SEO Keywords: How to Find and Use the Right Ones (Across Google and AI Search)

In this article, you’ll learn how to find SEO keywords worth your time, prioritize the ones that actually move pipeline, place them where search engines and AI engines reward you, and measure performance across both channels. You’ll get the same workflow used by teams that rank on Google while also getting cited inside ChatGPT, Perplexity, Claude, and Gemini.

What Are SEO Keywords?

SEO keywords are the words and phrases people type into a search engine to find information, products, or solutions. When the words on your page match the words in someone’s query, search engines treat your page as relevant. Relevance is what gets your page surfaced in results.

That definition still holds in AI search. ChatGPT, Perplexity, Claude, and Gemini do not show ten blue links, but they read pages the way Google does. They tokenize content, index meaning, and rank sources by topical relevance. The page that wins on Google for “best CRM for small business” is, more often than not, also the page being cited when someone asks ChatGPT the same question.

You are not optimizing for two different worlds. You are optimizing for one body of content that has to perform in two surfaces.

Why SEO Keywords Still Matter in 2026

Three reasons.

First, organic search still drives the largest share of inbound traffic for most B2B companies. You cannot opt out of Google.

Second, AI engines lean on the same content the SERP leans on. In our analysis of 83,670 AI citations, the pages cited inside ChatGPT, Claude, and Perplexity overwhelmingly come from sites that already rank well in traditional search. SEO is the upstream supply chain for AI visibility.

Third, the buyers using AI engines are deeper in the funnel than the average Google searcher. AI traffic converts at higher rates because the user has already let an LLM filter the options. By the time they click through, they are deciding, not browsing.

This is why you do not throw away the keyword playbook. You extend it. We argue this in more detail in our piece on GEO vs SEO.

The Four Types of SEO Keywords (and What Each One Is For)

Every keyword falls into one of four intent buckets.

Match the format to the intent or you will not rank, no matter how good the writing is. A blog post will not outrank a product page for “HubSpot pricing.” A pricing page will not outrank a guide for “what is a CRM.” If you want a deeper read on this, our piece on search intent breaks down how to spot it from the SERP.

Beyond intent, two more categorical splits matter.

Short-tail vs long-tail. Short-tail keywords like “CRM software” have high volume and brutal competition. Long-tail keywords like “best CRM for a 10-person consulting firm” have less volume but stronger intent and a much shorter path to a deal. If you are a newer site, you compound faster on long-tail. We have a deeper breakdown in long-tail vs short-tail keywords.

Branded vs non-branded. Branded queries (“HubSpot pricing,” “HubSpot vs Salesforce”) capture demand that already exists. Non-branded queries (“best CRM software”) generate demand that did not. You need both. If 80% of your traffic is branded, you are running a retention machine, not a growth one.

How to Find SEO Keywords (a Five-Step Process)

Most “keyword research” advice tells you to dump a seed term into a tool and download the CSV. That is not research. That is data extraction. Here is the actual workflow.

Step 1. Start With Five Seed Keywords

Seed keywords are short phrases that describe your category. Five is enough.

If you sell project management software, your seeds might be project management, task management, team collaboration, work management, and project planning. If you run an AI search analytics platform like Analyze AI, they might be AI search visibility, LLM brand monitoring, generative engine optimization, AI citation tracking, and AEO.

The point of seeds is to give the next steps a structured starting point. You are not trying to be exhaustive yet.

Step 2. Mine Google’s Free Surfaces

Before paying for a tool, Google itself will hand you keyword ideas at three different points on the SERP.

Autocomplete. Type your seed and read the dropdown.

People Also Ask. Scroll into the results and you will find the questions Google has clustered around your seed. Each one is a real question with real demand and a clear H2 if you write the article.

Related Searches. At the bottom of the page, Google surfaces eight to ten related queries. These are usually long-tail variants and they tend to be much less competitive than the seed.

You should walk away from this step with 60 to 100 candidate keywords. If you want to scrape this systematically, the 12 best free keyword research tools can do the legwork. Our free keyword generator tool is a good starting point if you do not yet have a paid subscription.

Step 3. Layer in Volume and Difficulty

Google’s free surfaces give you ideas. They do not give you volume, intent labels, or difficulty. For that, plug your candidate list into a keyword research tool.

What to look for.

Search volume tells you the size of the pond. It does not tell you whether you can fish in it.

Keyword difficulty tells you how hard the pond is to fish. Most tools score this 0 to 100 based on the link profile of currently ranking pages. As a rule, if your domain authority is below 40, do not target keywords above 50 in difficulty. The math is unpacked in keyword difficulty, and you can sanity-check any term with our free keyword difficulty checker.

Click-through potential matters more than volume. Some keywords have huge volume but get hijacked by featured snippets, AI Overviews, or knowledge panels. People see the answer and never click. Spot these with our free SERP checker.

A note on volume accuracy. Don’t over-trust it. Our study comparing GSC impressions to Google Keyword Planner volume shows the numbers rarely match. Use volume as a relative ranking signal, not an absolute forecast.

Step 4. Steal From Your Competitors

One reliable way to find keywords you should be ranking for is to find the keywords your competitors already rank for.

In Ahrefs or Semrush, drop a competitor domain into Site Explorer. Filter to keywords ranking in positions 1 to 10. Sort by traffic value. You are looking for two things.

Gaps. Keywords they rank for that you do not. These are direct opportunities.

Weak spots. Keywords where they rank in positions 8 to 15. These are pages you can outrank with a better article. Authoritative external sources like the Search Engine Land guide on optimizing content for people and search engines are useful checklists for what “better” actually means.

Do this for three competitors. Combine the lists. You now have a candidate set that is provably commercial because someone else is already paying writers to rank for it. For the deeper structure of this work, see SEO competitor analysis.

Step 5. Extend the Same Research Into AI Search Prompts

This is where most teams stop. It is exactly where you should not.

Your buyers are also typing into ChatGPT and Perplexity. The keyword “best CRM for small business” still exists in Google. But in ChatGPT it might land as: “What CRM should I use if I run a 12-person agency, need HubSpot-style contact tracking, and don’t want to pay over $100 per month?”

Same buyer. Same intent. Different language.

Prompt research is taking your high-priority keywords and identifying the natural-language prompts a real buyer would phrase around them, then tracking those prompts.

In Analyze AI, you do this in two motions. First, enter the prompts you already know matter and watch them.

The dashboard shows visibility, sentiment, position, and which competitors are getting cited alongside you.

The second motion is to let the system suggest prompts you have not thought of yet. Analyze AI looks at your category, your tracked competitors, and how AI engines actually phrase questions in your space, then proposes prompts you can track or reject with one click.

That suggestion engine is the equivalent of Autocomplete and People Also Ask, but for AI search. For more on the overlap between keyword and prompt research, see AI keyword research.

Organize Before You Write

A 200-keyword spreadsheet that nobody clusters is a graveyard. Organize first.

Sort your keywords by topic theme. Each theme becomes a content cluster. Each keyword inside that cluster becomes either a pillar page or a supporting page. This is also how you avoid keyword cannibalization, where two of your own pages compete for the same query.

For each keyword, capture five fields. Keyword, intent, volume, difficulty, and the URL you plan to use. We have a full breakdown of cluster structure in keyword grouping and the broader frame in our keyword strategy guide.

Then prioritize. The keywords worth writing now are the ones that pass three filters at once.

A 50,000-volume keyword that fails any of those three filters is not worth pursuing. A 400-volume keyword that passes all three is.

How to Use SEO Keywords on a Page

Finding keywords is research. Using them is execution. The execution part is simpler than people make it.

Title tag. Your primary keyword goes near the front of the title. Keep titles under 60 characters. We have a guide on title tags if you want the full treatment.

H1. The H1 should be a near-match of the title tag.

URL. Short, lowercase, and includes your primary keyword. Skip stop words. “/seo-keywords” beats “/blog/post-2026-the-best-guide-to-seo-keywords.”

First 100 words. Mention your primary keyword in the opening paragraph. Search engines and AI engines both weight the first chunk of a page heavily.

Meta description. Include the primary keyword. Write it like ad copy because it functions as ad copy. Our piece on writing meta descriptions covers the templates that work.

H2s and body. Use secondary keywords inside H2s and naturally throughout the body. Do not stuff. If a sentence sounds odd with the keyword in it, rewrite the sentence without it.

Internal links. Link from this new page to relevant pillar pages on your site, and from existing pages back to this new page. Anchor text should be the target keyword or a close variant.

The deeper move is comprehensive coverage. Modern ranking is less about keyword density and more about whether your page answers the full set of questions a searcher has. If the keyword is “email marketing software,” the page should cover features, pricing, deliverability, integrations, and migration. Coverage is the new keyword density. We unpack this in keyword optimization and our 9 on-page SEO factors.

How to Measure What’s Working (Across Both Channels)

You can’t improve what you don’t track. Three layers.

Layer 1. Google Performance

Search Console for impressions, clicks, average position, and queries. Pair with GA4 for sessions and conversions. This is table stakes.

Layer 2. AI Search Visibility

This is the layer most teams are blind on. You need to know which prompts mention your brand, where you sit in the answer, which competitors get cited alongside you, and which sources the AI engines pull from.

Once you have those numbers, you can do the work that actually moves them. Win the prompt your competitor is winning. Get cited on the third-party site the AI engines are quoting. Patch the prompt where your brand is missing entirely.

The third-party point matters. Roughly 83% of AI citations come from third-party sources, not from your own site. So tracking which sites the AI is pulling from is half the visibility game. The Sources view shows you that map.

If techradar.com is being cited 140 times for prompts in your category and you are not in their content, that is a single, named outreach target. The same view also exposes which competitors are showing up in citations you should be winning.

Layer 3. AI Referral Traffic

Citations are not the goal. Sessions and conversions are. Connect your Google Analytics and watch which pages on your site receive AI traffic, which engine sent it, and what those visitors do.

Page-level attribution is where you find what to double down on.

If your “Best CRM for Healthcare” page is pulling four times more AI traffic than your “What is CRM” page, the message is clear. Write more comparison content. For a fuller framework on this, see content optimization.

The Agent That Runs This Loop on Autopilot

Everything above is the manual version of the keyword-and-prompt research loop. It is the version most tools stop at.

Analyze AI does not stop there. Inside the platform, you can build agents that run the entire loop on schedule, on event, or on demand. The Agent Builder is the part of Analyze AI that genuinely changes the math on content operations.

A few patterns teams already run on it.

The Monday brief generator. Every Monday at 7am, an agent pulls the prompts where your visibility dropped, the competitor topics you are not covering, and the keyword opportunities your domain can rank for. It writes a brief and drops it in Notion. You walk in to a fresh editorial calendar.

The publish-gate agent. When a brief gets approved, an agent generates research, builds an outline, drafts the article with your brand voice rules injected, runs it through an AEO content scorecard, and only publishes if the score clears 80. If it does not, it pings the writer with the gaps.

The opportunity finder. Once a week, an agent identifies prompts where competitors get cited and you do not, cross-references those with your keyword opportunities, and produces a ranked list of pages to write next.

These are running today on the same platform you use for visibility tracking. The substrate has 180+ nodes, 34 pre-built data recipes, and three trigger modes (manual, schedule, webhook). It turns “we should track this” into “this is already done.”

If you are an agency lead, this makes a 30-account portfolio actually feasible. If you are a CMO, it ships board-grade intelligence without an analyst. If you are a content director, it gates publish on quality automatically.

A Unified Strategy in One Workflow

Pulling it all together, the modern keyword strategy looks like this.

1. Build a candidate keyword list from seeds, free Google surfaces, and competitor gaps.

2. Layer in volume, difficulty, and click potential. Filter by relevance, realism, and revenue.

3. Map each priority keyword to one or more natural-language prompts. Track those prompts in Analyze AI.

4. Write content that covers the topic comprehensively, ranks on Google, and gets cited inside AI engines. Place keywords where they belong.

5. Measure across both channels. Search Console for traditional, Analyze AI for AI visibility, citations, and referral traffic.

6. Refresh quarterly. Move keywords up the priority list as your domain authority grows.

If you want to skip the manual version, our SEO content strategy playbook covers how the writing fits into the same loop.

Closing

SEO is not dead. Buyers still type queries. Search engines still rank pages. Keywords are still the unit of organic discovery. What has changed is that “search” now spans Google plus six AI engines, and the same content has to perform in both. Treat keywords as the foundation. Treat prompts as the second layer. Measure both. That is the playbook.

Ready to see where your brand actually appears across Google and AI search? Analyze AI tracks visibility, citations, and referral traffic across ChatGPT, Perplexity, Claude, Gemini, and Copilot, then runs the optimization loop on autopilot through the Agent Builder.