From SEO Keyword Research to GEO Query Research
Discover what AI search terms to target as you create your Generative Engine Optimization (GEO) strategy.
How AI Search is Changing Keyword Research
Traditional keyword research entails finding the specific phrases that users type into search engines. For the past few decades, marketers have used different tools to discover keywords with high search volumes and then created web pages and content that include those exact terms.
However, AI search platforms are less concerned with exact keyword matches and more focused on the meaning behind the query. In other words, AI search prioritizes user intent and context over the specific phrasing of the search.
One major change from conventional search to AI is the nature of queries themselves. Users interacting with AI tend to use natural, conversational language, often phrasing searches as full questions or requests—a departure from the concise “Googling” we’ve been accustomed to.
For example, a traditional Google query might be “Best budget running shoes 2025,” whereas the same user on an AI platform might ask, “What are the best running shoes I can buy on a tight budget this year? I want to spend less than $100.” The intent is identical—finding affordable running shoes—but the AI query is longer and more nuanced.
Another key difference is that AI search can interpret and clarify intent in ways traditional search engines cannot. If a user’s question is ambiguous or broad, a Large Language Model (LLM) might ask a follow-up question or make an accurate assumption based on context.
In contrast, a traditional search engine would simply return a mix of results. The user would then have to click through to different sites or start over with a more refined search query. For example, if you ask ChatGPT, “How to improve my website?”, it may seek clarification (e.g., “Are you looking to improve traffic, usability, or performance?”) or just provide a multi-faceted answer. A search engine faced with the same query would likely return a mix of SEO tips, design tutorials, and performance optimization articles.
Additionally, AI search changes what it means to “rank” for a query. In traditional SEO, you aim to rank #1 for a keyword by optimizing for query terms and building authority signals. In Generative Engine Optimization (GEO), getting your content featured in an AI-generated answer in any capacity is the new win.
For instance, AI platforms might pull information from multiple sources to answer a single question. They’re not looking for one page that perfectly matches the query phrase. Instead, they’re piecing together information from different sites that together answer the user’s query. In effect, AI is performing the aggregation and ranking internally before presenting the user with content to explore.
The internalization of search has big implications for keyword research. Instead of obsessing over ranking for relevant keywords, marketers now need to think in terms of topics and semantic clusters. This means that traditional keyword research, which yields a list of specific search terms to target, must evolve into intent research—understanding the various ways a user might ask a question and what they actually want to know.
Keyword lists are still useful but they should be grouped by intent. For example, the keywords “buy running shoes online,” “best price running shoes,” and “cheap running shoes free shipping” all signal a similar intent (the user is trying to buy affordable running shoes online). That means rather than creating separate pages for each slight variation, an intent-focused approach is to create one comprehensive resource that addresses the core need in detail (affordable running shoe advice).
Search marketing has been moving in this direction for some time. Modern search engines like Google have incorporated semantic understanding (using algorithms like RankBrain and BERT) and started to reward content that addresses user intent rather than just repeating keywords.
AI search takes this to the next level. Exact-match keywords are far less important now—what matters is whether your content provides a satisfying and contextually relevant answer.
To adapt, marketing professionals must adjust their mindset and metrics. Instead of measuring success solely by SERP rankings, we must focus on visibility within AI answers. This means tracking if and how often your brand or content is mentioned by AI platforms. It also requires optimizing for longer, question-like query variations.
For example, you previously might have optimized a page for the keyword “Accounting software small business.” You should now ensure that the page explicitly answers a question like, “What’s the best accounting software for small businesses?” The content will likely remain similar but phrasing it in a question-and-answer format caters to AI systems that field conversational queries.
The takeaway is that AI search changes what queries we target by forcing us to think beyond 3-5 word search terms. Every query is now part of a larger conversation users are having with AI. Our job as marketers is to understand their intent within that conversation and provide content that directly addresses it.
Identifying AI Queries to Target
If intent is the new focus in search marketing, the challenge is figuring out the exact questions your audience is asking AI. This requires a new research approach since we can’t rely on traditional keyword planners to tell us what people ask ChatGPT and similar platforms.
Analyzing outputs
A logical starting point is to directly analyze the outputs from Large Language Models (LLMs). By seeing what answers AI systems give for certain topics, you gain insight into the questions that are asked and the information being prioritized.
For example, if you’re in the travel industry and want to know what prospective tourists ask AI, you could prompt ChatGPT with, “What are the top questions travelers have about visiting Japan?” The answer might list things like, “Do I need a visa to travel to Japan?”, “When is the best time to visit Tokyo?”, “How much Japanese do I need to know?”, etc. Each of those is a conversational query that real users might very well be asking AI. With this approach, you are using AI to surface common user questions based on the LLM’s training data and usage.
This AI-assisted query research can even complement traditional tools. In fact, SEO practitioners are increasingly using AI to supplement keyword research by generating lists of natural-language questions that can be cross-referenced against search volume.
Another way to identify common questions is to see what AI recommends for follow-up queries. Using a platform like Perplexity, you can input a broad topic and see what additional questions it suggests. Those recommendations effectively reveal user prompts that the AI deems relevant. Similarly, Google Search Generative Experience (SGE) sometimes provides search options that other users are asking.
Expand research beyond AI
Beyond using AI for research, it’s important to look at conversational search trends more broadly. Many SEO tools and analytics platforms now provide data on question-focused queries. For instance, Google Search Console shows long-tail query data (likely from voice searches) that can hint at conversational usage.
Additionally, community forums like Reddit and Quora are great for discovering how real people phrase their questions. These sites often contain full-sentence, natural language queries that are similar to what someone might ask an AI assistant. Compiling such questions helps you map out the landscape of user intent in your niche.
Conduct competitive analysis
Another effective tactic is to conduct competitive analysis within AI results. Punch in different query variations and see what sources are cited—just as you’d analyze what companies rank on Google results.
As we’ve covered, Microsoft Copilot and Perplexity—and increasingly ChatGPT and Google Gemini—provide multiple citations or source links within a single response so look out for brands and specific content that are repeatedly referenced.
For instance, if you ask Perplexity, “How do I consolidate credit card debt?”, it’s going to cite a handful of sites. Study the ones that appear. What queries are they covering? How is their content structured? And, importantly, what types of content does AI seem to prioritize—comprehensive guides, concise answers, or perhaps something else entirely?
Explore internal data
Finally, leverage any available data your company has, such as website chatbot or site search logs. The questions users ask might mirror how they phrase queries in a general AI platform and reveal what information they want to know about your brand.
And of course, there is Google Analytics. We’ll go into detail on how to attribute traffic from AI in another chapter but it’s worth noting now that GA4 can inform your query research. If your site gets visitors from Microsoft Copilot on Bing or Google SGE, you can glean—to some extent—what queries led them there. For example, if you notice traffic coming from Bing with unusually long query strings, it’s likely from conversational AI searches, although it won’t be specifically noted.
The Importance of Monitoring AI Usage Trends
Since Google has been the dominant search engine for nearly 30 years, SEO professionals mostly optimize for its algorithm over other search engines. However, there are currently four major AI search platforms (ChatGPT, Google Gemini, Microsoft Copilot, and Perplexity), plus Anthropic recently gave its Claude model access to the open web. That means query research (and GEO strategy as a whole) could be distributed across 4+ search tools, each of which is slightly different from the next.
Instead of spreading yourself thin, determine what platforms your audience tends to use the most. ChatGPT is used by the general public for a wide variety of questions and Gemini is integrated into Google search—so you’ll want to optimize for both in all likelihood.
However, if you work in the B2B industry, there is a good chance your potential customers are using Copilot (since it’s integrated into Microsoft Windows and Office). For those in academia or research-intensive fields, focusing on Perplexity probably makes sense.
The challenge is that we don’t have explicit query volume data for AI platforms like we do for Google. But we can identify macro trends from surveys and studies—e.g., the fact that ChatGPT and Google Gemini account for 78% of all AI traffic as of 2024 (Semrush, 2024). Keeping an eye on new research and usage stats will help you determine where to focus your efforts.
Exercise: Conducting Query Research
Let’s explore the example of a cybersecurity company that is ready to add a GEO strategy to its marketing efforts. Here are the steps their team could take to identify the right queries to optimize for:
Ask the AI search platform to provide relevant questions: They can start by asking, “What questions do small business owners have about protecting themselves from cybersecurity attacks?”
Note the phrasing of AI responses: The way AI responds can reveal queries to target. It might say, “There are a few aspects to protecting a business from cyber attacks: 1) employee training, 2) using antivirus software, 3) securing networks…” This suggests users might be asking questions like, “How do I train employees on cybersecurity?” and “What are some network security best practices for small businesses?”
Identify cited sources and references: As the team conducts research, they can build a list of sources that are frequently cited. Perhaps a competitor’s blog or a comprehensive report from the FTC.
Analyze the content cited: They can then visit the pages cited or referenced by AI. They notice that content is frequently structured as FAQs, includes expert insights, and refers to authoritative resources like NIST or CISA guidelines.
Find the gaps: The team can look for unanswered questions or insufficiently covered topics in existing content. Perhaps AI compiles information from multiple sources—one from a site briefly covering employee training, another with a basic overview of antivirus software, and another mentioning network security without actionable tips.
Create relevant content: Finally, they can create a comprehensive guide that covers every aspect of cybersecurity people ask AI about. They provide actionable tips at the end of each section, use an FAQ format, include insights from their CEO, and reference NIST and CISA guidelines to give it credibility.
Summary: A Step-by-Step AI Query Research Checklist
In traditional SEO, keyword research is a rather straightforward process. Semrush, Ahrefs, and similar solutions output a list of search terms and you target the ones that are most relevant with the highest search volumes.
GEO query research is not a perfect science. Since we provided multiple ways to discover what people are asking about and how they’re phrasing their questions, let’s conclude by recapping the different tactics:
Analyze AI outputs: Start with a broad query and review AI-generated answers to understand what questions users commonly ask.
Note suggested questions: Identify common follow-up queries and prompts recommended by Perplexity and Google SGE.
Cross-reference queries: Validate AI-suggested questions against traditional keyword data and search trends to make sure they’re worth targeting.
Monitor search trends: See what long-tail, conversational queries are bringing visitors to your site via traditional search using tools like Google Search Console.
Explore community forums: Discover natural-language queries that users pose on platforms such as Reddit and Quora.
Analyze AI-driven traffic: Identify queries from AI-generated search traffic using Google Analytics (GA4).
Leverage internal logs: Analyze data from your own site’s chatbot and internal search logs for query insights.
Perform competitive analysis: Investigate cited and referenced sources to learn which content AI prioritizes for specific queries.
Reverse engineer AI SERPs: Study responses and their referenced sources to look for informational gaps and understand what type of content AI prioritizes.
As we shift from SEO to GEO, the name of the game is no longer using keyword research tools to identify relevant search terms but rather piecing together information from multiple sources to understand how users are conversing with AI to get the information they need. Investing time in this exercise will help you build a comprehensive GEO strategy and, ultimately, create and structure content that surfaces in AI search results.