How to do keyword research for a blog (step by step, from 0 to calendar)

Don't lose focus with the noise of super pro tools, you have the best tool you can do keyword research on your shoulders

‍

Define the purpose of the blog and the type of reader (without this, the tool doesn't matter)

Before opening any tool, define the frame in 6—10 lines. Example for this case:

• What do you sell: Makeit Tool (SEMrush type suite) for SEO research, monitoring and prioritization.
  
  
• To whom: SEOs from niche to manager (includes consultants/agency).
  
  
• What actions do you expect: subscription to content, branded search, and for the reader to consider the tool as an option when they need to solve SEO tasks.
  
  
• What verticals will you cover: technical SEO, content, AI and visibility (AIO/LLMS), analytics (GA4/GSC), link building, reporting and management.
  
  
• What are you not going to do: generic posts “for everyone” without a clear job-to-be-done.
  
  

If this is not clear, research tends to generate “nice” but useless traffic: a lot of TOFU reading, little connection with the product and low brand building.

“Keyword value” ≠ volume: how to measure business value in an SEO blog

To prioritize by ROI, use value criteria that a blog can capture:

• Closeness to buying (decision): “alternatives to...”, “best tool for...”, “X vs Y”, “price”, “review”, “comparison”.
  
  
• Measurable how-to: tasks that the reader needs to execute and measure (GA4, logs, GSC, technical auditing, tracking SERP).
  
  
• Reference/citation potential: canonical pieces (glossary, definitive guides, checklists) that are linked and reused.
  
  

A practical trick: score each word/cluster from 1 to 5 for “business value” before looking at volume. If it doesn't get at least a 3, it usually doesn't deserve priority unless it's a pillar piece.

Step 1 Create your list of “seed topics” with product logic

In 30 minutes you can get 20—50 seeds if you build them from real “modules” and “pains”:

• Recurring tasks: technical auditing, keyword research, content briefs, ranking tracking, competitor analysis, reporting, internal linking, cannibalization, migrations.
  
  
• Dolores by profile: niche (quick wins, low-hanging fruit), manager (prioritization, governance, reporting), agency (repeatable processes).
  
  
• Comparisons and alternatives: by category (tools, methodologies, frameworks).
  
  
• Trends affecting work: AI Overviews, traffic measurement from LLMs, SERP changes.
  
  

The goal is not “interesting topics”, but rather a thematic universe that is already aligned with your business and with what your reader is really trying to solve.

Quick method: “Jobs-to-be-Done” from the SEO reader (niche vs manager)

Convert jobs to seeds. Examples:

1. “I want to find easy keywords to start a blog.”
   
   
2. “I want to prioritize an SEO backlog with limited resources.”
   
   
3. “I want to know if AI Overviews is lowering my CTR.”
   
   
4. “I want to measure traffic from ChatGPT in GA4.”
   
   
5. “I want to group keywords by intent and cluster.”
   
   
6. “I want to avoid cannibalization and define a canonical one.”
   
   
7. “I want to explain results to management with a dashboard.”
   
   
8. “I want to analyze competitors and detect gaps.”
   
   
9. “I want to decide what to update before I publish more.”
   
   
10. “I want to design 'citable' and useful (not generic) content.”
    
    

Each job gives you 2—5 seeds. If you can't write the job in a clear sentence, that's a bad seed.

Step 2 Expand seeds to real keywords (3 sources that don't fail)

Here you are looking for breadth without inventing terms. It combines three sources with different biases:

Fountain

What does it provide

Biase/limitation

Keyword Planner

Ideas + demand ranges from Google Ads

It's an Ads tool: estimates and groupings, not “exact truth” SEO

SERP mining

Real language (Autocomplete, PAA, related) and sub-questions

Requires manual work; varies by country/language/device

Search Console (if there is already traffic)

Queries where you already have impressions/closeness

It depends on your history; there are noise and changes due to seasonality

Google Keyword Planner: how to use it for ideation (even if you're SEO, not PPC)

Keyword Planner is useful for ideation and for understanding demand “families”. Basic flow:

1. In Google Ads, open Keyword Planner and use Discover new keywords.
   
   
2. Enter your seeds (and/or a relevant URL) and ask for ideas.
   
   
3. Filter by country/language and export.
   
   
4. Group by topic (even if the tool groups, you re-cluster later).
   
   

Correct reading: take it as estimation and as a generator of variations, not as a number that dictates decisions on its own.

SERP mining: Autocomplete, “People also ask” and related tips

For each seed, extract 10—30 variations and sub-questions. Rule of thumb:

• Save “intent” modifiers: nicest, vs., alternatives, priced, template, Checklist, instance, how, why, guides.
  
  
• Capture PAAs as potential sub-themes (these are natural “fan-out” tracks).
  
  
• Create a “dictionary” of modifiers per vertical (not all of them apply the same).
  
  

Minimum output: a sheet with columns seed → keyword → modifier → note of intent.

Search Console (if you already have traffic): queries by URL and gaps by cluster

If the blog already exists, here are usually the quick wins:

1. In GSC, open Performance and look at pages that are already receiving impressions.
   
   
2. For every top URL, it exports queries.
   
   
3. Group by intent and detect:
   
   
   
   • Long-tail that you already “rub” (high impressions, low CTR).
     
     
   • Cannibalization (several URLs for the same subintent).
     
     
   • Recurrent subthemes that you are missing as a dedicated piece.
     
     

This often provides faster opportunities than “starting from scratch” with external tools.

Step 3 Classify intent and type of content (what decides the format of the post)

The intention rules the format. If the top 10 is comparative and you publish a definition, it's not a problem of “keywords”: it's a problem of format and utility. This approach fits Google's guide to useful, “people-first” content: content created to help, not to manipulate rankings.

Simple taxonomy: informational, comparative, “how-to”, navigational, decision

• Informational: seeks to understand (e.g. “what is SEO cannibalization”).
  
  
• How-to: seeks to execute (e.g. “how to measure traffic from ChatGPT in GA4”).
  
  
• Comparison: seeks to choose between options (e.g. “Semrush vs Ahrefs vs...”)
  
  
• Navigational: Search for a brand or site (e.g. “Search Console performance report”).
  
  
• Decision: Look for criteria to buy or implement (e.g. “best tool for keyword research”).
  
  

In 2026, it is usually healthy for the blog not to rely only on informational TOFU: it prioritizes comparisons, how-to and decision making to maintain value even if there are more answers in the SERP.

Step 4 Evaluate difficulty and “click potential” (not every volume brings clicks)

Volume without “click potential” can fool you. In a scenario where a large part of searches end without a click, you need to look at the SERP and ask yourself: “Are people going to the web here or are they left solving the results page?”

SERP signals: features, aggregators, forums and results dominated by brand

Quick checklist of low potential signals (not absolute, but indicative):

• The SERP is full of modules (summaries, extensive PAA, carousels, etc.).
  
  
• The answer seems “closed” in SERP (basic definition, trivial list).
  
  
• Aggregators or “large” aggregators that are difficult to move dominate.
  
  
• Very brand-heavy results (if you're not a brand, the effort goes up).
  
  
• The real intention is different from your initial idea (format mismatch).
  
  

The recommendation: before deciding on a keyword, look at the top 10 and define what type of content “wins” there.

How to estimate effort vs return (2x2 model)

Use a simple array:

• Impact (high/low): business value + click potential + thematic authority.
  
  
• Effort (high/low): complexity of content, need for assets (tables, templates), competence, technical requirements.
  
  

Typical prioritization:

• High impact/low effort: quick wins.
  
  
• High impact/high effort: pillar parts.
  
  
• Low impact/low effort: strategic filler (little).
  
  
• Low impact/high effort: normally discard.
  
  

Step 5 Build clusters and architecture (giving you thematic authority)

Clusters help you scale and resist SERP changes: you don't depend on a keyword, but on covering a topic with real sub-questions.

Practical model:

• Pilar (canonical): reference guide with definitions, criteria, limits, checklist.
• Satellites: sub-topics that answer specific questions (and link to the canonical one).
• Internal links: coherent widths and paths to MOFU/BOFU.

To begin with, 3-5 well-built clusters usually win 30 individual posts.

From list to cluster: group by entity/topic, not by synonyms

Group by:

• Same job (what the reader is trying to achieve).
• Same stage of the funnel (learning vs decision).
• Recurrent subthemes in PAA and in top 10 content.
• Repeated entities (tools, metrics, processes, documents).

Avoid grouping only by “similar words”: this creates cannibalization and redundant content.

Map clusters to “canonical” reference pages

For each cluster, choose 1 canonical “reference” URL and decide what assets it will have (table, checklist, template, examples). This piece must be the most careful and maintained, because it is the one that accumulates internal links, mentions and updates.

Step 6 Prioritize and convert into an editorial calendar (with measurable criteria)

When you already have tagged keywords and defined clusters, the goal is to turn it into an executable calendar.

Suggested final table:

Practical decision: if you can't assign a date and “proposed piece”, the research isn't ready yet.

Rule 60/30/10 for an SEO blog (balancing objectives)

A reasonable mix to grow with stability:

• 60% how-to/guides (authority + recruitment).
  
  
• 30% comparatives/alternatives (decision).
  
  
• 10% trends/news (only if you can keep up to date and quality).
  
  

Adjust according to resources: if you are a niche with little time, reduce “news” to a minimum.

QA checklist before publishing (to avoid generic content, even if you use AI)

Before indexing, validate:

• Do you respond to the job in the first paragraph?
  
  
• Does it include concrete examples (not just theory)?
  
  
• Are there clear definitions and self-sufficient sections?
  
  
• Do you provide a useful asset (table, checklist, template) if the intention asks for it?
  
  
• Do you avoid dubious claims and exaggerations?
  
  
• Do you have consistent internal links to the cluster?
  
  
• Is it aligned with useful/people-first content (not written “for the robot”)?
  
  

Note on AI: Google prioritizes useful content regardless of how it is produced; the risk lies in publishing at scale without providing value or without quality review.

Workflow with Makeit Tool (SEMrush type suite): how to systematize the process

A SEMrush-like suite like Makeit Tool is a good fit so that this method isn't “a loose Excel”:

• Keyword research and expansion.
  
  
• SERP observation and competition.
  
  
• Clustering and intention.
  
  
• Prioritization by impact/effort.
  
  
• Tracking (what was published, what was updated, what changed in metrics).
  
  

The idea is to have a repeatable system that you can operate alone or with equipment.

From opportunity to backlog: how to document decisions for your team/client

Minimum brief per piece (reduces friction and improves consistency):

• Objective (what the reader should achieve).
  
  
• Intent and dominant format of the SERP.
  
  
• Outline with H2/H3 and “answer blocks” per section.
  
  
• Required assets (table, checklist, example, template).
  
  
• Internal links (to canonical and related pieces).
  
  
• Expected KPI (GSC: Impressions/CTR; GA4: engagement; business: leads/conversions).
  
  

Frequently asked questions about keyword research for a blog

How many keywords do I need to start a blog?

A practical starting range is usually 200—500 “real” keywords, but it's not the number that's important: it's having 3-5 well-defined clusters with one canonical per cluster. With that, you can plan 8—16 weeks of content without improvising or falling into individual posts.

Do I use Keyword Planner if I don't do Google Ads?

Yes, as a source of ideation and for demand estimates. Keyword Planner is designed for Ads, but it is used to discover variations, group themes and understand search ranges. Read it as an approximation, not as exact SEO data, and it always validates intent by looking at the SERP.

How do I know the correct intent of a keyword?

“Real” intent is defined by the SERP. Check out the top 10: if comparisons dominate, Google is rewarding decision; if they master step-by-step guides, it's how-to; if there are brand pages, it's navigational. Also note features (PAA, modules) and the type of content that is repeated.

What do I do if two posts are competing for the same keyword (cannibalization)?

First, identify which URL should be the canonical one for that subintention. Then consolidate: merge content if it contributes, redirects if appropriate, or reorients each piece to different subintentions (with clear titles and sections). It reinforces the internal link to the canonical one and avoids publishing almost identical variations.

How does AI affect keyword research in 2026?

The value of working in clusters and sub-questions increases, and the ROI of trivial “summarizable” content decreases. It is also more important to choose formats that provide criteria (comparisons, decision matrices, checklists) and to keep canonical pieces updated. In an environment with more no-click searches, click potential and utility weigh more.

What metrics are essential to prioritize?

At a minimum: business value, intent, relative difficulty, click potential (according to SERP) and available resources. If you already have a blog, add signs of closeness from the Search Console (high impressions with low CTR, queries in positions 8—20) to detect quick wins.