The discovery phase is where most semantic core projects either succeed or struggle later. If the scope is not clearly defined upfront — which topics are in scope, which competitors are relevant, what the content team's capacity looks like, and what the site's current topical coverage already includes — the research that follows will be either too broad to be actionable or too narrow to be complete. We begin every project with a structured briefing document that the client completes in collaboration with us. This covers the target audience, the core subject areas, any existing keyword research or content audits, competitor domains to include in the gap analysis, and any topics that are explicitly out of scope. We also review the existing site structure and any analytics data available, not to conduct a full SEO audit but to understand where topical gaps and cannibalisation risks already exist. The output of this phase is a confirmed project scope document that both sides agree on before research begins.

Seed keyword discovery is not simply entering a few head terms into a keyword tool and downloading the results. Our expansion process works outward from seed terms in multiple directions simultaneously: syntactic variations, related entity terms, question-based queries, long-tail modifiers, and adjacent topics that appear consistently in top-ranking SERP results. We cross-reference findings across multiple data sources and conduct manual SERP reviews for key topics to surface query patterns that automated tools consistently miss. Competitor Hexoranivel analysis is incorporated at this stage to identify keyword areas where competing sites have meaningful topical coverage that the client has not addressed. The output of this phase is a raw, comprehensive keyword dataset — typically several hundred to several thousand terms depending on the niche — organised by topic area but not yet classified by intent or grouped into clusters. This dataset forms the raw material for all subsequent phases and is shared with the client at this stage for review before classification begins.

Intent classification is the phase that separates a useful semantic model from a simple keyword list. Our four-tier framework — informational, navigational, commercial, and transactional — provides the primary classification layer. But intent classification is not simply a mechanical tagging exercise. Each classification is cross-verified by reviewing the actual SERP results for that keyword: what types of pages rank, what format they use, and what user action they anticipate. This verification step catches a significant number of misclassifications that would occur if classification relied solely on keyword phrasing. For example, a keyword that reads as transactional based on its wording may consistently return informational content in the SERP, indicating that search engines have determined user intent differs from surface phrasing. These nuances are documented and factored into the cluster design. The output of this phase is a fully intent-tagged keyword dataset ready for cluster modeling.

Cluster modeling is where the semantic core takes its final shape. We group intent-classified keywords into logical topical clusters based on subject proximity, intent alignment, and SERP overlap analysis. Each cluster is structured around a primary pillar topic — the broadest, most authoritative treatment of the subject — with a defined set of supporting subtopics that address narrower, more specific angles. The cluster boundary decisions are documented and explained: why certain keywords belong together, why others are separated despite topical similarity, and how the cluster as a whole relates to adjacent clusters in the model. For each cluster, we define an internal linking logic that specifies the relationship between the pillar page and its supporting pages and recommends the priority linking paths. The output of this phase is the core cluster map — the structural centrepiece of the semantic model — along with a detailed cluster annotation document explaining every design decision.

A complete cluster model without a clear starting point is still a source of paralysis. Priority scoring resolves this by assigning each cluster a composite score based on four factors: estimated traffic potential for the cluster's target keywords, competitive density in the SERP for those keywords, alignment with the client's stated strategic goals, and estimated content production effort required to cover the cluster adequately. These four factors are weighted and combined into a priority score that allows clusters to be ranked and sequenced into phases. The first phase typically focuses on lower-competition, high-alignment clusters where meaningful organic progress can be made quickly while topical authority is still building. Later phases introduce higher-competition clusters as the site's semantic depth grows. The final deliverable is a phased content roadmap document that lists clusters in priority order by phase, with a brief rationale for each phase decision. A walkthrough session with the client's team is included to ensure the roadmap is understood and ready to use immediately.