Hi everyone, this is Neo.

Quick question: when you pick content topics, is the first thing you do open a keyword tool and sort by search volume, high to low?

That habit is so natural we never question it. But today I’m here to pour cold water on it: in the AI search era, picking topics by search volume is systematically filtering out your best content opportunities.

That’s not an exaggeration. It’s the core argument of a recent Search Engine Journal article by Itamar Blauer — and after thinking about it for two days, I’m convinced every independent site seller needs to take this seriously. Because it’s a root-level problem: no matter how well you write, if the topic selection is wrong, everything else is wasted.

1. The Problem: People Don’t “Search” Anymore, They “Talk”

Look at what’s happening.

Most teams have learned to write for AI: structured, summarized, properly marked up — but very few have changed how they decide what to write about in the first place. The result: AI-friendly content aimed at a topic list assembled by sorting a spreadsheet on search volume. The writing moved on; the prioritization is stuck a decade back.

Why does volume sorting no longer work? Because user behavior changed:

Before: someone wants a CRM, opens Google, types “best CRM for small business,” clicks through results.

Now: the same person opens ChatGPT and asks, “Our team of 12, limited budget, want an easy CRM, Salesforce is too expensive — any recommendations? Is migration a pain?”

See it? Same need, different form: “search” became “conversation.” And that conversation carries constraints, decision intent, and implied objections — it’s worth far more commercially, because the asker is closer to buying.

Here’s the catch: keyword tools count strings typed into a search box. Nobody types that conversational question verbatim. Your tools never see this demand — the volume column reads 0.

So you sort by volume, traditional search terms always win, and the genuinely valuable conversational questions never make it onto your list. You end up prioritizing what’s easiest to count rather than what’s most valuable to answer.

2. The Mechanism: Query Fan-Out

Why is this a mechanism change, not just a tooling problem?

Google’s documentation is explicit:

“Both AI Overviews and AI Mode may use a ‘query fan-out’ technique, issuing multiple related searches across subtopics and data sources to develop a response.”

ChatGPT does something comparable — OpenAI’s docs describe it as rewriting the user’s prompt into search queries before retrieval. Different name, same logic:

1 user question → N searches by the model → material pulled from N result sets → one synthesized answer.

This shift is seismic:

• Search didn’t disappear; it moved — from the user’s hands into the model’s
• The competitive landscape changed — you’re no longer competing on one keyword’s SERP, but against every source that could be pulled into a synthesis
• Your page might get cited for a sub-question you never targeted — or get locked out of a topic you thought you owned, because you only covered the head term

This is the root cause of volume’s failure: volume counts what people type; in the AI era, what matters is what the model searches to answer a question.

3. Four Replacement Priorities

So what do you use instead of volume? The article gives four directions — let me expand each with practical applications for independent sites:

1. Prioritize the Sub-Questions, Not the Head Term

Since one question fans out into N sub-searches, those sub-searches are your real targets.

In practice: take your head term and write out the 8–10 things someone needs answered before they could actually act on it.

For “best CRM for small business”:

• What are the pricing tiers?
• How painful is migrating from another tool?
• Which tools does it integrate with?
• How long is the contract? Can I pay monthly?
• If I leave, what happens to my data?

Then check your page: does it genuinely answer those, or just gesture at them? Most pages do the latter — title says it, body skips it. AI synthesis matches on substance, not keyword repetition.

2. Prioritize Entities and Concepts, Not Strings

AI synthesis matches on meaning, not exact phrasing. Exact-match repetition buys you far less than it used to.

What earns you a place in the answer is covering the subject properly: naming the relevant products, standards, methods, alternatives, and how they relate to each other.

In practice: fewer pages built around variants of the same phrase, more pages built around a subject covered thoroughly. Don’t make ten pages for “CRM price,” “CRM cost,” “CRM how much” — make one page that genuinely explains CRM pricing logic. It’ll be cited by everyone asking about price.

3. Prioritize the Decision, Not the Definition

Almost nobody asks AI “what is X” anymore, because definition answers come instantly. People use AI assistants to decide.

So: comparisons, selection criteria, trade-offs, and objections — these upgraded from “bonus sections at the end of a buyer’s guide” to the main battlefield.

• Don’t write “What is Shopify” — write “Shopify vs WooCommerce: a 12-dimension comparison for small teams”
• Don’t write “What is a fulfillment center” — write “Should your first batch use a 3PL? Run these 4 numbers before you decide”
• If your content can’t support a comparison, it won’t be useful to a model answering a comparison prompt

4. Keep Volume Where It Still Decides

Don’t misunderstand — volume isn’t dead. Plenty of queries remain short, transactional, and settled by a normal SERP: brand terms, product terms, local intent, “near me” searches. In those cases volume is still a perfectly good signal, and you shouldn’t rebuild them around conversational prompts.

The key is knowing which category each of your pages belongs to:

• Transactional/brand pages → keep optimizing by volume
• Decision/research content → use the sub-question + entity + decision framework

Two page types, two logics. Don’t mix them.

4. How Do You Rank a Query With No Volume?

Every veteran SEO’s first reaction: “A question with zero volume — how do I compete with a term getting 2,400 searches a month?”

Answer: you can’t rank for the same term — but you don’t need to. Score on proxies instead:

• Does your page genuinely answer those sub-questions? → sub-questions bring long-tail rankings on their own
• Is your topical coverage more complete than competitors’? → completeness wins in AI citations
• Do you cover entities and concepts, not just strings? → semantic relevance decides who AI picks

The logic: content organized around the user’s real decision path blooms across sub-queries, long-tails, and AI citations. You win not “the big keyword,” but “the entire demand network behind it.”

5. How You’d Know If It Worked

Content people fear “did it work and I can’t tell.” The article gives two core signals:

First: is AI actually sending you people? This requires correct analytics setup — GA4’s default AI Assistant channel splits AI traffic across several channels and undercounts it. Fix that configuration first.

Second: ordinary organic performance. A page that properly covers its sub-questions naturally picks up long-tail rankings whether or not an assistant ever cites it. The page speaks for itself.

As for attributing a specific citation to a specific topic decision — you can’t, yet. Assistants don’t give you that data. Treat this as a directional, long-horizon program, not a campaign you can attribute to the tee.

6. Neo’s Take

1. Keyword research isn’t dead — the volume-sorting shortcut is

The author’s closing line is perfect: “Keyword research isn’t dead. What’s dead is treating a search volume column as the arbiter of what deserves a page.”

Volume as a signal was never wrong — treating it as the only arbiter was. In the traditional search era, volume approximated demand well enough. In the AI era, the gap between “strings typed into a search box” and “actual demand” has widened dramatically — because demand now flows mostly into conversations, not search boxes.

2. For independent sellers, this is a low-cost overtaking opportunity

Big companies’ content advantage is scale: more budget, more keywords, more pages. But the AI-era topic logic changed — winners aren’t those who publish the most, but those who understand the user’s decision path the deepest.

That’s precisely the small seller’s opening:

• How do customers phrase questions on sales calls? → that’s the user’s real “prompt”
• What do customers agonize over in support emails? → those are your most valuable sub-questions
• What keeps recurring in competitor reviews and community threads? → your topic goldmine

Big companies can’t easily get this material — their processes sit too far from users. Your advantage is proximity to the user. Use it.

3. Swap “volume thinking” for “decision-path thinking”

Here’s a quick self-check list I run before every topic decision:

• What 8 questions does my target customer need answered before making this decision?
• Am I helping them decide, or just explaining a definition?
• If the user said this question to an AI assistant, how would it sound? Can my page handle that?
• Am I building “one topic,” or a pile of “keyword variants”?

Answer those four, and topic quality basically takes care of itself.

4. Don’t abandon traditional search

One last reminder: this new logic is additive, not subtractive. For brand terms, product terms, and “near me” searches, volume is still the most reliable signal — keep using it. Just stop letting it decide everything. The content map of the AI era is far more complex than an Excel sheet.

AI searches on the user’s behalf — but users never wanted “search results”; they wanted answers. Whoever prepares the best answers wins.

It’s time to upgrade your topic logic. Instead of obsessing over volume numbers in your tool, dig through your sales records and support chats — that’s where the most valuable content topics of the AI era are hiding.