How to Find Best Keywords for SEO, Step by Step

Most keyword research advice collapses the minute you try to ship something real. We know because we’ve shipped the wrong pages for the “right” keywords, stared at flat traffic lines for months, and then found out the issue wasn’t the tool. It was us choosing the wrong target.

If you’re here to learn how to find best keywords for seo, the uncomfortable truth is that the “best” keyword is rarely the one with the biggest number in a tool. It’s the one that fits your audience’s intent, your site’s actual ability to rank, and the page you can realistically produce without making Google (or readers) roll their eyes.

What “best” means before you open a tool

A keyword is only “best” in relation to a goal: traffic for awareness, leads for sales, revenue for ecommerce, or brand visibility for category ownership. Pick one primary goal per page, not per site. That’s the first filter.

People treat high volume as the goal, not an input. That’s how you end up chasing keywords that can’t pay back the effort.

How to find best keywords for SEO: start with intent triage (or waste weeks)

Intent is where keyword research either compounds or dies. You can pick a phrase that’s topically perfect and still lose because Google is trying to satisfy a different kind of searcher than the page you’re planning to publish.

We learned this the dumb way. One of our earlier content pushes targeted a “best X” query with a pure educational guide because the topic felt educational. The SERP disagreed. Page one was full of listicles with comparisons, affiliate-style layouts, and clear “which should I buy?” framing. Our guide wasn’t bad. It was just the wrong shape. We sat on page two, then page three, then nowhere.

The SERP-pattern rubric we actually use

Forget the labels (informational, commercial investigation, transactional) for a second and look at what Google is already rewarding. We run a quick SERP tally and score it.

Open an incognito window, set the location if it matters (more on that later), search the keyword, then scan the top 10 results. Count how many are each page type.

Here’s our practical rubric:

• If 6 or more of the top 10 are long-form guides, definitions, or “how to” posts, treat it as informational and create a guide. People Also Ask boxes and featured snippets often show up here.
• If 6 or more are “best,” “top,” “vs,” “review,” or comparison pages, treat it as commercial investigation. Expect list-format structure, clear decision criteria, and internal comparisons. Video carousels are common.
• If 6 or more are category pages, product pages, or “pricing” pages, it’s transactional. Your blog post will struggle no matter how well written.
• If a local pack dominates above the fold (maps, “near me,” service listings), it’s local intent. You need location signals and usually a different kind of page.
• If the SERP is mixed (no type hits 6), your job is to find the pattern in the top 3, not the top 10. Google is telling you what it trusts most.

That rubric sounds simple. It isn’t always.

What trips people up is the “mixed” SERP: you’ll see a couple of guides, a couple of product pages, maybe a forum thread, and you’ll convince yourself your page can be “a bit of everything.” That usually produces a page that’s mediocre at all intents.

The fastest decision rule: new page type vs updating an existing one

We use a blunt rule to avoid endless debating:

If the intent matches an existing URL on our site, we update that URL. If the intent requires a different page format, we create a new page.

Examples:

If we already have a decent guide ranking for “backyard chicken breeds” and we discover “best backyard chicken breeds” has different SERP patterns (more listicles, more decision framing), we do not cram both intents into one page. We either add a dedicated “best” section that matches the SERP, or we publish a separate “best” page and cross-link carefully.

The annoying part is cannibalization: two pages targeting the same intent, splitting clicks and confusing Google. We’ve caused this by publishing a new post when the old one was already in striking distance, then watching both pages float around positions 11 to 20 like they’re allergic to page one.

Intent mismatch symptoms you can spot in 2 minutes

When we’re unsure, we check for these red flags:

If the query contains “best,” “top,” “vs,” “review,” “software,” “tool,” or “pricing,” and the SERP is comparison-heavy, a pure educational explainer is usually dead on arrival.

If the query is phrased as a question and the SERP shows People Also Ask plus short snippet-friendly results, we make sure our page answers early and cleanly. Not 900 words later.

If the SERP is filled with videos (especially YouTube), we assume the audience wants demonstration, not just text, and we decide whether we can meet that expectation.

Seed keyword engineering that doesn’t start with guessing

Most guides tell you to “brainstorm seed keywords.” That’s how you end up with industry jargon, one-word head terms, and suggestions that are irrelevant or impossible to rank for.

We build seeds from customer language, not our vocabulary. That means:

First, we pull exact phrasing from support tickets, chat logs, sales call notes, onboarding emails, and community threads. If you don’t have those, use reviews of your product or competitors. People tell on themselves in reviews: they describe what they tried, what failed, and what they wish existed.

Then we translate that phrasing into query formats people actually type. Customers speak in stories. Searchers speak in fragments.

A customer might say: “We want chickens that lay eggs even in winter and don’t freak out our neighbors.” A searchable seed becomes: “quiet chicken breeds for eggs” or “cold hardy laying hens.”

Where this falls apart is when you start with internal jargon like “omnichannel attribution” or “zero trust posture” and assume the market uses the same words. Sometimes they do. Often they don’t.

One throwaway moment: we once built a whole keyword list around what our product team called a feature. It turned out the only people using that term were us and two competitors. Anyway, back to the point.

Expand keyword ideas without tools-first tunnel vision

Tools are great at scaling, not at originality. We still start in the SERP because it shows real demand and real phrasing.

First, we type the seed into Google and record:

Autocomplete suggestions (these are not random, they’re behavior-driven).

People Also Ask questions (these are gold for subtopics and long-tail).

Related searches at the bottom (often the easiest cluster starters).

Then we leave Google. Your audience asks questions where they feel safe sounding dumb. That’s usually not on a polished landing page.

We mine:

YouTube autosuggest for “how to” phrasing and demo-driven intent.

Reddit threads and niche forums for problem language, constraints, and weird edge cases.

Instagram, Facebook groups, and comment sections for informal wording and objections.

The strategic reason: post-RankBrain, topical coverage matters. You can rank for queries you don’t repeat word-for-word if you genuinely cover the topic and match intent. RankBrain was announced in 2015 and framed as a major ranking signal. The exact weighting is debated, but the practical takeaway for us has been consistent: keyword lists are scaffolding, not a script.

The catch is that expansion creates chaos fast. You can collect 300 phrases in an afternoon and still have no plan because nothing is labeled.

Evaluation and prioritization: turning a messy list into a short list that can rank

This is the part everyone hand-waves: “check volume and difficulty.” Cool. Now pick between a keyword with 10K searches and brutal competition, and one with 200 searches and a clear ranking path.

We use four inputs as non-negotiable:

Search volume (demand).

Competition or difficulty (how hard it is to win).

Relevancy (does it belong on our site and on this page).

Intent fit (does the SERP want the page we can make).

Volume and competition are easy to see. Relevancy and intent are where teams lie to themselves.

Why we don’t worship high volume

High volume often means: broad meaning, more SERP features, bigger brands, and more link-heavy winners. Head terms like “chicken” are the extreme example: ambiguous, crowded, and usually pointless unless you already own the category.

Narrower terms can convert better because the searcher is self-qualifying.

A useful case pattern looks like this:

“best backyard chicken breeds” can show 100 to 1K monthly searches with low competition in tools, which might be worth targeting.

“backyard chicken breeds” can show higher volume but also higher competition, making it harder to rank for.

That doesn’t mean you ignore the broader term. It means you earn it by owning the long-tail first.

A simple scoring model we use (and the tie-breakers that matter)

We don’t pretend a spreadsheet knows the future. We use it to force clarity.

We score each keyword 1 to 5 on:

Intent fit (double-weighted). If the SERP wants a different page type, it’s a 1. No debate.

Relevancy. If it’s only loosely related, it’s a 2 or lower.

Achievability. We estimate based on competition metrics plus SERP reality: are the top results massive brands, is the content unusually strong, are backlinks clearly doing the heavy lifting.

Volume. We treat volume as a tiebreaker, not the driver.

Then we compute a rough score: (Intent x2) + Relevancy + Achievability + Volume. Maximum is 20.

Tie-breakers we use when scores are close:

If one keyword can be answered with a clean, better page than what’s ranking, we pick it even if volume is lower. Quality gaps are rare. When you see one, take it.

If a keyword maps cleanly to an existing page with some traction (Search Console shows impressions), we pick it. Updating beats starting from zero.

If competition is similar, we choose the variant with clearer wording and fewer meanings. Ambiguity kills.

Geo targeting: the underrated way to find easier wins

Google Keyword Planner lets you set a target area, and it changes what you see for volume and competition. That’s not a minor feature, it’s often the difference between “impossible” and “plausible.”

If you operate in a specific state or region, setting location to something like “Georgia, United States” can surface terms with lower competition and more realistic volume for your actual market.

A lot of teams do national keyword research for a local business and then wonder why they can’t rank. They picked a fight in the wrong arena.

Don’t trust difficulty scores without eyeballing the SERP

Tool metrics are directional. They’re not truth.

We’ve seen “low difficulty” keywords where the top 5 results are government sites and entrenched brands, plus a featured snippet and a video carousel that pushes organic results down. The tool didn’t lie, it just didn’t see what we needed it to see.

We always sanity-check:

Do the top pages have obvious link profiles (lots of referring domains) that we can’t match soon.

Is the content unusually deep or updated frequently.

Is the SERP dominated by formats we cannot produce (video-first, local pack, product grid).

Tool workflows that match real constraints

You can do this with expensive suites. You can also do a lot with free or cheap tools if you’re disciplined about the process.

Google Keyword Planner: what it’s good for, and what it isn’t

Keyword Planner gives you search volume ranges, trends, and competition data, and you can set the geographic target. We use it for directional validation, not as the final judge.

Our workflow is basic:

We enter a seed keyword idea. We set the target area to match the business reality. We review volume and trends, then we scan suggested variants for phrasing we missed.

What nobody mentions is that Keyword Planner is built for ads, not SEO. The competition metric is paid competition, not organic difficulty. It correlates sometimes. Sometimes it doesn’t.

Complementary tools and why we keep them around

If we need richer SEO metrics like keyword difficulty, intent labels, competitor backlinks, and ranking context, we pull from an SEO suite.

SE Ranking is one example that claims very large keyword databases (billions of keywords across Europe, North America, and Asia) with coverage across many geo databases, plus a 14-day trial without a credit card. We care about claims like that only insofar as they let us check a lot of variants by location without spending half a week.

Ubersuggest can be handy for quick checks, but the free plan limit (3 searches per day) is a real bottleneck. Plan for that or you’ll stall mid-research and start making “gut” decisions again.

Our honest recommendation: pick one primary tool for scale, then keep Keyword Planner for geo sanity checks and trend context. Don’t run five tools and pretend that’s a strategy.

Reverse engineering what already works: Search Console + competitor SERPs

If you already have a site with any history, the fastest wins come from queries you’re already showing for. Google Search Console is brutally underused.

We pull the last 3 months of queries and look for two buckets:

Striking distance: queries where the average position is roughly 8 to 20 and impressions are meaningful. These are pages that often need better intent match, stronger sections, or clearer on-page answers.

Near-miss variants: queries that are basically the same topic but different wording. These can become new sections on the same page, not new posts.

The main failure mode here is targeting the new keyword on the wrong URL. If a page is already ranking for the topic, publishing a new one can split signals and make both worse.

Competitor SERP review is the other half. We search the keyword and open the top results in separate tabs. We’re not looking to copy. We’re looking for patterns:

What subtopics appear across multiple winners.

What format wins (listicle, guide, template, tool).

What the page answers early vs buries.

If you see the same subheading on five of the top ten pages, that’s not originality. That’s a minimum requirement.

Using keywords without getting penalized: mapping clusters to pages

Old SEO advice from the 2000s taught people to stuff exact-match phrases everywhere. That era died for a reason.

Panda punished thin, low-quality, and duplicate content patterns. Penguin went after unnatural keyword usage and link behavior. The names are old, but the risks are still current: low-effort pages and manipulative patterns don’t age well.

We map a keyword cluster to a single page when the intent is the same. That page gets:

A primary query that defines the page’s promise.

A set of close variants that become headings or sections when they represent real sub-questions.

Related questions (often from People Also Ask) that become short, direct answers.

Then we write for coverage, not repetition. If the topic is actually covered, RankBrain-style systems can connect you to queries you never wrote verbatim.

The easiest mistake to spot: repeating the exact phrase in every H2. It reads like a robot and it’s unnecessary. Write like you’re helping someone, then check afterward that the language of the SERP is present where it belongs.

Validation: what we watch after publishing (and what we change)

We give a page enough time to collect signals. Not forever, but long enough to avoid thrashing.

We watch:

If impressions are rising but clicks are flat, we rewrite titles and snippets, and make sure the page answers the query faster.

If the page sits around positions 11 to 20, we usually have an intent mismatch or an authority gap. We adjust format first, then strengthen the content, then look at internal linking.

If the page gets clicks but the on-page behavior is bad (short time, bounces, no next step), the keyword might be fine but the page isn’t meeting the promise.

Expecting immediate results makes people do random edits that break what was working. We’ve done it. It’s painful.

Keyword research is not a one-time spreadsheet. It’s a loop: find, vet, publish, measure, revise, expand. If you keep the intent triage and prioritization discipline, you stop gambling and start building momentum.

FAQ

How do you find the best SEO keywords?

Start by checking the SERP to confirm intent, then validate demand and variants in Google Keyword Planner. Use Keyword Planner for direction and trends, not as the final judge of ranking difficulty.

What matters more: keyword volume or search intent?

Search intent. If your page type does not match what Google is rewarding on page one, high volume just means you fail faster.

How do you know if you should update an existing page or create a new one?

Update the existing URL when the intent matches and the page already has impressions or rankings. Create a new page when the SERP clearly favors a different format, like a “best” list versus a how-to guide.

What is the 80/20 rule in SEO keyword research?

It’s focusing on the small set of keywords and pages most likely to drive results. In practice, prioritize the terms with clear intent fit, high relevancy, and a realistic ranking path instead of chasing the biggest volumes.