Keyword research and SERP features

In this lesson

Find keyword ideas from tools, from the results page and from the questions that real people ask.
Use search volume, keyword difficulty and traffic potential to pick which keyword gets a page first.
Read a results page to learn the intent and the format that Google rewards for a query.
Group keywords into topic clusters, and find the gaps between your site and a competitor.
Name the main SERP features, and say which ones you can influence and how.

Days 03 and 03b made sure that search engines can reach and read your pages. This lesson asks a different question: which pages should exist at all? Keyword research gives the answer. Then you meet the parts of the results page that are not plain blue links. Day 04b covers on-page SEO: how to write the title, headings and text of each page.

Keyword research

A keyword is a word or phrase that people type into a search engine. Keyword research is the work of finding the keywords for your topic, and deciding which ones are worth a page. It connects what you can offer with what people actually ask for.

Keyword research has three steps:

Collect ideas. Start with a few seed keywords: the obvious words for your topic, such as “coffee roasting”. Put them in tools, and look at the results page and at forums. Collect many ideas first. Do not judge them yet.
Group the ideas. Many keywords mean the same thing. “How to roast coffee at home” and “home coffee roasting guide” need one page, not two. Ahrefs calls this grouping clustering. Its tools group keywords by the pages that rank for them.
Pick and plan. For each group, check the demand, the difficulty and the intent. Then decide which page to make first.

Google’s autocomplete is a free source of ideas. Google says that autocomplete predictions “reflect real searches that have been done on Google.” Type a seed keyword slowly and note the predictions.

Most people search with short head terms, but much traffic comes from long, specific phrases. Such phrases are called long-tail keywords. Each one has a small volume, but the competition is usually lower and the intent is clearer. Ahrefs calls it “a common rookie mistake to disregard low search volume keywords.”

Google warns about one way to use keywords badly. Its guide on helpful content asks: “Are you producing lots of content on many different topics in hopes that some of it might perform well?” Day 02 showed that scaled content abuse is a spam policy. Keyword research shows you what people need. It is not a list of pages to mass-produce.

Volume, difficulty and traffic potential

Three numbers help you compare keywords. All three are estimates, and each tool calculates them in its own way.

Metric	What it estimates	Watch out
Search volume	How many times people search for the keyword in one month, on average.	It counts searches, not people. It is a 12-month average, so a seasonal keyword can look small or large. It is for one country.
Keyword difficulty (KD)	How hard it is to rank in the top 10 for the keyword, on a scale of 0 to 100.	Each tool uses its own formula. Ahrefs bases KD on the number of websites that link to the top 10 pages. Semrush also uses the authority of the ranking domains and the features of the results page.
Traffic potential (TP)	How much search traffic the top page for the keyword gets from all the keywords that it ranks for.	It is an Ahrefs metric. Other tools have similar ideas under other names.

Ahrefs gives two examples of how volume misleads. A keyword with 120,000 searches in December and none in other months shows a volume of 10,000. “Black friday” shows an average of 282,000, but it had 2.5 million searches in November 2024. Ahrefs also says that a page at position 1 rarely gets more than 30% of the keyword’s volume. That is Ahrefs’ estimate, not a Google number.

Traffic potential often matters more than volume. One good page ranks for hundreds of related keywords, not only the one that you picked. So a keyword with a small volume can lead to a page with a large total traffic. The diagram shows this with example data.

Semrush divides its KD% into six bands, from “Very easy” (0 to 14) to “Very hard” (85 to 100). Ahrefs warns that you should not filter for easy keywords only. A high KD means that you need many links, so start early on the topics that matter most. Day 06 covers links.

Highest keyword difficulty (KD) that your site can win now:

Rank the keywords by Search volume Traffic potential

The page that you plan Any page type A guide (informational intent) A product page (commercial intent)

Example data for a site about home coffee roasting, not real tool output. Move the slider to set the difficulty that your site can win. Then switch from search volume to traffic potential. Notice that the first pick changes: a long keyword with a small volume can have the largest traffic potential.

The most important rule: intent beats numbers. A keyword is only worth a page if your page can satisfy the intent behind it. Day 01 explained search intent. A product page cannot rank for “how to roast coffee beans at home”, whatever the numbers say.

Keyword tools

Three tools appear on the roadmap. Each one has its own data, so their numbers do not agree.

Tool	What it is	Cost (9 October 2026)
Google Keyword Planner	A tool inside Google Ads, made for advertisers. It shows “average monthly searches” from Google’s own data, and bid estimates.	Free, but you need a Google Ads account with billing information.
Semrush	A full SEO platform: keyword research, competitor research, rank tracking, links and AI visibility.	A free account gives 10 reports a day. Paid plans start at $139 a month.
Ahrefs	A full SEO platform, known first for its large backlink index. It also has keyword research, traffic potential and parent topics.	Free tools, such as the Keyword Generator, show a limited list. Full data needs a paid plan.

Keyword Planner data has three limits for SEO:

Its “Competition” column is about advertisers. It is “the number of advertisers that showed on each keyword”. It is not organic difficulty.
Its volume is for exact matches, and Google says that the numbers are rounded.
It does not show keywords with “very low search volumes”.

Many SEO guides say that accounts with little ad spend see volume ranges, such as “1K – 10K”, not exact numbers. The Google help pages that I read for this lesson do not say this. So treat it as a claim that is not confirmed.

The best use: take Keyword Planner as Google’s baseline for demand. Use Semrush or Ahrefs for difficulty, traffic potential and competitor data.

SERP analysis

SERP analysis means that you look at the actual results page for a keyword before you write. Tools give numbers, but the results page shows what Google rewards for the query today. The results depend on location and personalization, as day 01 showed. So check from a private window, in the country that you target.

Look for four things:

Intent. Are the top results guides, product pages, category pages or tools? Semrush calls a query with mixed results “fractured search intent”. Then more than one page type can rank.
Format. Are they lists, step-by-step guides, comparisons or videos? Match the format that wins.
SERP features. Is there a featured snippet, a video row or a knowledge panel? They take space from the plain results. See the last section.
Gaps. What do all top pages miss? Your page can add that.

SERP analysis is the step that people skip most often. It is also the cheapest check. It costs one search and protects you from a page in the wrong format.

Find real questions

Keyword tools show short phrases without context. Question mining is the work of finding the full questions that real people ask about a topic. Four sources are on the roadmap:

People Also Ask (PAA). This is a box of related questions on the results page. Google calls it a “related questions group”. When you open a question, it shows a featured snippet. Each open question adds more questions, so one search can give dozens.
Forums and communities. On Reddit, Quora, Stack Overflow and niche forums, people ask questions in their own words. They also say what the existing answers miss. Google has a “Discussions and forums” feature, and forum sites can mark posts with DiscussionForumPosting structured data.
AI chat prompts. People ask AI assistants long, conversational questions. Semrush defines prompt research as finding “the questions that cause AI systems to compare options and recommend specific brands”. Semrush says that prompts have no history of search volume, unlike keywords. Day 07 covers AI search.
Your own data. Day 08 covers the Performance report in Search Console. It shows the queries that already bring people to your site.

Questions are good for two reasons. They show the exact words that people use, and they show the intent clearly.

Competitor gap analysis

Competitor gap analysis finds the keywords and topics that your competitors rank for and your site does not. Ahrefs defines a content gap analysis as “the process of finding topics your competitors have covered but you haven’t.”

The method is simple. Pick two or three sites that rank for your main keywords. Put them in a gap tool, such as Keyword Gap in Semrush or Competitive Analysis in Ahrefs. The tool lists the keywords where they rank and you do not. Then filter the list by intent and by what your site can really offer.

Pick your competitors from the results page, not from the market. A large shop can be your business competitor, while a small blog is your search competitor for a guide.

Topic clusters

A topic cluster is a group of pages about one topic. One pillar page gives a broad overview. Several cluster pages go deep into the subtopics. Internal links connect the pillar page and the cluster pages in both directions.

For example, a pillar page “Home coffee roasting” links to cluster pages about roasting levels, equipment and storage. Each cluster page links back to the pillar page.

Topic clusters turn a keyword list into a site plan. Each keyword group from your research becomes one page. Day 03 explained internal linking and site hierarchy. The cluster gives search engines a clear map of which page covers which subtopic. It also stops two of your pages from competing for one keyword.

The name “topic cluster” comes from SEO tools and agencies, not from Google. Google does not name it as a ranking factor. But Google does recommend a logical site structure and helpful internal links.

SERP features

SERP features are the parts of the results page that are not plain text results. Google’s Visual Elements gallery says that the look of the page can change by device, country, language “and many other factors”. The roadmap lists five features:

Feature	What it is	What you can do
Featured snippet	A box at the top that shows the answer text first, then the source link.	Answer the question clearly in a short passage. You cannot mark a page for it: “Google systems determine” this.
People Also Ask	Related questions that open to show a featured snippet.	The same work as for featured snippets.
Sitelinks	Extra links to pages of the same site, under one text result.	Google says that they “are automated”. Use a logical structure, informative titles and clear anchor text.
Knowledge panel	A box of facts about an entity (a person, place, organization or thing) in the Knowledge Graph.	It is automatic. The entity can claim the panel and suggest changes.
Image and video packs	Rows of image or video results, for queries where people want to see something.	Use images in `<img>` elements next to related text. Give each video an indexed watch page with a stable thumbnail.

Featured snippets

Google says that featured snippets are shown when people want answers “found in a short piece of a website”. A click on the snippet takes the user to that part of the page. A page can opt out. The nosnippet rule blocks all snippets. A low max-snippet value makes a featured snippet less likely, but Google does not guarantee it.

Sitelinks and knowledge panels

These two show that Google knows your site or your brand well. You cannot request them. Sitelinks come from a clear structure. If you must remove a sitelink, Google says to remove the page or use noindex. Knowledge panels come from the Knowledge Graph, which collects facts about entities. Day 05 covers entities and structured data.

Image and video packs

Google does not index images that are only in CSS. An image sitemap can help Google find images. For videos, Google says that the watch page “must be indexed” and must perform well in Search before the video can show. Day 04b covers image optimization on the page.

What matters most

SERP features change the value of a ranking. A position 1 under a featured snippet, a PAA box and a video row can be below the first screen of a phone. So check the features during SERP analysis, and count them in your plan. If a feature appears, ask whether your page can earn it.

Query how long to boil an egg eiffel tower a brand name small living room ideas

Features on this page (turn one off to see the effect)

Example layout, not a real results page. Pick a query. Then turn features off one at a time. Notice how far the first plain result moves up, and read how each feature is earned.

AI Overviews also sit on many results pages. They are a large change for SEO, so day 07 covers them in full.

Key terms

Keyword: a word or phrase that people type into a search engine.
Keyword research: finding the keywords for a topic, and deciding which ones are worth a page.
Seed keyword: an obvious starting keyword that you put into tools to get more ideas.
Long-tail keyword: a long, specific keyword with a small search volume and usually a clear intent.
Clustering: grouping keywords that one page can satisfy.
Search volume: the average number of searches for a keyword in one month, in one country.
Keyword difficulty (KD): a tool’s estimate, from 0 to 100, of how hard it is to rank in the top 10.
Traffic potential (TP): the search traffic that the top page for a keyword gets from all its keywords.
SERP analysis: reading the actual results page for a keyword to learn its intent, format and features.
Fractured search intent: a query whose top results show more than one intent.
Question mining: finding the full questions that real people ask about a topic.
Prompt research: finding the questions that people ask AI assistants about a topic.
Competitor gap analysis: finding the keywords that competitors rank for and your site does not.
Topic cluster: a pillar page and its cluster pages about one topic, with internal links between them.
SERP feature: a part of the results page that is not a plain text result.
Featured snippet: a box that shows an answer from a page first, then the link to the page.
Sitelinks: extra links to pages of the same site under one result.
Knowledge panel: a box of facts about an entity from Google’s Knowledge Graph.

Exercise

Exercise · about 15 min

Plan one page from one seed keyword

Use a private browser window. Pick a seed keyword for a topic that you know, or use home coffee roasting.

Type the seed into Google slowly. Write down five autocomplete predictions.
Search for the seed. Open three questions in the People Also Ask box. Write down the new questions that appear.
Look at the top five results. For each one, write down the page type (guide, product, category, video, forum) and the format (list, steps, comparison).
Write down every SERP feature on the page. Where does the first plain result start on your screen?
Open the Ahrefs Keyword Generator (free). Put in the seed and look at the “Questions” list. Note the volume and KD of three ideas.
Search Reddit for the seed. Write down one question that the top results do not answer.
Write a short plan: the main keyword, the intent, the format, and three subtopics for cluster pages.

Show a possible solution

Results depend on your country and the date, so your lists will be different. This run did not check live results. The notes below say what to look for, with a possible plan for home coffee roasting.

Autocomplete and PAA can give questions such as “is it cheaper to roast your own coffee”, “can you roast coffee in an oven” and “how long do home roasted beans last”.
If the top results mix guides with steps and product pages for roasters, the intent is fractured. Then a guide is the safer page type for a blog.
Look for People Also Ask and a video row, which often appear for how-to queries. A shopping feature can appear if some people want to buy a roaster.
In a keyword tool, a long question such as “how to roast coffee beans at home” often has a lower KD than the head term “coffee roaster”.
Forum posts can show practical problems, such as smoke or noise. Check if the top guides answer them.

Plan: main keyword how to roast coffee beans at home, informational intent, a step-by-step guide with a short direct answer at the top and a video. Cluster pages: roasting levels, a comparison of home roasting methods, and how to store roasted beans. Each cluster page links to the guide, and the guide links to each cluster page.

Check yourself

Keyword A has a search volume of 9,000 and a traffic potential of 7,000. Keyword B has a volume of 800 and a traffic potential of 12,000. Both fit your site. Why can B be the better first page?Low-volume keywords always rank fasterThe top page for B gets more total traffic, because it ranks for many related keywordsTraffic potential counts searches from all countriesSearch volume is not a real number

Keyword Planner shows 'High' in the Competition column for a keyword. What does this tell you about organic SEO?It is very hard to rank in the top 10 organic resultsNothing directly: the column is about how many advertisers showed ads for the keywordGoogle gives the keyword a high keyword difficulty scoreThe keyword has a high search volume

The top 10 results for your keyword are all product category pages. You planned a long guide. What should you do?Write the guide, because long content always ranksRethink the page: the results page shows that Google rewards category pages for this queryAdd the keyword to the guide more oftenAsk Google to show your guide as a featured snippet

A site has three articles that all target 'how to descale a kettle'. What does a topic cluster change?Nothing, more pages on one keyword give more chances to rankOne page targets the keyword; the others cover distinct subtopics and link to itAll three pages get noindexThe three pages are merged into the home page

Your site ranks at position 1, but the query also shows a featured snippet from another site, a People Also Ask box and a video row. Which statement is correct?You can add markup that marks your page as the featured snippetYour result can be below the first phone screen, so the position is worth less than it looksSitelinks will replace the featured snippetSERP features do not change clicks, only impressions

Sources

Keyword Research: The Beginner’s Guide (Ahrefs)
Free Keyword Generator Tool (Ahrefs)
How to Do a Content Gap Analysis (Ahrefs)
Use Keyword Planner (Google Ads Help)
About Keyword Planner forecasts and statistics (Google Ads Help)
What is KD%? (Semrush Knowledge Base)
Semrush pricing (Semrush)
What Is a SERP Analysis and How Can You Do One? (Semrush)
Topic clusters for SEO (Semrush)
People Also Ask: What It Is and How to Optimize for It (Semrush)
How to do prompt research for AI SEO (Semrush)
How Google autocomplete predictions work (Google Search Help)
Creating helpful, reliable, people-first content (Google Search Central)
Visual Elements gallery of Google Search (Google Search Central)
Featured snippets and your website (Google Search Central)
How Google’s featured snippets work (Google Search Help)
Sitelinks (Google Search Central)
About knowledge panels (Knowledge Panel Help)
Google image SEO best practices (Google Search Central)
Video SEO best practices (Google Search Central)
Discussion forum structured data (Google Search Central)
roadmap.sh SEO roadmap: Keyword Research, SERP Analysis, SERP Features (roadmap.sh)

Day 04b turns a keyword plan into a page: title tags, meta descriptions, headings, URL slugs, keyword placement, readability, anchor text and images.