How do you build a semantic core that fits your website?

A semantic core that fits your website is a list of search queries your target audience uses to find products or services, mapped to appropriate pages on the site. Start by identifying the search intent each page should address. Then collect keywords from comparable competing pages, expand the list using seed keywords, and remove queries that do not match your products or service area. The goal is not to collect as many terms as possible, but to ensure the remaining terms align with what the website actually offers. Adapt the approach to the market size, available time, and tools.

Define each page’s purpose before choosing competing pages

Semrush’s bakery case study, dated 2021, examines how a homepage can address searches for a local bakery, rather than tackling every topic across the entire site at once. It starts with a client-approved website structure and then develops a query list for each page to be optimized.

Before collecting competitors’ keywords, identify pages that address similar search intent. If a competitor’s homepage focuses on other needs, look instead at a category or service page that better matches your objective. Do not adopt the competitor’s keywords wholesale simply because the business is in the same industry.

• Check whether the product or service users are looking for matches the content of the target page.
• Set the correct search region for the analysis and limit its scope to the exact page URL.
• Similar queries in title tags can help identify relevant pages, but titles alone should not determine your choice.

Expand with seed keywords while removing unsuitable terms

A seed keyword can be the shortest phrase that expresses a user’s intent. The case study above uses phrases such as “german bread” to expand the list and also looks for additional seed keywords in search result snippets. This generates candidate queries, but does not mean every term found is suitable for the website.

During filtering, the case study retains non-branded terms that match the search intent, product type, and service area. For queries containing place names, check especially whether those locations fall within the service area. If there are too many candidate terms, narrow the scope of analysis by language, search volume, and keyword difficulty.

The case study’s list contained 60 terms after cleaning and 29 after client approval, which the author considered sufficient for writing the landing page copy. This is not a target number that every website should meet. The case study’s exclusion of zero-search-volume queries should not be treated as a fixed rule for all websites either.

Separate ranking targets from terms used in copy, then plan the content

A query’s relevance to the business does not necessarily mean the website is well placed to compete in its search results. The case study above notes that results for “custom bakery near me” were dominated by local dining aggregators and review services, suggesting that the bakery might struggle to gain organic search visibility. However, descriptive terms in the query could still be used in page copy. This was a judgment made in the case study, not a conclusion that applies to every region.

When mapping terms to pages, first list the core questions, likely follow-up questions, and information readers need to take action. Then organize the content into short paragraphs and clear sections, incorporating relevant terms naturally rather than overusing the main keyword. At Procrustes, we also turn the questions our clients genuinely care about into articles, with ongoing content planning, writing, and updates.

Search volume, keyword difficulty, cost per click, and competition levels change over time, so keep these metrics up to date. Figures from a single export should not be treated as a permanent basis for decisions.