Published:
Updated:
by Wayne Smith
Any serious work in SEM quickly reveals the limitations of optimizing for a single keyword. Most real-world search queries include multiple terms or modifiers. This behavior likely arose as users learned that adding extra terms helped them refine results and locate exactly what they wanted. Early organic search practitioners who recognized this pattern adopted long-tail keyword strategies.
Table of Contents
Part 1 – Relationship Between Keyword Research and Marketing: Keyword research and marketing research are closely connected. Fan-out queries—the modern evolution of long-tail strategies—reveal market size and uncover opportunities. Canonical entities act as authoritative targets within the knowledge graph, while individual user intent reflects audience-specific linguistic nuances. The “canonical loophole” explains why branded queries often rank effortlessly when the brand is already established as a canonical entity.
Part 2 – Long-Tail Keyword and Fan-Out Query Mapping: Mapping topics and entities using schema can help manage keyword cannibalization. Query research also informs topical structure, using hubs and spokes to build comprehensive coverage and establish topical authority.
People new to SEO often discover that the majority of search traffic—sometimes as high as 85%—comes from long-tail queries. Historically, advertisers could see this distribution more clearly because pay-per-click tools exposed long-tail keyword data. Many inexperienced SEO practitioners responded by creating separate pages for numerous long-tail variations, as these variations are often low-competition terms. However, this approach can easily lead to content cannibalization, where multiple pages compete for nearly identical terms with overlapping content. Historically, this risk was managed using on-page SEO for titles and headlines to avoid cannibalization.
The 2023 “Helpful Content Update” penalized many sites that had created multiple pages with nearly identical content targeting slight query variations. Post-HCU, it is essential for organic SEO to focus on entities—“things, not strings”—rather than just PPC keywords. Relying on on-page SEO to prevent cannibalization is fragile and often ineffective in AI-based RAG document retrieval systems, which may return duplicate results when identical entities appear.
Long-Tail Keyword and Fan-Out Query Research: Long-tail keyword research focuses on specific keyword variations and is particularly useful for pay-per-click marketing. Fan-out query research, on the other hand, emphasizes entities and examines how modern AI-powered answer engines and search engines leverage information to identify pages that demonstrate experience and expertise.
LSI Keywords, Honorable Mention: At one time, many SEO professionals believed that Google was using LSI technology and that including LSI terms on a page could influence ranking. However, Google clarified that they do not use LSI, but instead rely on layered AI technologies.
Finding long-tail and fan-out queries is covered in Keyword Research for Entities: Marketing Intelligence, Part One of this keyword research series. This page focuses on mapping out the keywords and entities for a page or for a cluster of related pages.
Topical Entity Mapping
While most organizations agree on what needs to be optimized, the methods for doing so efficiently are often proprietary. Implementation workflows directly affect SEO cost structures, operational efficiency, and competitive advantage.
Schema markup, however, provides a practical and transparent way to map the entities represented on a page, reinforcing experience and expertise signals. It also helps structure a site’s content to reduce cannibalization by clearly defining the relationships between pages that address similar or overlapping topics.
Within the "about" property in schema, the "@id" property can be used to uniquely identify entities, and a URL from the site is a valid and effective @id value. This organizational approach helps clarify and structure a topical hub when reviewing or planning the schema.
Before creating an entity map, the main (head) entity for each page—or for an entire topical hub—must be identified. Supporting entities can then be represented using schema properties such as "hasPart" or "mentions".
Consider this example for a tile installation company:
The URL https://example.com/tile-services/ represents the main entity for the page. Because the @id value matches the page’s URL, it serves as a stable identifier and makes it easy to recognize this page as the topical hub. Fan-out queries that belong on this page are those that fall under the same main entity. A real tile-services provider would likely offer installation, restoration, and repair—each of which represents a subtype of “tile services.”
For example, the installation page can reference the main entity using hasPart:
And on the https://example.com/tile-services/install/ page.
Together, these two pages clearly signal that the main topic is Tile Services, with installation identified as a supporting sub-entity. Both pages may be relevant for similar search queries; however, when a query is broadly about tile services, the hub page should surface, and when a query specifically targets tile installation, the installation page should appear instead. This type of entity mapping reduces cannibalization by explicitly defining how related pages connect and by clarifying their topical boundaries when overlapping subjects are present.
In Practice with layered AI
Layered AI evaluates the content of each page in context. Traditional SEO considerations, such as keyword density, should still be taken into account. The Tile Services page should provide a broad overview without being overly detailed about installation, while the Tile Installation page should focus on installation specifics and avoid repeating content from the Tile Services page. This ensures that each page maintains a distinct purpose and signals its unique entity to both users and AI-powered search systems.
It should also be noted that qualifiers are not entities. For example, creating a main page for “Tiles” and a subpage for “Red Tiles” can lead to cannibalization, because “red” is not a distinct entity—only the broader concept of tiles represents a meaningful topic.