Choosing keywords for blog posts comes down to three filters: the query has real search demand, the sites already ranking for it are ones you can plausibly beat, and the searcher wants an article rather than a product page. Miss any of the three and the post will not earn traffic no matter how well it is written.
One primary keyword per post. Five to fifteen secondary terms that occur naturally because you covered the topic properly. That is the whole model.
Start with intent, not volume
The fastest way to waste a post is to target a keyword where Google has already decided articles are not the answer. Search your candidate term and look at what actually ranks. If page one is ten product pages, a blog post will not break in. If it is nine articles and a forum thread, you have an informational query and your format matches.
Four intent types are worth naming. Informational queries (“how does composting work”) want explanation. Commercial investigation (“best compost bin”) wants comparison. Transactional (“buy compost bin”) wants a product page. Navigational (“subpod login”) wants a specific site. Blog posts win the first two and lose the last two.
Commercial investigation terms are the most valuable blog keywords in most niches. They convert far better than pure how-to content and still reward article format. Prioritize them.
Find candidates from four sources
- Seed expansion — put three to five broad terms from your niche into a keyword tool and pull the long-tail variants. SEO Rocket returns up to 150 ideas per search with volume, difficulty, CPC, and SERP features, and you can export the filtered list to CSV.
- Search Console — the highest-value source you already own. Filter queries with impressions above 100 and average position between 8 and 25. You are already visible; you just are not winning yet.
- Competitor gaps — pull the keywords three or four rivals rank for and you do not. A content gap report with per-rival position columns tells you which gaps are contested and which are open.
- Customer language — support tickets, sales calls, and community threads produce phrasings no tool suggests, and those phrasings often have zero competition.
Judge difficulty against the weakest page-one result
Difficulty scores are modeled estimates, not facts. A 34 in one tool may be a 51 in another. Use them to sort a list, never to make the final call.
The real test takes two minutes. Open page one for the keyword and look at the weakest result — usually somewhere between position seven and ten. Check its referring domain count, its word count, and how directly it answers the question. If that page is a 700-word post on a site with 40 referring domains and you can write something better, the keyword is winnable regardless of what the difficulty number says.
Benchmarking against the median or the top result is how people talk themselves out of every good opportunity. You are not trying to beat the strongest page on day one. You are trying to be better than the weakest.
Volume expectations, honestly
Third-party volume figures are modeled roughly twelve-month averages built from clickstream data and periodic crawls. They are directionally useful and frequently off by a factor of two on individual terms, especially below 100 searches a month.
That does not make them useless. It means you should treat 40 and 60 as the same number, treat 40 and 4,000 as genuinely different, and check your own Search Console data once the post is live — Google’s data about your site always wins over any estimate.
Low volume is not a disqualifier either. Twenty posts targeting 50-search terms with high commercial intent will out-earn one post chasing a 20,000-search head term you cannot rank for. Long-tail keywords also tend to convert better because the searcher has already narrowed what they want.
One primary keyword, many natural variants
Targeting three unrelated primary keywords in one post produces a page that ranks for none of them. Pick one. Then cover the topic thoroughly enough that the secondary terms appear because they had to.
If you are writing about container gardening for beginners, terms like potting mix, drainage holes, balcony sunlight, and watering frequency will appear on their own. If they do not, your coverage has holes. That is the real signal to watch — missing related terms usually means missing sections, not missing keyword placements.
Do watch for cannibalization. If you have three posts all aimed at slight variations of the same query, Google will pick one and suppress the others. Merge them into one stronger page and redirect the losers.
Where to place the keyword
- Title tag — near the front, under 60 characters so it does not truncate in the SERP.
- H1 — once, phrased for humans. It can differ from the title tag.
- First 100 words — once, in a sentence that answers the query directly.
- URL slug — short, hyphenated, no stop words. Set it before you publish; changing it later costs you.
- One or two H2s — where it fits the section honestly.
- Meta description — 140–155 characters. It is not a ranking factor, but it drives click-through, which matters.
- Body — two to four total mentions across the article is plenty. Density targets are a relic.
Also worth doing: answer the core question in the first two sentences of the post. That helps human readers, and it is the passage most likely to be pulled into a featured snippet or an AI Overview.
Build a keyword-to-post map before you write
Keep a simple sheet: keyword, intent, estimated volume, weakest page-one competitor’s referring domains, target URL, status. This prevents two writers from covering the same term and makes it obvious when a topic needs a new post versus an update to an existing one.
Group related keywords into clusters of one pillar page and five to ten supporting posts. The pillar targets the broad term; each supporting post takes one specific sub-question. Clusters rank better than scattered posts because internal links between them concentrate relevance.
Saving your chosen terms into a project keyword pool — as SEO Rocket does — means the same list feeds the writer and the rank tracker, so you stop retyping keywords into three different tools.
Measure the right thing after publishing
Give a post at least eight to twelve weeks before judging it. New pages typically bounce around while Google works out where they belong, and daily movement of two or three positions is normal noise, not a trend.
Watch impressions first. Rising impressions with a flat position means Google is testing you on more queries — a good sign. Flat impressions after three months means the page is not being considered at all, which usually points at intent mismatch or insufficient depth rather than keyword placement.
Then check which queries the page actually earned in Search Console. You will often find it ranking for something adjacent to your target that converts better. Rewrite toward that term rather than forcing the original.
Do that consistently and keyword selection stops being guesswork. If you want the research, the gap analysis, the writing gates, and the tracking in one workspace, SEO Rocket runs all of it on industry-grade data at a flat $50 a month — but the method above works with any tool that reports honest numbers.
