Discussion
  • Entity SEO

Topical entity maps for brands: useful for planning content and PR, or consultant theatre? What did one actually change for you?

Published 1 reply

Discussion starter: an entity SEO question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.

Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.

1 reply

  • admin
    Admin
    Original poster

    Useful when it changes decisions, theatre when the graph is the deliverable. A topical entity map earns its cost when every node ends up attached to something concrete: a page to build, merge or retire, a PR angle, a spokesperson, or a brand description to correct, each with an owner and a way to measure it. If nobody can name a decision the map changed, it was theatre, however good the visualisation looked.

    The evidence supports the idea underneath entity maps more than the way they are often sold. Google has organised information around entities since it launched its Knowledge Graph in 2012, and it said its February 2026 Discover core update would show more in-depth, original content from sites with expertise in a topic, judged topic by topic. What no public evidence supports is optimising towards the scores many maps arrive with: salience values, entity density, or a tool’s topical authority grade. The map is a planning and coordination tool. It is not a model of how Google ranks.

    Three different products are sold as an “entity map”

    Brand entity hygiene

    Making the company, its products and its people unambiguous: one entity home (usually the About page), Organization markup, and consistent names, descriptions and profile links. This has the clearest mechanism, because it reduces the chance of being confused with something else. It matters most for shared or generic brand names, rebrands, mergers and multi-brand companies.

    Topical coverage map

    The concepts, problems, product attributes and questions the brand should be connected with, and how they relate. Its value is indirect: it shapes scope, page architecture and internal links. The map does nothing by itself; the pages built from it do.

    Association map for PR

    Which publications, communities and third-party pages discuss those topics, and whether the brand appears in that conversation. It has the loosest connection to rankings and the closest link to how journalists, reviewers and AI answers describe the brand, though no controlled public test exists that I can point to tying such a map to citations.

    Why the distinction matters

    If a map disappointed you, check whether you bought the second or third product while being told it carried the mechanism of the first. Each one should be judged against what it can plausibly change.

    What the evidence supports, and where it thins out

    Claim Evidence What it means for a map
    Search engines represent entities and need them disambiguated Well established: Google’s Knowledge Graph dates from 2012, and Organization markup plus consistent profiles are the standard disambiguation tools Hygiene work has a real mechanism
    Expertise is judged topic by topic Google statement, specific to the February 2026 Discover core update Supports narrowing scope to topics you can cover in depth; says nothing about graphs or scores
    Entity salience predicts rankings No public evidence I can point to; salience is the output of a separate Cloud product A drafting sanity check, not a KPI
    A tool’s topical authority score reflects how Google sees you A vendor metric built on the vendor’s own method Directional at best; do not report it as an outcome
    Manufactured co-occurrence of brand and topic lifts AI visibility Google’s generative AI guide says seeking inauthentic mentions is not helpful Genuine coverage is the only version worth planning

    On the AI side, Google’s May 2026 guide to its generative AI features says foundational SEO still applies, that you do not need special schema or AI-specific rewording, and that seeking inauthentic mentions is not helpful. That undercuts the version of entity mapping sold as “engineering co-occurrence for language models”. It does not undercut the PR version, where the goal is genuine coverage connecting the brand with topics it really works in.

    Salience deserves its own note because it appears in many entity audits. Google’s Cloud Natural Language API returns the entities it finds in a text with a salience value and metadata that can include a Knowledge Graph ID and a Wikipedia URL. Salience is a 0 to 1 estimate of how central an entity is to that particular text, according to that model. It is a reasonable check that a page meant to be about commercial solar financing reads as being about that. Google has not documented any use of it in Search, and pushing a page’s score from 0.4 to 0.6 by repeating a term is keyword density with a new label.

    Decisions a good map should change

    When a map earns its fee, the change shows up in the plan rather than in the diagram. These are the decisions worth expecting:

    • Scope. Topics to drop because the brand has no product connection or credible expertise, even where there is search volume. A keyword-led plan has no natural point at which to make this decision.
    • Page architecture. Pages that target different keywords but cover the same concept get merged into one page per concept or intent. That reduces cannibalisation and gives each concept one URL to earn links.
    • Internal links. Relationships in the map become linking rules: product to the problems it solves, problem to the guide that explains it, guide to the expert who wrote it.
    • Missing attribute content. The properties buyers compare on (materials, compatibility, certifications, service areas, pricing model) can sit in PDFs, sales decks or nowhere. Stating them plainly gives search engines, comparison writers and AI answers something accurate to use.
    • PR focus and spokespeople. Topics where the company has proprietary data or a credible expert become the PR focus. Topics where it has neither stop being pitched. Each spokesperson is tied to a few topics instead of commenting on everything.
    • One brand description. A short statement of what the company is, where it operates and who it serves, reused on the About page, profiles, press kit and author bios, so third parties stop paraphrasing it five different ways.

    Hypothetical example:

    A software company sells project management tools to construction firms. Its blog has 180 posts across about 60 topics, most of them generic productivity advice. The map is built from Search Console queries, sales call notes and support tickets rather than from Wikipedia. It shows buyers talking about requests for information, submittals, daily site reports and subcontractor compliance, which the site mentions only in passing. It also shows PR pitching “future of work” commentary that never mentions construction. The decisions that follow: merge about 40 generic posts into eight hub pages on construction workflows, publish feature pages for requests for information and submittals, retire the productivity category, and move the CEO’s commentary onto construction data from anonymised product usage. The map itself is a spreadsheet. Its value is those four decisions, and each can be measured.

    Signs you are looking at consultant theatre

    • The main deliverable is a network diagram with hundreds of nodes and no column for page, owner or action.
    • Success is reported as salience scores, entity density or a tool’s authority grade rather than visibility, coverage, leads or revenue.
    • The headline recommendation is adding long lists of about and mentions properties pointing at Wikipedia or Wikidata. They produce no rich result, and no controlled public test exists that I can point to showing a ranking effect.
    • The entities come only from Wikipedia, Wikidata or an NLP API, with no input from customers, sales or Search Console. That produces a map of the encyclopaedia, not of your market.
    • It promises knowledge panels or AI citations. Nothing guarantees either.
    • It looks like the map the same consultant made for a company in another industry.

    When a map is worth more than the effort

    Situation Likely value Why
    Small site, one product line, one team Low A topic list in a spreadsheet does the same job in a day
    Shared or generic brand name, rebrand, merger, several product brands High for hygiene Ambiguity costs clicks and produces wrong descriptions
    SEO, content, PR and product marketing planning separately High The map becomes a shared vocabulary and an ownership list
    Large library with overlapping pages Medium to high Consolidation decisions need a concept-level view
    Health, finance or other expertise-sensitive topics Medium to high Scope decisions tie each topic to someone qualified to write about it

    The coordination value is easy to underrate. In a company where four teams publish under one brand, the map’s biggest contribution can be agreeing who owns which topic. That stops the blog, the help centre and the PR agency from producing competing pages and messages on the same subject.

    How to test whether a map changed anything

    1. Write a hypothesis per cluster. For example: merging the construction-workflow posts into hubs and linking them from product pages will lift non-brand impressions and clicks for that cluster’s queries.
    2. Act on some clusters and hold others back. Change half of a set of comparable clusters and leave the rest as a comparison. It is not a controlled experiment, because clusters differ in demand and competition, but it separates the map’s effect from sitewide movement such as a core update.
    3. Measure at cluster level. Use regex filters in Search Console for each cluster’s queries and pages. Track non-brand impressions and clicks, the number of your URLs ranking for the same query (fewer after consolidation is a good sign), and, for PR clusters, the share of coverage that mentions the brand together with the topic. Google’s “Search generative AI performance” report adds impressions in its AI features by page, though not clicks.
    4. Time it to the mechanism. Architecture and linking changes can only show up after the affected URLs are recrawled, so confirm that in logs or URL Inspection before reading results. PR effects build as coverage accumulates, so give them two quarters and discount single-story spikes.
    5. Decide in advance what would change your mind. If the clusters you acted on do not beat the holdouts after recrawl, the map’s value was in hygiene and scope, and paying for periodic refreshes is hard to justify.

    My position: build the map as a spreadsheet, not a poster, and do not start until you can name three decisions it will inform in the next quarter. Judge it by those decisions and the cluster-level results. If you are buying one, ask for an anonymised action list from a previous engagement. A consultant who can only show you the diagram is showing you the theatre.

    Need help with this on your own site?

    Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.

    See SEO services

Add a reply

Replies are open to approved members. Log in or apply to join.