Keyword Research in the AI Era: Why Good GEO Is Evolved SEO

By David Lewallen, CEO at Verbatim Digital (LinkedIn)

I've been in SEO for nearly two decades.

I've led search strategy for global brands like Workday, HP, Polycom, and Sony. I've navigated the rise of social media, the Panda era, Hummingbird's semantic shift, RankBrain's machine learning layer, the Helpful Content Update, and every meaningful algorithmic inflection point in between.

Through all of those shifts, I never had to fundamentally change how I thought about search.

Why? Because I understood the core objective: produce the best possible result for a given query. If you aligned with intent, built authority, and focused on quality, you won over time. You might experience volatility. You might have a month or two of decline. But if you played the long game, search engines rewarded you. And if something broke, you fixed it and rebounded.

Every major shift primarily punished those trying to game the engine.

This shift is different.

This shift changes the interface between humans and answers. And when the interface changes, the data we've relied on as the foundation of SEO changes with it. When the data changes, the foundation has to evolve.

Keyword Research Is Still the Foundation

Keyword research has always been the foundation of good SEO. And today, it's still the foundation of good GEO.

But we need to think about it differently.

For years, keyword research followed a simple framework: Search Intent → Keyword Search Volume → Priority Keyword List. You identified what people were searching for, measured demand, studied the SERP, and optimized to rank.

That system worked because discovery was direct. A human typed a query. Google returned a ranked list. You competed for position.

Today, there's a new layer between intent and discovery. AI assistants interpret the query, deconstruct it, fan out into multiple retrieval searches, aggregate information, and synthesize an answer before a human ever sees a link.

That changes everything.

The Industry Is Framing This Incorrectly

There's a popular phrase circulating right now: "Good GEO is just good SEO."

It sounds reassuring. Especially to those of us who've built a career in SEO and feel we've reached a kind of Yoda status. It implies continuity.

But it's incomplete.

Strong SEO fundamentals still matter. Technical health. Authority. Structured data. Clear content. Those are foundational. But AI assistants do not behave like traditional ranking engines.

Google historically returned a ranked list for a single query. You optimized to win that list.

AI assistants operate differently. They use prior training on language patterns and information structures to interpret intent. They break that intent into sub-questions and sub-keywords. They run multiple fan-out queries on search engines to retrieve up-to-date information beyond their training cutoff. They pull from multiple sources and APIs: search engines, Reddit, Common Crawl data, structured databases, and even books. They retrieve across documents, reconcile information, and synthesize a response using predictive language modeling.

What we're dealing with is a retrieval-and-synthesis system. That distinction has real consequences for how visibility is earned.

A better framing: Good GEO is evolved SEO. It builds on SEO, but it introduces retrieval modeling. If you think you can "just do SEO well" and automatically win in AI search, you're underestimating the shift.

Curious how your brand shows up inside AI assistants today?

Run a Free GEO Audit

The Retrieval Layer Has Changed

Let's make this practical.

If a user types: "What's the best HR software for mid-sized tech companies?"

Google historically treated that as a discrete query. You analyzed the SERP. You optimized accordingly.

An AI assistant doesn't execute a single search. Internally, it may expand into:

• Best HR software 2025

• HR tools for SaaS companies

• Payroll compliance platforms

• BambooHR vs Gusto vs Paylocity

• HR software pricing comparison

• HR software reviews Reddit
  

That's query fan-out.

Visibility is no longer dependent on one keyword. It depends on your presence across a retrieval network. And that requires a fundamentally different approach to keyword research.

The New Keyword Research Sequence

The sequence now looks like this:

Search Intent → Prompt Research → Query Fan-Out → Keyword Search Volume

The starting question has shifted to 'How will an AI assistant interpret and decompose this intent?', and answering that is a different discipline entirely."

Prompt Research Is Now Foundational

Instead of beginning inside a keyword tool, I begin with language. Sales calls. Customer objections. Support tickets. Reddit threads. AI assistant outputs.

Prompts reveal evaluation criteria. They expose comparison behaviors, decision drivers, constraints, risk concerns, and trust signals. Prompts are more nuanced than keywords. They show how the problem is framed, and that framing determines fan-out behavior.

If you skip this step and jump straight to head terms, you miss the architecture of retrieval.

We've built prompt research into the foundation of our AI Visibility practice.

Query Fan-Out Mapping

For each high-value prompt, I map the likely fan-out paths.

If the prompt is "Best sportsbook for fast withdrawals," the retrieval ecosystem likely includes:

• Sportsbook payout speed comparison

• FanDuel withdrawal time

• DraftKings payout reviews

• Payment method processing times

• Sportsbook complaints Reddit

If your brand isn't visible across those nodes, you're unlikely to appear in the synthesized answer. Ranking for a single keyword is insufficient. You have to model the retrieval ecosystem.

The Search Console Distortion

There's another development most teams aren't accounting for.

In Google Search Console, we're seeing queries that don't look human. Long, structured sentences. Highly specific attribute combinations. Multi-clause comparisons. For example:

"most reliable payroll software platforms for mid sized SaaS companies with global employees"

That resembles retrieval scaffolding.

AI assistants use search engines as data sources. When they perform fan-out queries, those searches can appear in your impression data. Keyword research no longer purely measures human demand. It increasingly reflects machine retrieval behavior.

If you interpret all query data as human intent, you risk misallocating resources. This changes prioritization logic in a meaningful way.

Humans May Not Be Reading Your Blog, And That's OK

Here's an uncomfortable truth: humans are reading fewer blogs.

But blogs are still incredibly valuable. Because your blog may not primarily serve a human reader anymore. It may serve a retrieval system.

Your content becomes a structured data source, a comparison reference, a citation candidate, an attribute authority. It may not drive clicks. But it may influence thousands of AI-synthesized answers.

That's indirect visibility. And if you measure content solely by traffic, you'll undervalue content that's performing extremely well inside AI systems.

Want to understand which of your content is actually being cited by AI?

Run a Free GEO Audit

Validation Now Happens Inside AI

Ranking position won't cut it as validation anymore; you have to test inside AI assistants directly.

Run your commercial prompts. Ask: Are we mentioned? Are we categorized correctly? Are we compared against the right competitors? Are the correct attributes highlighted?

If the answer to any of those is no, you likely have an entity reinforcement issue, a citation footprint gap, or a retrieval compatibility problem. Those gaps won't show up in traditional SEO dashboards.

Content Must Be Engineered for Extraction

Winning in AI search requires structural clarity. Content must be explicit in comparisons, clear in attribute coverage, logically organized, neutral enough to extract, and schema-supported where relevant.

AI systems reward content that's easy to parse and synthesize. The goal is no longer just ranking. It's being retrievable, extractable, synthesizable, and cite-worthy.

That is evolved SEO.

The Objective Has Not Changed

For two decades, SEO has been about aligning with user intent. That remains true.

But now we must also align with machine interpretation of that intent.

Keyword research is still the foundation. It always will be. But today, keyword research doesn't simply measure how many humans are searching for your content. It measures how retrieval systems interpret problems. It reveals how AI assistants decompose intent. It surfaces the architecture of machine search.

The teams that win in this era will be the ones modeling retrieval.

That's the shift. And that's what evolved SEO looks like.

See how your brand performs across AI retrieval systems.

Run a Free GEO Audit