Most stores do keyword research exactly once. Someone picks a handful of terms for the homepage during the launch, the categories get named after whatever the supplier’s catalog called them, and from then on the “strategy” is adding products. Two years later the store ranks for its own brand name and not much else.
The search landscape punishes that approach more every year. Ahrefs’ analysis of its US database found that almost 93% of all keywords get fewer than 10 searches a month, and roughly 15% of the searches Google sees each day have never been typed before. Demand is spread across millions of specific phrases, and a store that only targets its ten biggest category names is competing for the smallest, most crowded slice of it.
Keyword research for an online store is not about finding one perfect term per page. It is about understanding how people describe what you sell, at every stage from “what should I buy” to “buy this exact model”, and then building your categories, filters, product pages, and guides around that language. This guide walks through the process in the order that produces a usable keyword map rather than a spreadsheet nobody opens again.
Why keyword research for a store is different
A blog maps one keyword to one article. A store maps keywords to a hierarchy, and most of the commercial value sits in the middle of it, on category and subcategory pages, not on the homepage and not on individual products. Before collecting a single keyword, it helps to know which page type each kind of query belongs to:
- Category keywords such as “women’s trail running shoes” carry commercial intent and the most revenue. They belong on category and subcategory pages.
- Product keywords such as a brand plus a model name are transactional and often branded. They belong on product pages, and you rarely have to research them because the product name is the keyword.
- Comparison and “best” keywords such as “best trail running shoes for wide feet” belong on buying guides that link down into the categories they discuss.
- Informational keywords such as “how to clean trail running shoes” belong on blog content whose job is to build authority and pass links to the money pages.
- Branded keywords, meaning your own store name and variations, which you will rank for anyway and can mostly ignore during research.
Getting this mapping wrong is one of the reasons stores struggle even when they do everything else right. If you want the wider picture of how these pieces fit, our guide to getting your ecommerce store found on Google covers it; the rest of this article is the keyword side alone.
Step 1: Start with what you already rank for
The cheapest keyword data you will ever get is sitting in Google Search Console. The Performance report shows every query that produced an impression for your store, with clicks, click-through rate, and average position, going back 16 months. Before opening any paid tool, export it and look for three things.
First, queries where you sit between positions 8 and 30. These are terms Google already associates with your pages. Moving them onto page one is usually a matter of improving the page that ranks, not creating a new one, and it is the fastest win in this whole process.
Second, queries that land on the wrong page. A product page ranking for a category term, or a blog post ranking for a query that should go to a category, is a signal that the right page either does not exist or is too weak to compete with your own content.
Third, queries you did not expect at all. Stores routinely discover that customers describe products with words the store never uses: “rain jacket” when every page says “waterproof shell”, or a regional term for a product type. Those words go straight into the seed list.
Step 2: Build the seed list from your catalog, not from a tool
Keyword tools are expansion engines. They multiply whatever you feed them, so the quality of the seed list decides the quality of everything after it. Build it from the business rather than from guesses:
- Every category and subcategory name exactly as it appears in your navigation
- Product types and their plural forms, since “backpack” and “backpacks” often show different intent
- Attributes customers filter by: material, size, color, capacity, compatibility, use case
- Problems the products solve, phrased the way a customer would say them to a friend
- Synonyms and slang pulled from reviews, support tickets, chat transcripts, and your own site search logs
- Competitor navigation, because the category names your three closest rivals use are a free map of the terms they consider worth a page
Site search logs deserve extra attention. When someone types a term into your store’s search box, they are telling you both the word they use and the fact that they could not find it in the navigation. If a term shows up repeatedly and you have products for it but no page, that is a category waiting to be built. How those categories then fit together matters as much as the words themselves, which is why keyword research and site architecture should be planned as one exercise rather than two.
Step 3: Expand the list
Now the tools earn their keep. Feed the seed list into a keyword tool and pull matching terms, related terms, and questions for each seed. For a store with 30 categories this will produce several thousand keywords, which is the point; you will filter hard in the next step.
If budget is zero, Google’s Keyword Planner works with a Google Ads account, though accounts without ad spend see broad volume ranges rather than exact figures. Google autocomplete, the “People also ask” boxes, and the related searches at the bottom of results are free too, and for commercial terms they surface the modifiers people actually add: “for men”, “under 100”, “with pockets”, “vs”, “alternative”.
Paid tools (Ahrefs, Semrush, and similar) add three things worth paying for at this stage: keyword difficulty scores, the ability to see every keyword a competitor’s category page ranks for, and clustering, which groups thousands of variations into the handful of pages that should exist. Run the expansion per seed, not for the whole list at once, so that each category ends up with its own group of candidate terms.
Step 4: Read volume and difficulty without fooling yourself
Volume numbers are estimates, and treating them as facts is the most common mistake at this stage. Ahrefs’ comparison found that Google’s Keyword Planner overestimates search volume 54.28% of the time, and every third-party tool models its numbers from a sample. Use volume to rank keywords against each other, not to forecast traffic to the dollar.
Difficulty scores need the same skepticism. Most tools calculate difficulty from the backlinks of the pages currently ranking, which says nothing about who those pages belong to. A category term with a low difficulty score where the top ten is Amazon, Walmart, and two national retailers is not easy; it is a term where the ranking pages simply do not need links. For each candidate keyword, open the results and check:
- Who ranks: marketplaces and national brands across the whole page usually means the term is out of reach for now, whatever the score says
- What ranks: if the top results are category pages, Google wants a category page; if they are guides, a category page will not get there
- What surrounds the results: Shopping ads, an AI summary, and image packs all pull clicks away from the organic listings, so a 1,000-search keyword with all three may deliver fewer visits than a 300-search keyword with none
Then set a realistic ceiling. A newer store should be filling its map with terms where comparable stores rank, not where the biggest retailers in the country rank, and it should come back for the harder terms once the easier ones are producing revenue.
Step 5: Sort by intent and map every keyword to a page
Every keyword that survives Step 4 gets two labels: an intent and a destination URL. The intent labels are the five from the start of this article. The destination is either an existing page or a new page that needs to be created, and you should be able to say which page type it is.
This is where the earlier work pays off. Terms like “insulated water bottle” and “insulated water bottles” and “insulated stainless steel water bottle” are one category page, not three. “Best insulated water bottle for hiking” is a buying guide. “How long does an insulated bottle keep ice” is a blog post. “Hydro Flask 32 oz wide mouth” is a product page. Google’s own SEO Starter Guide makes the point that different people search for the same thing with different words, and the practical response is to gather those words on one page rather than spread them across several.
The output is a keyword map: a spreadsheet with one row per page, listing the primary keyword, the secondary variations the page should also cover, the intent, the page type, and whether the page exists yet. That document is the actual deliverable of keyword research. The list of keywords is only raw material.
Step 6: Check for cannibalization
Stores create cannibalization faster than any other kind of site because the platform generates pages automatically. A filter for “waterproof” inside the jackets category can produce an indexable URL that competes with a “waterproof jackets” subcategory. A tag page competes with a category. A product page for a single bestseller competes with the category it lives in.
For every primary keyword on the map, search your own site for other pages that target the same term, including auto-generated filter and tag URLs, and pick one winner. The losers either get consolidated, canonicalized to the winner, or re-pointed at a different term. Product pages ranking for terms they should not own is a frequent symptom, and it usually traces back to the reasons we covered in why most ecommerce product pages never rank: thin category pages leave Google no better option than a product page.
Step 7: Find the long tail that converts
The head terms in your map will take months. The long tail pays the bills while you wait, and for a store it is more structured than most people expect. Work through these modifier groups against every category:
- Use case: “for hiking”, “for small kitchens”, “for sensitive skin”
- Audience: “for kids”, “for beginners”, “for seniors”
- Attribute combinations: material plus color plus size, which are often worth their own indexable filter page when volume exists
- Price framing: “cheap”, “budget”, “under 50”, “premium”
- Occasion: “gift”, “wedding”, “back to school”
- Comparison: “X vs Y”, “X alternative”, “is X worth it”
- Questions: “how to choose”, “what size”, “how long does”, “can you”
Question keywords are where most stores put blog content, and where most store blogs go wrong by answering questions that never lead to a purchase. Each informational keyword on the map should be tied to the category it supports, and the post that targets it should link to that category in the first few paragraphs. We covered the pattern in detail in why your ecommerce blog isn’t driving sales; at the research stage, the rule is simple: if you cannot name the category a question supports, it does not go on the map.
Step 8: Steal the gaps from competitors
Pick the three stores that outrank you most often for your category terms and pull the full list of keywords each one ranks for. Remove everything on your own map and everything branded, and what remains is a list of demand you have not addressed, sorted by someone else’s success. Category ideas, attribute pages, buying guides, and product types you do not stock yet all show up here.
Most of the ecommerce SEO tools on the market do this in a single report, usually called a content gap or keyword gap, and it is the single most efficient hour in the whole process. Feed the results back into Step 4 and Step 5 so that every new term gets the same intent and difficulty check as the rest.
Turning the research into a plan
A keyword map with 400 rows is not a plan. Sort it by three things: how much revenue the page type carries (categories first, then guides, then product and informational pages), how achievable the primary keyword is based on who currently ranks, and whether the page already exists. Pages that exist and rank between positions 8 and 30 for a real commercial term go to the top. New category pages with achievable head terms come next. Informational content comes last, in batches, and always tied to a category.
Before building anything new, make sure the pages you already have are indexable and not tripping over each other; our ecommerce SEO audit checklist covers that in order. Then schedule the research itself as a quarterly task. Products change, seasons change, and a meaningful share of what people search for this month did not exist last year. The stores that treat keyword research as a living document, rather than a launch checklist, are the ones that stop ranking for their brand name alone.
About the Author
Matija Konjić is the founder of Link Inbound, a link building and content marketing agency working with B2B and B2C brands. He has built campaigns across 40+ industries and obsesses over the data behind what actually moves rankings.