Topical Map: Build a Cluster Plan Search and AI Can Read, and Know What Topical Authority Is Not

A topical map is a planning document that lists the topics a site intends to cover, the specific questions inside each topic, the one page that will answer each distinct intent, and the links that connect those pages. It organizes editorial work. No search engine reads it, and nothing about having one makes a site rank.

Topical authority is the idea that a site that covers a subject thoroughly earns more trust on that subject. The idea is plausible and widely repeated. It is also an industry construct. We checked four Google documents on 7 October 2026: the SEO Starter Guide, the guide to Search ranking systems, the guidance on creating helpful, reliable, people-first content, and the page on AI features and your website. None uses the phrase topical authority. The nearest language is a question in the helpful-content guidance: does your site have a primary purpose or focus? The same page asks whether content provides a substantial, complete or comprehensive description of a topic, and whether it offers original information. Those are quality questions for a page and a site. They are not a score.

Google has also said, in a 2016 hangout reported by Search Engine Roundtable, that it does not have anything like a website authority score. In the same report the author notes that Mueller had said earlier in the year that Google does look at a website overall for some things. Both statements are consistent with this reading: sitewide signals may exist, and no single authority number is published. HubSpot's current guide to topic clusters goes further and asserts that clusters build topical authority by design and that answer engines choose sources by how authoritatively a site covers a subject. We found no independent evidence behind that assertion in the guide, so we treat it as a vendor claim. Plan clusters because they make a site easier for people to use and easier for you to maintain. Do not promise a ranking effect to anyone.

The three documents get confused because they share rows. They answer different questions, and each one fails when asked to do the others' job. A keyword list tells you what people type. A calendar tells you who writes what and when. The map sits between them and decides which pages should exist at all.

Start with the decisions your buyers make, not with the product. A buyer choosing a maintenance contractor weighs scope, price structure, response time and risk. Each of those is a cluster seed. Keyword tools can enlarge each seed afterward, but they cannot tell you which decisions matter. Sales calls can.

Next, collect the questions around each seed from sources you own: Search Console queries for the site, the internal search box, support tickets, and the questions reps hear. Our posts on keyword clustering and on query fan-out and content architecture cover the grouping step in more detail. The rule that matters here is that one distinct intent gets one page. If two queries would be satisfied by the same answer, they share a page. If you give each its own page, you are building the cannibalization problem we describe in our keyword cannibalization post.

Then assign a role to each page. A hub page frames the decision and links to the pages that answer its parts. A supporting page answers one question in full. A comparison page helps someone choose between options. A reference page holds a definition or a table that other pages cite. Write the role down, because it tells the writer what the page must not do. A supporting page that tries to be a hub becomes an unfocused page.

Finally, set the links before the writing, and set the priority. Which cluster comes first? The one closest to revenue, where your own evidence is strongest, and where you can add something that existing pages lack. The map is finished when you can say, for any page on it, who it is for, what question it answers, what it links to and who owns it.

The map below is illustrative. It describes an invented commercial heating and cooling maintenance contractor serving one metro area, built to show the shape of the document. It is not a client, it reflects no real search data, and the page titles are examples of intents, not recommendations to publish those exact pages.

Mostly with ordinary work. Google's documentation says the best practices for SEO remain relevant for its AI features, and that there are no additional requirements to appear in AI Overviews or AI Mode. It also says those features may use a technique called query fan-out, which issues multiple related searches across subtopics to build a response. A cluster that answers the sub-questions of a decision in separate, focused pages gives that kind of retrieval more to match. It is a reasonable design rationale. It is not a guaranteed effect, and we know of no published test that proves it.

Links come first. Google's link guidance says it uses links as a signal of relevance and to find new pages, that every page you care about should be linked from at least one other page on your site, and that good anchor text is descriptive, reasonably concise and relevant to both pages. Crawlable links are plain anchor elements with an href. A hub page that links to its supporting pages through a script-only menu has not linked them in the sense that matters. Our internal link governance post covers keeping this honest as a site grows.

Organization comes second. Google's SEO Starter Guide says organizing a site logically can help search engines and users understand how pages relate. For sites with more than a few thousand URLs, it adds that grouping similar topics in directories can help Google learn how often URLs in each directory change. For a smaller site, a tidy URL structure is a convenience rather than a lever.

Page design is third. Each page should answer one question in its first paragraph, use a heading that states the question, and put comparisons in real tables. If you use structured data, the schema.org about property describes the subject matter of a page, and Google's structured data guidelines require that markup describe content visible to readers. Use it to state plainly what the page already says. Markup does not prove a page covers a topic.

A map earns its keep by what it leaves out. Google's helpful-content guidance asks whether content provides original information, reporting, research or analysis, and whether it goes beyond the obvious. A page that restates what ten others say, written only to fill a cell in a cluster, fails that test and dilutes the pages around it. The decline list is where the map says no.

Four tests help. Does this intent differ from another page's? Do we have information of our own to add, such as first-party data, a process, a real example or an opinion we can defend? Can someone own and update the page? Does it help a person decide something? A page that fails any of the four stays on the decline list until that changes. Declining a page also keeps the corpus honest. Programmatic or templated variations across places and services read as doorway pages when each adds nothing, which is a separate trap our doorway page audit post describes.

Measure the cluster as a set. Check indexing for each page first, because a page that is not indexed cannot do anything. Then watch impressions and clicks for the cluster's queries in Search Console, reading the cluster's branded and non-branded queries separately. Google notes that AI Overviews and AI Mode traffic is included in the Search Console performance report under the Web search type, so a cluster's numbers may include it. Then count what matters to the business: conversations started from cluster pages, and which pages the sales team actually send to prospects.

Set expectations in advance. Write down the date you will review, the queries you expect to move, and what would make you merge, rewrite or retire a page. A cluster that has not moved after a fair period is information about the pages or the intent, not about authority. If a page ranks for something other than its intent, fix the page or the map. Do not report a score called authority, because no one can verify one. The honest summary is a table: pages published, pages indexed, queries gained, conversations started, decisions made.