Keyword Research For the AI Era: What Changed and What Still Works

By Garry M. Callis Jr.

Keyword Research For the AI Era: What Changed and What Still Works

A lot has happened this year, and it's time to look under the hood of your keyword research strategy. Is it still effective? Or are there some kinks that need to be worked out? Hopefully this article puts some things into perspective.

Key Takeaways



You pull up your keyword research tool. A target term shows 4,200 monthly searches, moderate difficulty, clear informational intent. You optimize the page, build the internal links, wait out the indexing cycle. The page climbs to position two. Traffic barely moves. And now you're looking at your analytics, scratching your head. You feel as if you have done everything right, but are still missing something, and you'd be right.

I've seen this complaint all over Reddit lately: practitioners ranking for exactly the term they targeted and getting nothing for it. AI has basically rendered the click optional for a huge share of queries, and the keyword research process built for a click-based world hasn't caught up. It's the same shift I wrote about in the AI buyer's journey: brands don't control discovery the way they used to, and that includes the keyword that used to guarantee a visit.

You check the SERP. There's an AI Overview sitting above your result, answering the question in 180 words. The user got what they needed. They didn't click. We're fully in the Zero-Click era now. Where queries are answered and sales convert from surfacing an answer in AI Overviews and AI Search.

This is the problem with applying a 2022 keyword research process in 2026. The volume, the intent, and the ranking is real. Don't be surprised when you're trying to use an outdated method to get a modern result and get a shocked Pikachu face when what you wanted didn't pan out.

What I've realized is that keyword research is still necessary. The thing is, this process is going to need a fresh coat of paint. Hopefully this article gives you everything you'll need to put keyword research back in your arsenal.


What Traditional Keyword Research Still Gets Right



The fundamentals of keyword research are still exactly the same. Intent classification still matters. Understanding whether a query is informational, commercial, navigational or transactional still determines what kind of content to build. Competitor gap analysis still surfaces topics your audience cares about that you haven't covered. Keyword research will still tell you something real about the competitive landscape and where to start.

Contrary to popular belief, long-tail keywords still work, arguably better than ever. AI Overviews dominate high-volume head terms, synthesizing answers that satisfy the query without a click. Long-tail queries are specific, conversational and made up of multiple words. They trigger AI Overviews far less consistently. A query like "best project management software" is almost certainly going to return an AI Overview. A query like "project management software for remote teams under 10 people with a free tier" is more likely to route a user to actual results they need to click through.

The practitioners writing off keyword research entirely are making a mistake. The ones applying it exactly as they did three years ago are leaving money on the table. The right move is a calibrated update, not a rebuild from scratch.


What Changes When AI Enters the Picture


AI Overview Presence is (arguably) The New First Filter

According to Xponent21's own analysis, Google AI Overviews now appear in over 60% of U.S. search queries. That number was under 10% two years ago. For a practitioner doing keyword research today, that means a significant share of your target keyword list is returning a result format that actively reduces clicks to source pages.

Before you evaluate volume or difficulty, you need to know: does this query trigger an AI Overview, and if it does, does that Overview leave something unanswered?

The AI Overview trigger test is simple. Run the query in Google. If an AI Overview appears, read it. Ask whether a user who got this answer would still need to click somewhere. If the answer is no, the Overview gave them everything, and that keyword's click value has dropped significantly. It still has citation value, which we'll cover shortly. But it is no longer a reliable traffic driver at its stated search volume.


The Question Behind the Query Matters More Than the Query Itself

Traditional keyword research optimizes for terms, specific strings people type. AI engines respond to intent and conversational natural language.

A user who types "CRM software" and a user who asks "what's the best CRM for a 5-person sales team that already uses Slack" have the same surface-level keyword but completely different intents. AI engines handle the second query with a synthesized recommendation. Traditional search returns a list of results for both. By thinking of your strategy from the conversational queries your customers are having, you can create better content for them.

The implication for keyword research: the question behind your target keyword is as important as the keyword itself. Content that explicitly answers the question, not just ranks for the term, earns citations and satisfies the intent that drove the search.


Keyword Clusters Replace Individual Keywords as The Unit of Strategy

When an AI engine handles a query, it doesn't just answer the question asked. It expands into related sub-questions, draws from multiple angles of the topic, and synthesizes across sources. The practitioner who has built out a cluster of content around a topic gives the AI more surface area to cite from. The practitioner who has one optimized page targeting one keyword gives the AI one opportunity, which it may or may not take.

The keyword cluster, a set of related queries and questions that map to a topic from multiple angles, is the unit of AI-era keyword strategy. Individual keyword targeting still informs each piece. But the cluster is what earns sustained citation presence across a topic. A recurring theme of this article is going to be "query fan out," the set of related queries, and you'll have to cast a wide net in order to catch all of these.


How to Research What AI is Actually Answering in Your Space

Two gentlemen seated at a desk looking at a computer.

Prompt Testing as Keyword Research

The most underused keyword research method available right now is running your target queries directly in AI engines and logging what they say.

Open ChatGPT, Perplexity and Google AI Mode. Run the top 15 queries from your keyword research. For each one, note whether the AI answers it directly, what sources it cites, what related questions it suggests, and what it doesn't answer.

The gaps in the AI's answers are your content opportunities. If every AI engine gives a solid answer on "what is generative engine optimization" but hedges or goes vague on "how to measure GEO performance," that's a gap with active search demand behind it. Build the content that fills it. It also wouldn't be a bad idea to cross reference those queries with your own Google Search Console data and see if you're already close to answering these questions. That's one of the reasons this article is being written in the first place.

It's honestly been a little funny running this exercise on our own back catalog. A surprising number of queries I tested in ChatGPT and Perplexity pulled answers that traced straight back to pain points our community listening work had already flagged. That's not a coincidence we take lightly. It tells us Discover AIO's read on what practitioners are struggling with is staying current, and as a learning platform, staying current with our own readers' pain points is the whole job. Don't be afraid to meet people where they are, that's how you're able to answer the BIG questions they may have, and they can cite you as not only an answer, but a source of truth.

Citation Gap Research

Ahrefs' research on AI Overview citations found that over 60% of cited sources fall outside the top 10 organic results for the same query. Your competitors may be getting cited for queries you rank for but don't appear in. That's a citation gap, and it has a content solution.

Run your target queries and check which domains appear as cited sources in AI Overviews. If a competitor is consistently cited for a topic cluster where you rank organically, look at what their cited content does differently. It's almost always one of two things: their content answers the specific question more directly, or they have more supporting content around the topic that gives the AI more to draw from.

Community Listening as a Keyword Source

A man putting his ear out to listen

Forums, community threads and comment sections surface questions practitioners ask before they formalize them into search queries. A marketer asking "how do I know if my AI SEO work is actually doing anything" in a Slack community hasn't typed that into Google yet. But they will. Or someone like them will.

Community listening is prospective keyword research. The questions being debated in communities today are the queries that will have search volume in six months. Discover AIO's own community is a source of this signal. The questions members bring to calls and threads are an early indicator of where practitioner demand is heading. Take a couple of hours, go into Reddit and other web forums and just interact with people. Talk to people, answer questions, post about stuff. You never know what your community feels or goes through unless you're there to answer and talk about these things directly and with intent.

I'll give an example. There's a trend that's been going around on Reddit. A lot of people are asking variations of the same question. "Is SEO dead?" Obviously the answer is no, but of course people are going to fan-out and ask variations in order to catch the umbrella. Query fan out happens to be a huge part of Google's new generative engine optimization update. (which you should totally check out) When you're creating content around your research in community listening, you're going to encounter contrarian thought processes, and that's totally okay. As a matter of fact, factor those into your research. It's goof to have rigor and thoughts that may go against the grain, because it brings in perspectives from something you didn't consider initially.

A Revised Keyword Research Workflow

The updated process adds two steps to the front of the traditional workflow and one step at the end. Everything in between stays the same.

Step 1: Standard keyword research. Volume, difficulty, intent classification. Nothing changes here. Build your list the way you always have.

Step 2: AI Overview trigger test. For every keyword on your shortlist, check whether it returns an AI Overview. Categorize each: no AI Overview (traffic play), AI Overview with citation opportunity (citation play), AI Overview that fully resolves the query (low traffic value, high citation value only).

Step 3: Click behavior assessment. For queries that trigger AI Overviews, read the Overview. Does it leave something unanswered, a follow-up question, a specific use case, a decision the user still has to make? If yes, there's still click potential. If no, the keyword's click value is lower than its volume suggests.

Step 4: Question cluster mapping. For each keyword you decide to target, run it in AI engines and log the related questions the AI surfaces. Those questions are your supporting content targets. Build content that answers them and you increase your citation surface area across the cluster.

Step 5: Citation gap check. Who is AI citing for this topic right now? Are you among them? If not, what do cited pages do that yours doesn't? That analysis directly informs your content approach.

Tools That Still Work

A box of tools, hammer, ruler, vice, etc.

Standard keyword tools, including Semrush, Ahrefs and Moz, remain valuable for Steps 1 and 2. Their volume and difficulty data is still accurate. What they don't tell you is whether a keyword's traffic value has been compressed by AI Overview presence. That step is manual for now.

For citation monitoring, tools like Otterly.ai and the prompt testing method above give you the most direct signal of where you're getting cited and where you're not. DiscoverAIO's community is actively tracking what's working across different niches and verticals. That practitioner signal is hard to replicate with any single tool.


What This Means for Your Content Strategy

The thought process around content strategy needs to change for the time you're in. Otherwise, you'll be left at the starting line, looking like John Travolta on that one scene in Pulp Fiction. (IYKYK)

High-volume head terms with AI Overview dominance: target for citation, not click. Optimize these pages to be cited sources within the Overview, not click destinations. The measure of success changes from traffic to citation frequency. The A in AEO is Answer. Optimize your content so you can answer your customers' queries immediately without them having to do too much work.

Mid-volume conversational queries: highest value targets right now. These trigger AI Overviews less consistently, still drive meaningful clicks, and are often underserved because competitors are chasing the head terms. A 600-search/month conversational query with no AI Overview dominance is often worth more than a 4,000-search/month head term that's been absorbed into an Overview.

Long-tail specifics: highest click-through potential, lowest AI Overview interference. Build these deliberately as part of your cluster strategy, not as afterthoughts, but as the pieces that capture the users who need more than a synthesized answer.

Talk about everything: some keywords are worth targeting even if they drive no direct traffic. This is more related to the fan-out we talked about earlier. If you're in the plumbing biz, and you're trying to cover as many queries as possible, there's no issue to devoting time to casting a wider net. There's a more than 0 percent chance that someone, at some point, is going to look for the hyper-specific thing you made that one piece of content on, and BAM!, you're immediately the source of truth because you took the time to talk about it.

Picture this: a content manager at a B2B SaaS company has been falling behind in impressions on GSC and GA4. They read this article and started taking the steps outlaid here. Yes they may have seen a decrease in clicks, but their impressions skyrocketed when they started creating content unique to them, their business, and the use cases outlaid in their research. Because of how the buyer's journey is changing, be prepared to think more about how AI is changing the playing field, my aforementioned article should cover that, so give it a read.

Frequently Asked Questions

3 blocks with the letters F. A. Q. on a table. Meaning frequently asked questions.

Is Keyword Research Still Worth Doing in 2026?

Yes, but the question it's answering has changed. Traditional keyword research told you which terms to rank for. In the AI era, keyword research tells you which queries to target for traffic, which to target for citation visibility, and which questions to build content around that AI engines aren't answering well yet. The tool is the same. The decision framework around it is new.

What Types of Keywords Are Most Affected by AI Overviews?

Informational head terms are most affected, queries with clear answers that AI can synthesize and deliver directly. Commercial and transactional queries are less affected because AI engines are more cautious about making purchasing recommendations without user context. Local queries route to local packs rather than AI Overviews for many intent types.

How do I Know if a Keyword Triggers an AI Overview?

Run the query in Google and check. AI Overview presence varies by user, location and query phrasing, so run it several times and in incognito mode for a cleaner signal. Tools like Semrush and SE Ranking are building AI Overview detection into their SERP features tracking, and that data is increasingly available at scale.

Should I Stop Targeting High-Volume Keywords?

No, but rethink your KPIs. High-volume keywords that trigger AI Overviews are citation targets, not traffic targets. Optimize the content to be the source the Overview cites. Track citation frequency, not clicks, as the measure of performance for those keywords. In marketing, if you hear something more than 3 times, it's a good chance it'll be ingrained in your mind. Citations work in the same manner.

What Tools are Best for Keyword Research in the AI Era?

Standard keyword tools, including Semrush, Ahrefs and Moz, remain the foundation for volume and difficulty data. Layer in manual AI engine testing across ChatGPT, Perplexity and Google AI Mode to assess citation gaps and question clusters. Citation monitoring tools like Otterly.ai give you direct visibility into where you're being cited. No single tool does all of it yet. The workflow is still partially manual.

Where To Go From Here

If you're rebuilding your keyword process for AI search, you're not doing it alone. DiscoverAIO is where practitioners are working through exactly this, sharing what's moving the needle, what tools are worth using, and what the data is actually showing. Join the community and bring your unique perspectives. We're all still learning to be better marketers, and your insights could be the key to advancing us all.