How to build a content strategy for AI search beyond keywords

Planning content around search volume data alone while LLMs are running numerous backend queries is the fastest way to miss out on content opportunities.

In a 2025 study of 501 prompts using Gemini 3 with Google Search grounding enabled, Seer Interactive found that 95% of the generated fan-out queries had zero tracked monthly search volume. Zero.

So every time you open your keyword tool and filter by volume alone, you’re potentially missing the specific questions AI search systems investigate when answering buyer prompts.

Discovery is moving from keywords and rankings to AI-driven chat experiences. And the real advantage isn’t when you train AI to recognize your brand in these experiences,  when you get AI to recommend it.

Alex Birkett walked through how to extract topics from customer data, reverse-engineer citation patterns, and use a three-part validation model to create an effective content strategy for AI search.

Here’s the system.

TL;DR

Analyze which pages earn AI citations and compare them with traditional search performance. Use those findings to identify content gaps and prioritize formats such as comparison pages, knowledge base articles, case studies, and original research.

Keyword volume alone misses many questions AI search systems investigate. In Seer Interactive’s 2025 Gemini 3 study, 95% of generated fan-out queries had zero tracked monthly search volume.

Build your AI search content strategy around actual buyer questions from sales calls, support tickets, and customer research. Group them into five prompt patterns: product discovery, jobs to be done, pain points, comparisons, and brand-aware questions.

Prioritize topics using Alex Birkett’s three-part validation model: buyer data, product alignment, and channel validation. Score each from 1–10 instead of choosing topics solely by search volume.

Why does traditional keyword research miss AI search opportunities?

Traditional keyword research misses many AI search opportunities because LLMs can break a buyer’s question into multiple detailed sub-queries that have little or no tracked search volume. A content strategy based only on conventional keywords can overlook the specific questions and information AI systems use to construct their answers.

LLMs don’t process queries the way Google does. 

When someone asks an answer engine, “I need a CRM that integrates with Slack and my email,” the system doesn’t search for that exact phrase. It fans out into multiple sub-queries: CRM features, Slack integration capabilities, email sync requirements, and use cases for teams using both tools.

Most of those fanout queries never appear in keyword tools. 

They have no volume data because humans don’t type them into search boxes. Instead, LLMs generate them on the backend to match user intent. And these fanout queries are expanding into longtail variations specific to the context in which the user is searching. 

Seer Interactive’s research found that fanout queries averaged 6.7 words, but could range anywhere from 2 to 17 words.

(Image Source)

This is why keyword planning alone is ineffective when creating a content strategy for AI search. 

You’re optimizing for the question users ask, rather than the 10 questions AI generates to answer it. Keyword tools show search volume for single queries. LLMs use query fanout. The gap between the two is where your content strategy is failing.

Customer research catches what keyword tools miss, while sales call transcripts reveal the specific, multi-part problems prospects ask about. Someone saying “we need to scale our ABM campaigns but keep them personalized” on a call triggers different content needs than “ABM software” in a keyword tool.

How is content planning for AI search different from traditional SEO?

Traditional SEO content planning relies on keywords, search demand, and rankings to identify opportunities. AI search content planning adds buyer questions, conversational prompts, and citation analysis to understand what information AI systems retrieve and reference. The strongest approach combines both rather than abandoning keyword research.

Search volume doesn’t tell the whole story anymore. 

AI search systems can expand prompts into multiple queries and retrieve sources beyond the top organic rankings. Traditional SEO still matters, but rankings alone don’t explain which pages earn AI citations.

Here’s how to build an answer engine content strategy that keeps you visible.

How do you find buyer questions for an AI search content strategy?

Start with sales call transcripts, support tickets, and customer surveys to identify the questions prospects repeatedly ask. Group those questions by buyer problem and intent, then turn them into conversational prompts that reflect how people research products and solutions.

Run customer research before you open a keyword tool. Not alongside it. Before it.

Mine three sources: sales call transcripts, support tickets, and NPS survey responses. These contain actual questions prospects ask when they’re trying to solve problems. Upload anonymized sales calls to tools like Notebook LM and extract themes. Look for repeated pain points, instead of one-off complaints.

And yes, the format matters. 

You need to structure findings as prompts, rather than keywords. 

“The channel… is going to include keywords mostly as validation—nowadays it also includes prompts.”

— Alex Birkett

For example, “I need an AI-based CRM that integrates with Slack” instead of “best CRM software.” Buyers use longer, more specific queries in LLMs because they’re conversing, not searching.

This shift changes what you prioritize. A question that appears in five sales calls matters more than a keyword with 1,000 monthly searches. Especially if those five calls represent your ideal customer profile, asking the exact question your product answers.

Map customer research to five prompt patterns

Group customer questions into categories that match how people use answer engines:

1. Product discovery: “Best X for Y” queries where buyers are exploring solutions in a category.
2. Jobs to be done: “I need to accomplish X” statements that describe desired outcomes rather than product features.
3. Pain points: “I’m struggling with Y” complaints that surface when current solutions fail.
4. Comparisons: Alternative and versus queries that happen during evaluation stages.
5. Brand-aware: Product FAQs and feature questions from people already familiar with your solution.

“From an AI planning perspective…product comparisons, solution evaluation, solving pressing pain points—those are the ones…to prioritize for prompt tracking.” 

— Alex Birkett

Focus on pain point queries first. When users describe specific problems, AI tools surface solutions and product recommendations in responses. 

For example, someone searching “my oven won’t stop smoking” is more likely to get product suggestions than just troubleshooting steps. 

Complaint-based prompts trigger product recommendations more reliably than generic category searches.

This is more important than most teams realize. Pain point content positions your brand as the solution to a problem the AI just validated, rather than one option in a list of alternatives.

How can AI search citations help identify content gaps?

Analyze the pages AI search tools cite for your priority buyer questions, then compare those sources with your own content and organic rankings. Look for topics, page formats, and specific information competitors are being cited for that your website doesn’t adequately cover.

Export citation data from tools like Profound, or use an AI visibility tracker built with n8n to collect citations from monitored Google AI Overviews. Run batch URL analysis in Ahrefs to check which cited pages also rank in traditional search.

Pages that perform in both channels are your highest priorities as they get AI citations and drive organic traffic. But pages with zero keyword rankings that get frequent AI citations are opportunities, too. They can earn visibility in answer engines even when traditional SEO reports show little organic search performance.

Check what page types show up most in citation data. Eight formats are prominent:

1. Listicles ranking solutions in a category
2. Reviews and competitive comparison pages
3. Knowledge base articles answering specific questions
4. Product pages with detailed specs
5. Persona and use case content for specific buyer types
6. Case studies providing validation-stage proof
7. Product-led content like calculators and how-tos
8. Original research and data studies

Build these formats for your priority topics. They perform for both SEO and LLM citations because they match how AI systems structure answers. 

Keep this in mind when choosing page types to include in your LLM citation strategy:

• A comparison page gives the AI clear alternatives to present; 
• A knowledge base article provides specific data points to cite; and 
• An original research offers unique information that isn’t duplicated across competitor sites.

Get this right, and every AI citation becomes a new entry point to your funnel.

The three-part validation model

Planning content with a binary yes/no filter for AI search is futile. Every topic needs three validations scored on a 1-10 scale:

1. Buyer data confirms customers ask about it. 

How often does this question appear in sales calls, support tickets, or customer research?

2. Product alignment shows your solution solves it. 

Can you credibly answer this question with your product or service? Does creating content here reinforce your positioning or dilute it?

3. Channel validation proves people search for it. 

Is there search volume, citation data, or discussion forum activity related to this topic?

For example, a prompt like “SEO agencies in New York” scores high on all three if you’re a New York agency. Buyers ask for local providers, your service fits, and search volume exists. Prioritize it.

Meanwhile, a topic like “B2B buying behavior” might score lower on volume but higher on buyer relevance and product fit if your sales calls reveal this is what prospects want to understand. Use scores to prioritize with limited resources.

This model forces honest evaluation. You can’t create content just because it has volume if your product doesn’t solve that problem. There’s no point in chasing buyer questions if nobody’s searching for answers. And ignoring high-intent customer questions just because they lack volume data shows you never really understood intent in the first place.

How can review sites and online communities reveal AI search content opportunities?

Review sites and specialist communities reveal buyer questions, frustrations, and product comparisons that traditional keyword tools often miss. Sources like G2, Capterra, LinkedIn discussions, and industry forums can help uncover specific problems worth addressing in your content.

Look for recurring complaints, feature requests, and implementation challenges. A pattern of users struggling with a particular integration or workflow, for example, can reveal an opportunity for a comparison page, troubleshooting guide, or practical how-to.

This is where volume-based planning breaks down completely.

These are real questions from real prospects that your keyword tool may never surface because they’re discussed in communities and reviews rather than typed into Google.

How should you prioritize prompts for AI search content?

Prioritize prompts by how frequently they appear in customer research, how closely they align with your product, and whether they indicate commercial intent. Combine those signals with search volume and AI citation opportunities, rather than automatically excluding questions with little or no tracked demand.

Frequency shows importance. Cross-reference customer questions with search volume where it exists. If a repeated customer question has 1,000 monthly searches, prioritize it immediately. If it has zero volume, check if it’s a pain point with high intent, as those are still topics worth considering.

Prioritize bottom and middle funnel topics first. Product comparisons, alternatives, and solution evaluation queries are where your brand appears in AI responses. Top funnel content builds authority but rarely drives direct visibility in answer engines.

This inverts traditional content strategy. 

You’re not building a funnel from awareness to decision. You’re starting at the decision stage where AI tools make recommendations, then working backward to build authority.

How do you turn buyer questions into an AI search content plan?

Start by analyzing sales call transcripts to identify recurring buyer questions, then group them by product and stage of the buying journey. Score each question against buyer relevance, product fit, and channel potential to prioritize the topics most likely to generate meaningful search visibility and qualified demand.

1. Pull three months of sales call transcripts. Upload them to Notebook LM and run a thematic analysis. 
2. Extract 20-30 questions prospects ask. Not questions you wish they’d ask, but questions they actually voice on calls.
3. Format those questions as prompts. Group them by product line and buyer journey stage. Score each on buyer relevance (1-10), product fit (1-10), and channel potential (1-10).
4. Pick your top five prompts based on combined scores. 
5. Create one piece of content for the highest-scoring topic. Use a format AI tools cite: comparison page, how-to guide, or knowledge base article.

Don’t overthink the execution. A 1,200-word comparison article built around a real customer question will outperform a 3,000-word thought leadership piece about industry trends. 

“I would highly advise against scaling out thousands of long-tail pages… They might work now, but they risk SEO performance.”

— Alex Birkett

Answer engines cite specific, actionable content that solves defined problems. That’s your focus.

Why should customer research drive your AI search content strategy?

Customer research connects your content strategy to the problems, questions, and comparisons buyers actually care about. Combined with AI citation analysis and traditional keyword data, it gives you a more reliable way to prioritize content than search volume alone.

Search volume data optimized for Google’s algorithm doesn’t predict what ChatGPT will cite or recommend.

This takes longer to get right. 

You can’t build a useful AI search content strategy by plugging keywords into a content brief generator and spinning up dozens of articles. You need customer research, citation analysis, and validation scoring for every topic. Build a repeatable process, publish content around your highest-priority questions, and measure whether that content earns citations, relevant traffic, and qualified leads.

This is a longer-term investment than producing content from keyword lists, but it gives you a clearer way to test which buyer questions matter, where your brand earns AI search visibility, and which topics contribute to qualified pipeline.

Answer engines are already handling 10-15% of search queries in some industries, and that percentage is growing.

Frequently asked questions

What is query fan-out in AI search?

Query fan-out is when an AI search system breaks a user’s question into multiple related searches to gather information before generating an answer. These searches can include specific features, comparisons, use cases, and questions that don’t appear in traditional keyword research.

How do you find content topics for AI search?

Start with questions from customer interviews, sales calls, support tickets, and online communities. Group recurring questions by buyer intent, then use search data and AI citation research to identify which topics are relevant to your audience and product.

Should B2B marketers stop using keyword research for AI search?

No. Keyword research still helps identify demand and understand traditional search opportunities. AI search content planning adds conversational prompts, customer research, and citation analysis to uncover topics that keyword volume alone may miss.

How do you prioritize AI search topics with zero search volume?

Evaluate whether the question appears in customer research, aligns with your product, and addresses a meaningful buyer problem. A topic with zero tracked search volume can still deserve investment if it reflects high-intent customer needs.

What content formats are useful for AI search citations?

Useful formats include comparison pages, knowledge base articles, product guides, case studies, and original research. Choose the format that best answers the buyer’s question and provides specific, credible information rather than assuming one format will always earn more citations.

What to do next

If you want to apply Alex Birkett’s framework, CXL’s Content strategy for LLM visibility and changing search habits course covers turning customer research into a content roadmap, analyzing AI citation opportunities, and building visibility across search platforms. For the execution side, Optimize pages for AI search with GEO/AEO, led by Steve Toth, covers identifying content gaps, building comparison pages, and optimizing content for AI-driven discovery.