Skip to content

How to Find Keywords for SEO: A Working Process That Does Not Need Enterprise Tools

Keyword research fails two ways: guessing with no data, or drowning in a tool export nobody actions. The fix is a small process biased toward searches you can verify and win.

John Cravey with AIFounder5 min readUpdated Aug 6, 2026

Keyword research is deciding which searches deserve a page. Most guides make it a tool tutorial. This is the process instead: five steps, mostly free sources, biased hard toward searches you can verify exist and realistically win. It is the same process we run at the start of every engagement, and it ends in a list short enough to execute.

Free estimate · 2 minutes

Your Search Console data already shows the gap.

The estimate doubles as intake for an audit: we read your data, show where the clicks are being lost, and price the fix.

The plain-English version

A keyword is worth targeting when three things are true: the searcher might pay you, you can actually win the position, and the position is worth money. Every failure mode in keyword research is skipping one of the three: chasing volume with no intent, chasing head terms you will never hold, or ranking for things that never convert. The process below is just those three tests, run in order, against phrasings you gathered from real sources instead of guessed.

Step 1: harvest the phrasings buyers actually use

Start with language, not volume. Four sources, one afternoon:

  • Your own Search Console, first and always. The Performance report lists real queries you already appear for, with impressions and position attached. Queries at position 8 to 20 with impressions are your fastest wins; we walked the workflow in reading the GSC performance report.
  • Sales conversations and intake forms. The exact sentences prospects use are queries, near-verbatim. "Someone to fix our drainage" is a keyword; your service name usually is not.
  • Google autocomplete and People Also Ask. Type your service and your city, harvest every completion and question. These are demand-verified phrasings by definition.
  • Competitors' page titles. What the currently-ranking local sites target tells you what an engine already believes matches the demand.

Step 2: attach numbers, with appropriate distrust

Now quantify, knowing every number is an estimate. Google Keyword Planner gives volume ranges free (wide bands without an active campaign, but the relative sizes are what you need). Google Trends adds direction and seasonality: a term at modest volume growing 20 percent yearly beats a bigger one in decline, and seasonal terms need content shipped a quarter early. Scope everything to your geography; national volume is a vanity number for a local business. Where tools disagree, keep the disagreement visible rather than averaging it away. If you want one paid-tool workflow to compare against, Ahrefs' keyword research guide is the honest version of it.

Step 3: score winnability by looking at who ranks

The most skipped step, and the one that saves quarters of wasted work: search each candidate term and look at what currently wins. A results page of national directories and household brands means years of fighting; a page of thin local sites and forum threads means months. Note the map pack too, since for local terms the pack is the real first page. Grade each term honestly: winnable now, winnable in a year, or someone else's trophy. The systematic version of this read is the competitor analysis every SMB should run.

Step 4: group into pages by intent, not by spreadsheet row

Keywords do not get pages; intents do. "Water heater replacement cost," "how much to replace a water heater," and "water heater installation price" are one page. A commercial intent (someone ready to hire) gets a service page; an informational intent (someone diagnosing) gets an answer page that links to the service page. Map every surviving term to a page, and let the grouping tell you what the site needs: most small businesses discover they need 15 to 25 pages, not 200 blog posts. Moz's beginner guide covers the intent taxonomy if the concept is new to your team.

Step 5: prioritize by value, then execute in order

Final sort: expected value. Winnable terms with buying intent and real job value first; supporting question content second; long-fight head terms last or never. Put the list in a sheet with one row per page (target intent, phrasings, current position, status) and work it top to bottom. Re-run the whole loop quarterly: Search Console fills with new queries as you rank, which feeds step 1, which is how the process compounds instead of ending.

The searches to deliberately ignore

A good keyword list is defined as much by what it leaves out. Ignore pure-informational head terms with no plausible path to your service (the definition queries every publisher already owns), ignore other brands' names unless you sell against them honestly on a comparison page, and ignore any term whose results page is wall-to-wall national directories, because that is a rent you cannot outbid. Ignore vanity volume in general: the 10,000-search term that converts at nothing loses to the 40-search term where every searcher is holding a credit card. Write the ignore list down next to the target list; it is the half of the strategy that keeps the other half executable.

A worked example: a plumber runs the five steps

Concrete beats abstract, so here is the process on a two-truck plumbing company in a mid-size metro. Step 1 harvests about 60 phrasings: Search Console shows "water heater replacement [city]" sitting at position 11 with 400 monthly impressions, intake calls surface "tankless worth it" and "why is my water bill high," autocomplete adds "emergency plumber near me open now." Step 2 attaches ranges: the water heater family reads a few hundred searches a month locally, drain terms similar, tankless growing in Trends, repipe small but high-ticket. Step 3 checks the results pages: the emergency terms are a map-pack fight (winnable through the profile), the water heater terms show two thin local competitors (winnable in months), "plumber [city]" is held by directories (deprioritized). Step 4 groups the survivors into nine pages: five service pages, four answer pages. Step 5 sorts by value: water heater first (near-miss position plus real volume plus a four-figure ticket), emergency second, tankless third. Nine pages, a clear order, and the position-11 term alone likely pays for the whole quarter's work when it reaches the top five.

Where this feeds next

The list is an input, not an outcome. The terms become pages, the pages need the local and technical layers, and the loop needs measurement: that wider system is laid out in SEO for small business, and the data side, including how to verify demand with sources beyond keyword tools, is in data for SEO.

Want the demand map done for you? Run the estimator and we will pull the real search demand for your business and metro, scored for winnability, before any sales call. Or see how the process plays out for your industry.

Answers

Frequently asked questions

How do I find keywords for SEO for free?

Five sources cover most of what paid tools provide: your own Google Search Console (real queries you already appear for), Google autocomplete and People Also Ask (real phrasings), Google Keyword Planner (volume ranges), and Google Trends (direction and seasonality). Together they give you demand, phrasing, and trajectory without a subscription.

What makes a keyword worth targeting?

Three tests together: intent (would the searcher plausibly pay for what you sell), winnability (can a site like yours realistically hold a top spot, judged by who ranks today), and value (what a ranked position would be worth in revenue, not traffic). A term must pass all three; volume alone passes none of them.

How many keywords should a small business target?

A focused set of 20 to 40 terms, grouped into pages by shared intent, beats a spreadsheet of 500 every time. Each page targets one intent cluster: the head phrasing plus its variants. When those are won, expand outward. Coverage you cannot execute is not strategy.
The unit shifted from keyword to question, but the discipline is the same: find the phrasings real buyers use and answer them better than anyone. AI answers get assembled from pages that answer specific questions precisely, so question-shaped research (People Also Ask, your own sales calls) matters more, not less.

What are the two ways keyword research usually fails?

Guessing with no data, and drowning in a tool export nobody actions. The first produces pages targeting phrases nobody searches. The second produces a spreadsheet of ten thousand rows that never becomes a page. The fix is a small process biased toward searches you can verify and realistically win.

Where do I harvest the phrasings buyers actually use?

From the places they already speak: sales calls, intake forms, the questions you answer on every first call, review text, and the People Also Ask box. Those are the words customers use, which is not the same as the words the industry uses, and the gap between the two is where most keyword lists go wrong.

How much should I trust keyword volume numbers?

Directionally, not precisely. Volume figures are estimates, they bucket aggressively, and they miss the long tail that actually pays a small business. Use them to compare the relative size of options, never as a promise of traffic, and weight verified sources like your own Search Console data far higher.

How do I judge whether a keyword is winnable?

By looking at who currently ranks, not at a difficulty score. Open the results page. If it is directories, national brands, and pages with far more depth than you can produce, that is a fight to skip. If it is thin pages, generic content, or nobody answering the question directly, that is an opening.

How do I turn a keyword list into pages?

Group by intent, not by spreadsheet row. Every phrasing that wants the same answer belongs on one page, however differently people type it. That collapses a thousand-row export into a manageable set of pages, and it prevents the classic mistake of building near-duplicate pages that compete with each other.

Which searches should I deliberately ignore?

Head terms owned by directories or national brands. Informational queries an AI answer now satisfies completely. Phrases with volume but no buying intent. And anything you cannot serve, since ranking for work you do not do produces enquiries that waste both sides' time.

What does the five-step process look like end to end?

Harvest the phrasings buyers actually use. Attach volume numbers with appropriate distrust. Score winnability by looking at who currently ranks rather than at a difficulty score. Group the phrasings into pages by intent. Then prioritize by value and execute in order. Five steps, no enterprise tool required at any of them.

Why prioritize by value rather than by volume?

Because volume and revenue are different axes and only one pays. A phrase with a tenth of the searches and direct buying intent is worth more than a popular informational term, and a search you can actually win beats a bigger one you cannot. Value ranking is search volume filtered through intent and winnability.

Question we did not answer? Ask us directly and we will answer it here.

Written by
John Cravey
Founder

Founder of Frontend Horizon. Writes most of the long-form work on the FH blog.

Newer post
SEO for Financial Services: Ranking in a YMYL Category Without Fighting Compliance
Older post
Data for SEO: The Numbers That Should Drive Your Strategy and Where to Get Them Free
Keep reading

More from the blog

SEO·5 min

How to Do an SEO Audit: The Checklist We Run on Every New Site

A tool can score your site in ninety seconds. A diagnosis tells you which three of the two hundred flagged items actually move revenue.

SEO·9 min

Answer Engine Optimization: Getting Cited by AI, Not Just Ranked

Ranking first means nothing if the AI answers the question above your link and the user never scrolls. The new goal is being the source it quotes.

Search Console·10 min

Turn Google Search Console Into an AI Opportunity Finder

You are already ranking on page two for things you never wrote about. An AI agent's job is to find them before you waste effort elsewhere.