Keyword research for blog posts
Keyword research for blog posts decides which page each search belongs to: the right page for the wrong search brings no one, and two pages for one search beat each other. Give a topic its own page only when the data proves people search for it, then study the results to find the gap your post fills.
The research also decides whether a post is worth quoting. It must hold facts the reader can't get from the next ten results, and every outside fact carries its source.
When a topic gets its own page
- A topic gets its own page only when a keyword tool and the real results both show demand and competing pages. Search the exact phrase. If nothing comes back, step up one level. Then try another angle, for example by adding "price".
- The splitting rule: if a non-expert would not see two variants as the same thing, they get separate pages. True synonyms share one page.
- The homepage owns the domain's main term. No other page targets it.
- Use the exact keyphrase at least 3 times on the page, in its base form. The keyphrase is what people type, so don't bolt extra words onto it by default.
- Place the subject's important keywords in the title, meta, headings and body where they read naturally. A keyword that can't be placed naturally is left out. Never stuff; chapter 12 runs an over-optimisation check.
- Never let two pages compete for one search. Give the search to one page, and point the other page at its own question.
- The "competitors miss" list is the content angle: cover what the top pages leave out.
Tools and demand checks
- Use: Google Keyword Planner set to the target country, Google autocomplete walked through the alphabet, Search Console once the site is live, a paid competitor tool if your budget allows it, and Google Trends for seasonality.
- Keywords come from search data, never from guesses: a keyword tool for volume and difficulty, the live results for intent.
- Audit the top 3 to 5 results: domain strength, heading quality, indexed pages and depth. Stacked weaknesses mean you can win without backlinks by covering the topic more deeply. Some brand-search giants can't be beaten; pick another phrase.
- Build keyword clusters for your business. For a plumber: emergency, frost, water heater, heating, floor heating and pipework. For a running shoe store: trail, road, wide fit, stability, racing and care.
For a local service business: don't append the city name to the keyphrase by default. Keyword tools are unreliable for hyper-local terms, so trust the live results and the map pack, and Search Console once the page is live. Cross each keyword cluster with every place you really serve.
Intent
Intent is what the searcher wants to do: learn, compare or buy.
- Buyer intent shows in modifiers: "price", "cost", "best", "vs", "alternative", "near me", "subsidy", "installation", "per m² price" and a place name.
- Prioritise buying-intent questions (price, availability, "near me", "best for") over broad informational ones.
- People now ask whole questions ("how much does a boiler replacement cost in Leeds?"). Build pages on those questions.
- Match copy to what the reader already knows. Problem-aware readers get the problem named in their own words. Solution-aware readers get the difference between you and the alternative. Jargon headlines quietly shut out the first group.
- Research keywords per language; never translate them. Include the spellings without diacritics that people actually type.
The searches a post may take depend on what the other pages already own.
- For an online store: "buy", "shop", "price" and product-name searches belong to collection and product pages. Posts take the informational and comparison searches ("how to choose", "[A] vs [B]", "best [product] for [use]", care and how-to) and hand the reader to the collection (chapter 14).
- For a software or B2B company: "[product] pricing" and "[product] login" belong to the product site. Posts take "[problem] how to fix", "[tool A] vs [tool B]", "[category] for [industry]" and "[tool] alternatives".
Customer questions and phrases
- Collect the questions customers really ask from Google searches, People Also Ask, assistant answers and the enquiries the business already gets, then match each question to a page.
- Keep a customer phrase bank: the words heard on calls, in reviews, in support tickets and in searches, in every language you sell in. Headings borrow from it first.
- Write a SERP note per post (SERP means the search engine results page): the queries in each language, the top-ranking URLs, the intent, the keyword per language, and a "competitors miss" list of the gaps your post covers. The SERP analysis below produces it.
The SERP analysis for every article
Run it on the top 10 Google results for the target keyword, in a real browser, in the target country and language. Think it through in the research notes; none of it appears on the page.
- Search intent: what exactly the searcher is trying to accomplish, and whether the ranking pages satisfy it.
- Competitor weaknesses: for each ranking page, the missing information, outdated content, shallow explanations and unanswered questions.
- Content gaps: 5 to 10 specific things a new article can provide that the current results don't cover well.
- SERP opportunities: weaker results, forums, Reddit threads and pages that don't fully answer the query, and what each reveals about the opening.
- Unique angles: 3 differentiated angles, each with a realistic reason a searcher would prefer it over the current results.
- Original value: the first-hand research, photos, measurements, comparisons or tools that would make the article genuinely more useful.
- Ranking feasibility: a score out of 20, with 4 points each for search intent, competition, content gaps, original advantages and relevance to the site.
- Article blueprint: the title, the structure, the key questions to answer and the evidence needed before publishing.
- Final verdict: pursue the keyword, target a narrower variation, or skip it.
- Real results only. Use the search results and page content you can actually open. Never invent traffic estimates, rankings, competitor weaknesses or keyword data, and name anything you couldn't verify.
- The goal is a genuine gap. Another 2,000-word article that says what the top 10 already say is worthless. Find the gap in the results and build something worth ranking.
- Research a distinct strategy for each article that earns traffic, citations and backlinks (a chart others will cite, a data point journalists need, a small tool), and put it in the blueprint.

Sources and evidence
- Numbers come from three places only: your own records, your Search Console or analytics, or a named, dated public source. Never invent a market, competitor, legal or search figure. Your own figures come from your own records (chapter 6).
- Every fact traces to a source in the research notes. Sources live in the notes; the page carries the link where a reader needs it.
- Keep what you know apart from what you're guessing: a fact from your records, a figure from a source, and a hunch you still have to test.
- Cite laws, standards and regulations by section and date, and find the local equivalent of every rule or guideline you follow.
- For any claim about how a search engine or AI feature works, use the primary source (Google's Search Central documentation on AI features and Search Essentials), and check the current policy at the time you implement.
- Read the real sources: the law, the standard, the manufacturer's manual, the official guidance, before any marketing blog that summarises them.
- Real conversations in your niche (forums, Reddit) inform wording and needs. Never copy the comments.
- Claims must fit the business. A business that isn't open day and night never says "24/7".
Prices, measurements and process
- Your prices come from your own price list, stated once, exactly, the same way on every page. A post about a job you do quotes what you charge.
- A market price belongs only in a post about market prices ("what a boiler replacement costs in Leeds"), and it is labelled as a market average with its sources. To build one, take the published prices of 10 competitors for the same item:
- Normalise the unit and the tax basis first.
- Compare "from" prices only with "from" prices.
- Drop outliers above 3× or below a third of the median.
- With fewer than 3 comparable prices, don't publish an average.
- Round to one committed figure (table below).
- Measurements and durability figures come from the manufacturer's data or your own measurements. Where you quote what others state, say whose figure it is.
- Process steps use the words your customers and the trade use, and each step is a concrete action with its figure.
- Every post whose topic is a job or a purchase carries a cost breakdown and a cost per unit (per m², per running metre, per hour, per item, per seat, per month).
| Value | Round to |
|---|---|
| Prices from 100 | nearest 10 |
| Prices from 20 to 100 | nearest 1 |
| Prices under 20 | nearest 0.10 |
| Per-unit rates under 10 | nearest 0.10 |
| Per-unit rates from 10 | nearest 1 |
| Per-kilometre rates | nearest 0.01 |
| Guarantees | whole years |
The source of a price changes with the business.
- For a local service business: every service topic also carries the response time and a "how long does it take" section with the real time breakdown for that job.
- For an online store: product prices are the store's own catalogue prices. A market average serves only a comparison post ("what a good [product] costs"), and buyer questions come from product reviews, support tickets and return reasons (chapter 14).
- For a software or B2B company: competitor prices come from their public pricing pages, normalised to the same billing period and seat count, with the date you checked them.
Entities and local conditions
Entities are the named things a topic involves: places, materials, standards, organisations, products and people. Search engines and assistants use them to understand what a page is about.
- Name them exactly: the material, the standard and its number, the product model, the district.
- Use real distinctions, never filler. A distinction is real when it changes the answer for the reader.
For a local service business: describe your own area from what you know and can source. Heating differs between panel blocks on district heating and private houses with boilers or heat pumps; old wooden districts need different roofing from concrete estates; hard water wears a boiler faster. Open a place page with the conditions there that change the job, in the first paragraph. A roofer whose town page explains that town's houses, weather and roofing methods has a page no competitor can copy. Never use travel-guide language on local pages ("nestled", "in the heart of", "vibrant", "boasts").
One canonical entity
- Keep one canonical entity: one company name, registration code, address, phone, email and offer list across the site, the Google Business Profile, national directories and social profiles, in every language.
- Record a focus keyphrase for every page, internally.
- Write every SEO field from scratch for each page, never pulled from an old plugin or an unfinished site. The fields per page: title tag, meta description, H1, slug, image alts and structured data.
For a local service business: use an email address on your own domain in your name, address and phone details (NAP), and keep those details identical, character for character, everywhere they appear.

Research hygiene
- Check a site in its own language. Browser translation creates quirks that don't exist in the original.
- Back every recommendation with real data and its source. No number without a source.
Edited by Andres LepikCo-founder & Head of Growth
Fact-checked by Siim SaarHead of SEO
Post by Rasmus Kivi
Co-founder & CEO
Co-founded REND and still reads every client's monthly report himself before it goes out.