AM Niche keyword research is the job of finding the exact searches real people type about a narrow topic, then deciding which of those searches a new site can realistically win. For a beginner, the hard part is not finding keywords. Any tool will hand you thousands. The hard part is cutting that pile down to a short list of pages worth writing first.
This guide walks through a method you can run with a browser and a spreadsheet: start with seeds, mine Google autocomplete, expand with questions and modifiers, group the results into clusters, check difficulty by looking at the actual search results, and pick your first 30 pages. It works for blogs, affiliate sites and small stores alike.
What niche keyword research is (and what it is not)
Broad keyword research asks what the whole internet searches for. Niche keyword research asks a narrower question: inside this one topic, which searches have clear intent, are answered poorly today, and fit a site like mine? The output is not a giant export. It is a map of topics grouped into clusters, with each cluster tied to one page.
Three ideas guide everything below:
- Real searches beat estimated volume. Volume numbers from third-party tools are modeled estimates, and for long, specific phrases they are often zero or missing even when people clearly search them. Autocomplete suggestions are drawn from searches people actually run, so a phrase that shows up there is being typed by someone.
- One page per intent. Ten keyword variations that want the same answer belong on one page, not ten.
- The search results tell you the difficulty. A difficulty score is mostly a guess based on backlinks. Looking at who ranks today tells you whether a small site has a chance.
Step 1: Build a seed list
Seeds are the short, obvious words that describe your niche. If your niche is home espresso, your seeds might be "espresso machine", "espresso grinder", "tamper", "milk frother", "espresso beans" and "descaling". Aim for 10 to 20 seeds. Fewer and you miss whole branches of the topic. More and you are probably drifting outside the niche.
Good places to find seeds:
- The category menus of the biggest stores in your niche. Their navigation is a ready-made list of product types.
- Subreddit sidebars, forum section names and the tags people add to posts.
- The table of contents of a popular beginner book on the topic.
- Your own experience: the tools, problems and steps you know from doing the thing.
Write each seed into column A of a spreadsheet. Add a column for where the seed came from so you can see later which sources produced useful keywords.
Step 2: Mine Google autocomplete from A to Z
Google autocomplete is the list of suggestions that drops down as you type. Each suggestion reflects real search activity, which makes it the most honest free data source a beginner has. The trick is to query it systematically instead of typing one phrase and stopping.
- Open a private or incognito window so your own search history does not shape the suggestions.
- Type a seed followed by a space, for example "espresso grinder ", and record every suggestion.
- Type the seed plus each letter of the alphabet: "espresso grinder a", "espresso grinder b", and so on through "z". Each letter surfaces a different set of suggestions.
- Repeat with words in front of the seed where it makes sense ("best espresso grinder", "cheap espresso grinder", "quiet espresso grinder"), since many real searches put the modifier first.
- Paste everything into your sheet, one keyword per row, next to the seed it came from.
- Remove exact duplicates and anything off-topic, such as navigational searches for a brand's login page or customer service number.
Done by hand, one seed takes 10 to 15 minutes, so 15 seeds fill an afternoon. Suggestions also vary by country: "descaling solution" in the UK returns different phrasing and brands than the same search in the US. Decide which country you are writing for before you start and stay with it.
If the manual version gets tedious, the Niche Research and Keyword Miner in AM Jarvis runs the same A to Z autocomplete mining across 18 countries and groups the output into a keyword map, so you can spend your time judging keywords instead of copying them.
Step 3: Expand with questions and modifiers
A to Z mining finds what follows your seed. Question and modifier expansion finds the searches that wrap around it. Run your seeds through autocomplete again using these patterns:
| Modifier type | What to type | Example result | Page it suggests |
|---|---|---|---|
| Questions | how, what, why, can, does, is | how to clean espresso grinder burrs | Step-by-step how-to |
| Comparisons | vs, or, compared to, alternative | flat vs conical burr grinder | Comparison article |
| Buyer | best, top, cheap, under $X | best espresso grinder under 300 | Roundup or buying guide |
| Problems | not working, won't, too, fix | espresso grinder clumping | Troubleshooting guide |
| Audience | for beginners, for small kitchen, for travel | espresso machine for small kitchen | Filtered roundup |
| Specs | size, settings, temperature, ratio | espresso grind size chart | Reference page |
Two more free sources are worth a few minutes each. The "People also ask" box on a results page shows questions Google has connected to the search, and opening a few makes more appear. The "Related searches" block at the bottom of the page shows how people rephrase the same need. Add anything useful to the sheet with a note saying where it came from.
By the end of this step, a niche of moderate size usually gives you several hundred rows, sometimes a couple of thousand. That is normal. The next step is where the list gets smaller.
Step 4: Group keywords into clusters
A cluster is a group of keywords that one page can fully answer. "How to descale an espresso machine", "descaling espresso machine with vinegar" and "how often to descale espresso machine" might look like three topics, but a person searching any of them would be happy with one thorough descaling guide. That is one cluster and one page.
A quick test for whether two keywords belong together: search both and compare the top 10 results. If several of the same URLs rank for both, Google treats them as the same intent and they belong on one page. If the results are almost entirely different, they need separate pages.
To cluster a sheet by hand:
- Sort by the seed column so related phrases sit next to each other.
- Add a "cluster" column and give each group a short label, such as "descaling" or "grinder-vs".
- Pick the clearest phrasing in each group as the primary keyword. The rest become secondary keywords and likely subheadings.
- Add an "intent" column: informational (wants to learn), commercial (comparing options) or transactional (ready to buy).
- Delete clusters that do not fit your site, such as local "near me" searches for a national blog.
A few hundred raw keywords usually collapse into somewhere between 40 and 120 clusters. Each surviving cluster is a candidate page. If you want to see how clusters stack into pillars and a full site structure, the companion guide on keyword research for niche sites covers topical maps and keyword-to-page mapping in detail.
Step 5: Judge difficulty by looking at the SERP
Keyword difficulty scores from tools are mostly built from how many backlinks the top pages have. They are a reasonable first filter, but they miss what matters most to a new site: whether the pages ranking today answer the search well. The fastest way to find out is to look.
For each cluster's primary keyword, search it in a private window for your target country and read the first page. Score it against these signals:
| Signal | Suggests an opening | Suggests a hard fight |
|---|---|---|
| Who ranks | Forums, Reddit threads, Q and A sites, small blogs | Major brands, large publishers, government sites |
| Page match | Results answer a related but different question | Every result is a dedicated page on the exact topic |
| Content quality | Thin, outdated or obviously rewritten pages | Detailed, current pages with original photos or testing |
| Freshness | Top results are several years old | Recently updated pages throughout |
| SERP features | Plain results, few ads | Heavy ads, shopping units, an AI answer that fully resolves the question |
A forum thread in the top five is the classic sign of an opening. It means Google could not find a dedicated page good enough to outrank a conversation. Several small sites on page one is another good sign, because it shows the query does not require a large, established domain.
Give each cluster a simple score from 1 (wide open) to 5 (dominated by big brands). You do not need precision. You are sorting, not predicting.
Tip: Note the page type that ranks, not only who ranks. If the top results for "espresso tamper" are all product and category pages, a blog post will struggle there no matter how good it is. Save that keyword for a store page, or skip it.
Step 6: Pick your first 30 pages
Thirty pages is a practical first target. It is enough to show search engines that the site covers a topic in depth, and small enough to finish in a month or two of steady work. Choose them with these rules:
- Start with difficulty 1 and 2 clusters. Early rankings build impressions and show you what Google responds to on your site.
- Stay inside two or three sub-topics. Thirty pages spread across ten sub-topics look thin everywhere. Thirty pages in three sub-topics look like real expertise.
- Include one or two pillar pages. A broad guide such as "espresso grinder guide" gives your cluster pages a hub to link up to.
- Mix intents. Roughly two thirds informational and one third commercial gives you pages that earn trust and pages that can earn money.
- Prefer topics you can answer better than the current results. If you have tested the product or solved the problem yourself, that experience is your edge.
Put the 30 in the order you will publish them, with pillars and their closest clusters first so internal links make sense from day one. Then write. Research that never turns into published pages is wasted time.
Common beginner mistakes
- Chasing volume. The biggest keywords in a niche usually belong to the biggest sites. A page ranking third for a small, specific phrase brings more visitors than a page stuck on page five for a big one.
- Writing one page per keyword variation. This splits your effort and makes your own pages compete with each other. Cluster first, then write.
- Ignoring intent. If the results are all product pages, an article will not fit. If they are all how-to guides, a product page will not fit.
- Researching forever. A keyword list goes stale if you sit on it. Once 30 pages are chosen, start publishing and do more research after you have real data.
- Skipping the country setting. Spelling, brands and phrasing change between the US, UK, Canada and Australia. Research for the market you will actually serve.
Once pages are live, open Google Search Console after a few weeks and go to Performance, then Search results. The queries your pages already get impressions for are the best keyword research you will ever have, because they show what Google thinks your site is about. Feed those back into your sheet and plan the next 30. When you are ready to turn a keyword map into pages, the AI Site Builder can build pillar and cluster pages from it. If you are still deciding what topic to cover at all, start with how to find a niche.