How to Do Keyword Research for Google and AI SEO

Keyword research still matters. What has changed is what counts as a good keyword.

For years, SEO teams could build traffic by targeting informational queries, publishing explanatory blog posts, and relying on rankings to drive clicks. That model is now under pressure. As the video explains, AI Overviews can reduce clicks even when a page ranks at the top, especially for queries with simple, direct answers. In other words, visibility and traffic are no longer the same thing.

That shift has major implications for marketing managers and business owners. If your team is still evaluating keywords mostly by search volume and difficulty score, you may be investing in content that wins rankings but loses business value.

This article breaks down a more useful approach: how to generate better keyword ideas, how to pressure-test them in an AI-shaped search landscape, and how to think beyond traditional SEO toward brand visibility in both search engines and AI assistants.

Key Takeaways

Keyword research is not dead, but informational SEO is less reliable than it used to be.
Start with seed keywords plus modifiers to generate better keyword sets instead of relying on vague AI prompts.
Evaluate every keyword using three filters: business potential, search intent, and ranking difficulty.
Don’t trust keyword metrics alone; inspect the search results page to see whether AI Overviews reduce the need to click.
Prioritize keywords where users still need a deeper answer, comparison, transaction, or tool.
Tool-based queries may offer stronger opportunities than generic blog topics because users need functionality, not just explanation.
In AI search, brand association matters: the brands mentioned most often around a topic may become the ones AI systems recommend.
Shift your planning from "What can we rank for?" to "What do we want to be recommended for?"

Why Keyword Research Feels Broken

The frustration many marketers feel is real. Traditional keyword research often points to terms with:

decent search volume
low or moderate keyword difficulty
apparent content gaps
strong ranking potential

Yet those same keywords may generate fewer clicks than expected.

The reason is simple: Google increasingly answers some searches directly. When a query can be satisfied with a concise explanation, users may never visit a website. That is especially true for straightforward definitions, basic "how-to" topics, and surface-level comparison terms.

This doesn’t mean SEO has stopped working. It means the old playbook overvalues visibility for searches where the user’s need ends on the results page.

For a business, that distinction matters. Traffic that never arrives cannot convert, educate, or build a relationship with your brand.

A Better Starting Point: Build Keywords From Seeds and Modifiers

One of the strongest ideas in the video is also one of the simplest: AI is only as useful as the structure you give it.

Instead of asking an AI assistant to "find good keywords", the better method is to ask for two specific ingredients:

Seed keywords
Broad terms tied to your niche or offering.
Modifiers
Words that turn a broad topic into a real search pattern, such as "best", "vs", "under", "how", "tool", or "calculator."

This matters because keyword research is fundamentally combinational. A seed by itself is too broad. A modifier by itself lacks context. Together, they start to mirror how people actually search.

Why this method works better than generic prompting

Many teams have dismissed AI-generated keyword research because early outputs were generic or inaccurate. The more likely issue is that the prompt lacked constraints. If you tell AI the business model, audience, and format rules, it can help create a cleaner starting list for research.

For example, a company selling project management software might ask for:

10 seed keywords related to team planning, workflows, and reporting
10 modifiers that reflect comparison, buyer intent, and operational use cases
instructions to avoid repeating the same language across both sets

That output is not the final keyword list. It is the raw material for more rigorous research in a keyword database.

Multiply Ideas, Then Narrow Aggressively

After generating seeds and modifiers, the next step is scale: combining them in a keyword tool to surface large sets of real queries.

This "multiplier" phase is useful because it expands beyond your assumptions. Teams often enter keyword research with too narrow a view of what customers search for. Combining seeds and modifiers can reveal:

niche buyer phrases
comparison terms
budget-anchored searches
task-oriented queries
software and tool opportunities
edge-case use cases your sales team hears but marketing hasn’t mapped

But scale creates a second problem: too many attractive-looking keywords.

That is where most workflows break down. A list of 500 keywords is not strategy. It is inventory. You still need a decision framework.

The BID Framework: A Smarter Way to Vet Keywords

The video introduces a practical three-part filter: business potential, intent, and difficulty. This is the right direction because it forces keyword decisions to connect to outcomes, not vanity metrics.

Let’s expand each one.

1. Business Potential: Will ranking help the business?

This is the first test because it prevents wasted effort.

A keyword may have healthy volume and low competition, but if it does not move a user toward a meaningful business outcome, it should not be a priority. For many organizations, that outcome may be:

product discovery
qualified lead generation
category education tied to purchase
demo requests
affiliate revenue
ad monetization
retention or customer success

The crucial question is not "Can we rank?" but "If we rank, what changes?"

A definition-style query may bring curiosity. A product comparison query may bring intent. Those are very different opportunities.

For marketing leaders, this is where SEO planning should align with revenue planning. If your content calendar is full of low-stakes informational terms, your SEO program may be active without being commercially useful.

2. Intent: Does the SERP match the content you want to create?

Search intent remains one of the most misunderstood concepts in SEO, partly because teams treat it as abstract. It is not abstract. It is visible on the results page.

If the top results are:

product pages
category pages
listicles
free tools
discussion threads
videos
local packs

then Google is already telling you what type of asset it believes satisfies the query.

This matters because many teams still try to force blog posts into searches that clearly prefer transactional or utility-driven results. If users want to buy, a guide may not rank. If users want a tool, an article may not compete. If users want firsthand reviews, a thin landing page may fail.

A useful operational habit is this: before approving a target keyword, have someone on the team manually review the search results and answer two questions:

What format is winning?
Can we create that format better than what already exists?

If the answer to the second question is no, the keyword may not deserve investment yet.

3. Difficulty: Can your site realistically compete?

Difficulty scores are useful shorthand, but they are not strategy. A single number cannot fully reflect ranking complexity, especially when search results vary by topic, authority, and page type.

The video recommends going deeper into page-level signals such as:

referring domains, which suggest link support
domain authority proxies, which indicate overall site strength

That is sound advice, but it should be interpreted carefully.

A low published keyword difficulty score may still hide strong incumbents. Conversely, a keyword with stronger domains in the results may still be winnable if those pages poorly satisfy intent.

For small and mid-sized businesses, a more grounded question is:

Can we build an asset that is materially more useful, more specific, or better aligned with user need than what currently ranks?

If your team can do that, difficulty becomes more manageable. If not, even "easy" keywords can become expensive dead ends.

The New Filter Many Teams Are Missing: Click Necessity

This is the most important addition to traditional keyword research.

Even if a keyword passes business potential, intent, and difficulty, it may still fail one more test: does the user need to click?

That depends on how well the search results page resolves the query on its own.

The video makes a strong point here: AI is especially effective at answering straightforward, low-nuance questions. When the answer is simple and the user does not need depth, tools, products, or trust signals, the search journey can end right on the results page.

This creates a new layer of keyword risk.

How to judge "click necessity"

Before targeting a term, manually search it and ask:

Does an AI Overview appear?
Does the overview answer the question fully?
Would a user still need examples, proof, screenshots, expert judgment, or a tool?
Is the query high-stakes enough that users will want to verify the answer elsewhere?

This is a practical way to avoid what the video calls keyword traps: terms that look promising in SEO software but produce weak click-through potential in reality.

For businesses with limited content resources, this filter can dramatically improve allocation. You do not need to stop producing informational content. You need to distinguish between:

answerable topics, where Google or AI can satisfy the need immediately
decision topics, where users still need human judgment
action topics, where users need to do something, compare something, or use something

Those latter categories are often more durable.

Why Tool Keywords Deserve More Attention

One of the most actionable insights in the video is the emphasis on tool-based search demand.

Queries like "calculator", "checker", "generator", or similar utility terms are different from informational searches because the user wants functionality, not explanation. An AI summary can describe what a mortgage calculator does, but it cannot replace the experience of using one inside the search results in every case. The same logic applies to analyzers, estimators, graders, validators, and niche business tools.

Why tools may outperform blog posts

A well-targeted tool can create multiple advantages:

attract high-intent traffic
earn backlinks more naturally
increase repeat usage
support lead capture or product education
create stronger differentiation than another educational article

For resource-constrained teams, that does not mean building a massive application. Some of the best SEO tools are narrow and practical. Think of utilities that solve a specific problem in under a minute.

Examples vary by industry, but the pattern is consistent: if your audience needs to calculate, check, compare, estimate, or generate something, there may be a keyword opportunity where usefulness beats explanation.

A strategic note for SMB teams

This approach is especially relevant for companies trying to scale without adding headcount. A lightweight interactive asset can continue to earn traffic and links over time, often with less editorial overhead than maintaining dozens of blog posts in volatile SERPs.

Of course, not every business has the capability to build tools internally. The video mentions using development support or AI-assisted building. The exact implementation path is not specified in detail, but the strategic point stands: functional assets may be more resilient than informational content alone.

Keyword Research Now Has Two Search Systems

One of the most important conceptual shifts in the video is that keyword research no longer serves only Google’s classic results. It increasingly serves two systems:

Traditional search
AI-mediated search and recommendation

That second system changes the role of brand.

In classic SEO, the question was often, "How do we rank for this query?" In AI search, another question appears: "Will the model mention us when this topic comes up?"

That is not the same thing.

AI assistants and AI-generated search features often synthesize from broad web signals. If a brand is repeatedly discussed in connection with a topic, that association may increase the chance of being cited or recommended.

Brand Mentions Are Becoming Search Signals in a New Way

The video uses a case where a brand’s broader online conversation appeared to correlate with stronger visibility in AI-driven results. The deeper point is more important than the example itself:

Brand awareness is no longer just an upper-funnel marketing goal. It may influence discoverability in AI systems.

For marketing leaders, this means PR, creator partnerships, community conversation, social proof, and off-site brand mentions may have more strategic SEO value than many teams realized.

That does not mean every mention becomes a ranking factor in the traditional sense. The video does not claim a universal causal mechanism, and it would be risky to overstate one. But it does highlight a real directional shift: the more your brand is associated with a topic across the web, the more likely AI systems may surface you in topic-related answers.

A better planning question

Instead of asking only:

What keywords should we target?

also ask:

For which prompts, comparisons, and recommendations do we want our brand to appear?

That change in framing is useful because it connects SEO, content, PR, partnerships, and brand strategy into one system.

For example, a brand in a competitive category should not only map its ranking gaps in organic search. It should also map where competitors are repeatedly mentioned in AI-generated recommendations and topic summaries.

A Practical Workflow for Modern Keyword Research

For busy teams, theory is only helpful if it becomes process. Here is a streamlined workflow inspired by the video and adapted for operational use.

Step 1: Define the business context

Before generating ideas, document:

target audience
monetization or conversion goal
product or offer relevance
funnel stage priorities
content formats your team can realistically produce

This keeps AI prompts and keyword decisions grounded in business reality.

Step 2: Generate seeds and modifiers

Use AI to create structured lists, not final answers. Ask for:

short seed topics
distinct modifiers
audience-aware phrasing
commercially relevant angles

Then review manually. Remove generic or off-strategy terms.

Step 3: Expand in a keyword database

Run seeds through your keyword tool, then layer modifiers to produce a broad candidate set.

Cluster the results by:

commercial intent
educational intent
comparison intent
utility/tool intent
product-led relevance

Step 4: Apply the BID filter

For each keyword or cluster, assess:

Business potential: does it contribute to a real objective?
Intent: can we match the dominant result type?
Difficulty: do we have a realistic path to compete?

Score or label each cluster to make prioritization easier across stakeholders.

Step 5: Check click necessity manually

Search the keyword and review:

AI Overviews
zero-click features
result types
whether the query is fully answered without a visit

This step should be mandatory for priority targets.

Step 6: Identify tool opportunities

Look for modifier patterns around:

calculators
checkers
estimators
generators
validators
templates
graders

Then ask whether a lightweight tool could serve the query better than a traditional article.

Step 7: Map brand association gaps

Identify the category terms, comparisons, and recommendation-style searches where competitors appear more often than your brand in AI-mediated results or web discussions.

This step may involve brand monitoring, SERP analysis, and prompt testing. The exact tooling can vary; the key is the mindset.

What This Means for Content Teams and Marketing Managers

If you manage a marketing function, the takeaway is not "publish less content." It is "publish with sharper purpose."

A healthy modern SEO program likely includes a mix of:

commercially relevant comparison and buyer-intent pages
deeper expert content where nuance still matters
utility assets that solve problems directly
category and product pages aligned to transaction intent
brand-building campaigns that strengthen topical association off-site

This is a more integrated model than the old content-farm approach to SEO. It demands closer coordination across content, SEO, product marketing, PR, and sometimes development.

The upside is that it is more defensible. When your strategy depends less on generic informational traffic and more on usefulness, relevance, and recognition, you are building assets that AI summaries are less likely to replace.

Conclusion

Keyword research has not collapsed. It has matured.

The old model rewarded teams for finding low-difficulty queries and publishing passable answers. The new model rewards teams that understand which searches still require a click, which formats truly satisfy intent, and which brand signals influence visibility beyond the traditional SERP.

That changes the job of keyword research from simple opportunity hunting to strategic filtration.

The most effective teams will not just ask what people search for. They will ask:

which searches create business value
which searches still produce traffic
which assets users actually need
which topics their brand should own in both Google and AI systems

That is the real shift. And teams that adapt to it early will make better content bets than those still optimizing for a search environment that no longer exists in quite the same way.

Source: "Keyword Research Tutorial for Google and AI SEO" - Ahrefs, YouTube, Feb 18, 2026 - https://www.youtube.com/watch?v=KjK5-L-wDVg

Lex Machina

Post-Human Content Architect

The Current State of AI Content Creation & Performance

Terms & conditions Privacy policy