. The word “latent” means something that exists but cannot be seen. When a user issues a complex query, the LLM must not only understand the query but also map out follow-up questions that a user might ask as part of an information journey about the topic. Those questions that comprise the follow-up questions are latent questions. Virtually every query contains latent questions.
Google’s Information Gain Patent
The issue of latent queries poses a new problem for SEO: How do you optimize for questions that are unknown? Optimizing for AI search means optimizing for the entire range of questions that are related to the initial or head query. But even the concept of a head query is going away because users are now asking complex queries which demand complex answers. This is precisely why it may be useful for AI SEO purposes to optimize not just for one query but for the immediate information needs of the user.
Identifying Latent Questions
One way to look at AI search results is to break them down into the questions that the AI answers are satisfying, to identify the hidden query fan-out questions. For example, if you ask Google’s AI Mode how to make pizza dough, the AI Mode will generate a pizza dough recipe. The answer is synthesized from multiple queries.
Example of Initial Query and Query Fan-Out Questions
Initial Query: How to make pizza dough?
Query Fan-Out Questions:
- What ingredients are needed to make pizza dough?
- How much pizza dough does this recipe make?
- What indicates that the yeast is active?
- How long should the dough be kneaded by hand?
- How long should the dough be kneaded with a mixer?
- How long should the dough rise during the first proofing?
- How can you develop more complex flavor in the dough?
- What should the oven temperature be for baking the pizza?
- How long should the pizza bake?
Actual AI Mode Question and Answers
Here are some of the hidden query fan-out questions accompanied by the answers that Google’s AI Mode generated.
Question: What ingredients are needed to make pizza dough?
Answer: “1 and 1/3 cups warm water (around 105-110°F or 40-43°C); 1 tablespoon granulated sugar (or honey); 1 packet active dry or instant yeast (2 1/4 teaspoons); 3 1/2 to 4 cups all-purpose or bread flour, plus more for dusting; 2 tablespoons extra-virgin olive oil, plus more for greasing; 1 teaspoon salt.”
Question: How much pizza dough does this recipe make?
Answer: “This recipe makes enough dough for two 10-12 inch pizzas.”
Question: What indicates that the yeast is active?
Answer: “Cover and let it sit for about 5-10 minutes, until the mixture is foamy and bubbly. This indicates the yeast is active.”
Reverse Question Answering Prompt
You can use the Reverse Question Answering Prompt to identify the underlying questions in any AI Mode answer. You can even use a similar but more precise prompt to analyze your own content to identify what questions the document answers. It’s a good way to check if your content does or does not answer the questions you want it to answer.
Prompt to Extract Questions from AI Mode
Analyze the document and extract a list of questions that are directly and completely answered by full sentences in the text. Only include questions if the document contains a full sentence or sentences that clearly answers it. Do not include any questions that are answered only partially, implicitly, or by inference.
Cultivate Being Mentioned on Other Sites
Something that is becoming increasingly apparent is that AI search tends to rank companies whose websites are recommended by other sites. Research by Ahrefs found a strong correlation between sites that appear in AI Overviews and branded mentions. According to Ahrefs, “So we looked at these factors that correlate with the amount of times a brand appears in AI overviews, tested tons of different things, and by far the strongest correlation, very, very strong correlation, almost 0.67, was branded web mentions.”
Post-Keyword Ranking Era
We are in a post-keyword ranking era. Google’s organic search was already using AI and a core topicality system to better understand queries and the topic that web pages were about. The big difference now is that Google’s AI Mode has enabled users to search with long and complex conversational queries that aren’t necessarily answered by web pages that are focused on being relevant to keywords instead of to what people are actually looking for.
Write About Topics
Writing about topics seems like a straightforward approach but what it means depends on the context of the topic. What “topic writing” proposes is that instead of writing about the keyword Blue Widget, the writer must write about the topic of Blue Widget. The old way of SEO was to think about Blue Widget and all the associated Blue Widget keyword phrases.
Images and Videos
Images and videos, if they’re helpful and answer questions, could become increasingly important as users begin to search with images and increasingly expect to see more videos in the search results, both short and longform videos.
Conclusion
In conclusion, optimizing for AI search requires a new approach. It’s no longer just about keywords, but about understanding the latent questions that users may ask and providing answers to those questions. It’s also about being mentioned on other sites and creating content that is helpful and answers questions. By following these steps, you can improve your chances of ranking well in AI search results.