Samuel Schmitt recently shared something that caught my attention: "I spent the last few weeks building a content strategy framework for AI search. Not theory. A step-by-step process with a real example." He then teased the outcome - a 2,700-word guide for building a multi-format strategy optimized for Google AI Overview, ChatGPT, and Gemini.

That framing matters. Most content advice still assumes a simple equation: target keyword -> write article -> rank. Samuel is pointing to a different reality: AI search surfaces answers by pulling from many sources and formats, and your strategy has to reflect that.

In this post, I want to expand on Samuel Schmitt's core idea and turn it into a practical blueprint you can apply quickly. The goal is not to chase shiny AI trends. It is to align your content with how modern search products actually assemble answers.

"Text-only content strategies are leaving visibility on the table." - Samuel Schmitt

What changed: from ranking pages to feeding answer systems

Samuel highlights a few stats that should reset your instincts:

Reddit accounts for 40.1% of all AI citations.
YouTube is cited in up to 29.5% of AI Overviews.
68% of terms triggering AI summaries have fewer than 100 monthly searches.

Read that again: low-volume queries can still trigger AI summaries, and AI systems often cite forums and videos. So if your entire plan is "publish 10 blog posts per month," you are probably under-serving the exact sources AI engines like to cite.

The big shift is that visibility is no longer only about a single blue-link position. It is about being present in the set of sources an AI system considers credible for many related micro-questions.

Samuel Schmitt's framework, expanded into a repeatable process

Samuel listed five core components:

Build a keyword universe using PAA at scale
Cluster keywords by SERP similarity
Identify the best content format per topic
Set priorities based on real SERP data
Map the approach to how AI search works (Query Fan-Out)

Let me unpack each one and show how they connect.

1) Build a keyword universe with People Also Ask (PAA) at scale

Traditional keyword research starts with a few head terms, then expands with variations. The problem is that AI answers are often triggered by question-shaped queries and long-tail intent.

PAA is a cheat code because it exposes the question graph around a topic: definitions, comparisons, steps, troubleshooting, edge cases, and buyer concerns.

Practical way to do it fast:

Pick 3-5 seed topics (products, pain points, category terms).
Pull PAA questions for each seed at scale (via your SEO tool, scraping, or APIs).
Normalize them (remove duplicates, unify phrasing).
Keep the questions as questions. Do not prematurely convert everything into a single "keyword." The question format is useful later when you decide format and angles.

Example: if your seed is "AI meeting notes," PAA often reveals clusters like:

"Is it safe to record meetings with AI?"
"Which tools work with Zoom/Teams?"
"How accurate are AI summaries?"
"Can it handle multiple speakers?"
"How do you share notes with the team?"

This is already hinting that you will not win with one generic article. You need a set of assets.

2) Cluster by SERP similarity to cut the noise

Samuel's point about clustering by SERP similarity is crucial because it anchors your strategy in what Google actually ranks, not just what a spreadsheet suggests.

SERP similarity clustering asks: when I search these two queries, do I see mostly the same URLs and content types? If yes, they likely belong to the same intent cluster and could be served by one strong piece (or one hub with supporting assets).

Why this reduces wasted work:

You stop creating multiple pages that cannibalize each other.
You stop forcing blog posts when the SERP clearly rewards forums, videos, or tools.
You get an intent map that is much closer to how AI summaries fan out across sub-questions.

A simple workflow:

For each question/keyword, capture the top results and their content types.
Group terms where 50-70% of top URLs overlap.
Label each cluster with the dominant intent (learn, compare, solve, choose, configure).

3) Pick the best format per topic (not everything should be an article)

Samuel explicitly calls out choosing the best format per topic: article, video, forum, or tool. This is where most teams need to change their default behavior.

A quick decision guide based on what tends to get cited and ranked:

Article (guides, explainers, comparisons)

Best when the SERP is dominated by editorial content and the intent is:

"What is..." and "how does..."
"X vs Y"
"best..." lists with clear criteria

Video (demos, walkthroughs, proof)

Best when the query implies seeing the thing:

"how to" with steps that benefit from a screen recording
product comparisons where viewers want to watch the workflow
troubleshooting where visuals save time

Forum/UGC (Reddit, community posts, Q&A)

Best when the query is subjective, experience-based, or controversial:

"Is [tool] worth it?"
"What are the downsides of..."
"Alternatives to..."

If Reddit is already heavily cited by AI systems (Samuel's 40.1% stat), then participating in communities and earning real discussion can be an actual distribution strategy, not just "brand awareness."

Tool (calculators, templates, checkers)

Best when the user wants an output, not an explanation:

ROI calculators
checklists and audits
generators and frameworks that deliver a result in 30 seconds

In AI search, tools also earn natural mentions and links, which can improve inclusion in the broader citation set.

4) Prioritize with real SERP data (not gut feel)

Samuel notes that the entire process can take less than 30 minutes. The only way that is true is if you prioritize ruthlessly.

What to prioritize:

Clusters where AI Overviews appear frequently (higher chance of citation visibility).
SERPs with mixed formats (a sign that Google is open to multiple asset types).
Terms with low volume but high intent (remember: 68% of AI-summary terms are under 100 searches).
Gaps where the top results are thin, outdated, or lacking firsthand experience.

What to deprioritize:

Clusters where the SERP is dominated by giant brands and the format is locked.
Topics where you cannot add unique evidence (data, screenshots, expert POV, user stories).

5) Make it match how AI search works: Query Fan-Out

Samuel ties the framework to "Query Fan-Out," which is the idea that AI systems break a prompt into many sub-queries, retrieve information from multiple sources, then synthesize an answer.

If that is the mechanism, then your content should be designed as:

A hub that answers the main question clearly.
Supporting assets that each win a sub-question.
Multiple formats so you can appear in different retrieval channels (web pages, videos, forums, tools).

A practical mental model: one big question, ten small questions. Your goal is not to rank for only the big question. Your goal is to be the best source for several of the small questions so you are repeatedly eligible for citations.

The 30-minute execution plan (a realistic version)

If I were applying Samuel Schmitt's approach today, I would do this:

Choose one topic cluster with commercial relevance.
Pull 30-60 PAA questions for that topic.
Group them into 3-6 intent clusters using SERP overlap.
For each cluster, note the dominant formats in the top results.
Commit to a mini "multi-format set" for the top 1-2 clusters:
- 1 strong guide or comparison page
- 1 short demo video
- 1 community thread or Q&A participation plan
- 1 lightweight template or calculator (if the intent fits)

You are not building an empire in 30 minutes. You are making a sharper plan that avoids the most common failure mode: producing lots of text that AI systems do not cite.

Common mistakes to avoid

Treating AI visibility as a separate channel. It is still search, but with more surfaces and more retrieval sources.
Publishing without checking which formats the SERP rewards.
Ignoring forums and YouTube because they are "not SEO." Samuel's citation stats say otherwise.
Overvaluing volume. Low-volume queries can still be high-impact because they trigger summaries and show strong intent.

Closing thought

Samuel Schmitt's post is a reminder that content strategy is becoming a format strategy. If AI engines assemble answers from a basket of sources, you want to be in that basket in more than one way.

If you want Samuel's full guide, he linked it here: https://lnkd.in/e96sGeiC

This blog post expands on a viral LinkedIn post by Samuel Schmitt, Customer Experience is Everything. View the original LinkedIn post →