Brand promotion: this article introduces Guangsuan's on-site technical SEO and keyword mapping services. It is a vendor introduction, not an independent review or an editorial endorsement.
Short answer, with its conditions: a product page answers whether a capability fits your situation, a documentation page answers how to complete a task, and a community page answers how that task works in practice for people like you. Give each buyer question to the page that already holds the facts needed to answer it, and repair that page before you plan another URL. When two pages answer the same question, decide with the reader's next action and the evidence you can supply, not by how often the word appears. Nothing here promises a ranking position, an indexing date or an enquiry.
Three page types, three different questions
For a product with a public site, a documentation area and a customer community, the page-role question splits three ways. The mistake is to treat those three areas as one body of text to be keyworded, because they then start answering the same question in three slightly different voices.
| Reader's question | Page that should own it | What that page must carry | Next action it should end with |
|---|---|---|---|
| Does this capability do what we need, and does it fit the way we work? | Product page | What the capability does, who it is for, the surfaces and regions it supports, its documented limits, which plan carries it, and a route to pricing or a trial | Evaluate it, start a trial, or look elsewhere |
| How do I set this up, and what will stop me? | Documentation page | Prerequisites, permissions and plan requirements, ordered steps, error conditions, and the version or scope the steps apply to | Complete the task inside the product |
| How do people like us actually do this, and what happens at the edges? | Community page | Peer answers, real workflows, edge cases and workarounds, plus a clear route to the official instruction when the answer is version-specific | Solve a problem in context and continue working |
The third row needs the most care, because a community page and a documentation page can drift toward each other. The dividing line is authority and stability. A documentation page states what the product does in the version the reader has. A community page carries what people found when they used it, including workarounds and the cases the documentation does not mention. Google's helpful-content guidance counts forum discussions, customer reviews, comments and member-uploaded media as main content when they directly serve the page's purpose, so peer answers can carry a page properly — as long as the page answers something a reader actually asked.
Two rules follow from that. When a community answer becomes official and version-bound, it belongs in the documentation, and the thread should point there. When a documentation answer starts depending on the reader's context — their volume, their workflow, their region — the context-specific part belongs in the community. If your community lives on a separate domain or subdomain, the crawl and index rules for that area are a separate decision from the page-role assignment; check the rules that actually apply to that path rather than assuming the main site's rules carry over.
Write the page's sentence before you look for a keyword
Before any keyword work, give every URL one sentence: this page is where someone asking [question] finds [answer], and the next step is [what]. If you cannot write that sentence, the page has no job yet, and a traffic budget will not give it one. This costs nothing, and it is the fastest way to see which URLs are duplicates before any tool does.
The published keyword mapping guide behind this approach makes the same point in a different way: the final assignment follows search intent and what the business can actually provide, and a keyword group should organise the answer rather than prescribe repeated phrases. The same guide notes that the decision for a candidate may be to update an existing page, to create a distinct answer, to go and gather missing information, or to defer the candidate. Keep established addresses stable, and give genuinely new pages readable, durable addresses rather than moving an old one to insert another word. The method is written out in full in the keyword mapping guide.
Assigning queries to pages without a paid keyword tool
You do not need a subscription to run the assignment. Three inputs carry most of the work.
- Your own material. Support questions, product records, on-site search strings, sales enquiries and the pages you already publish. Support and sales teams use different words than marketing does, and that difference is often the first sign that a page is missing.
- Your own Search Console rows. Export the relevant queries and pages from the Performance report and keep the dates, country, device and search type beside the export. Read that export as what your property was shown for in that window, not as total market demand: anonymised and truncated rows mean the query table is incomplete, so a query you cannot see is unknown rather than zero.
- Representative search results. Open a few pages that already serve the same task and read what they actually answer. Search competitors for a query need not be the companies you sell against.
If you do use a keyword tool, record the tool, the country database, the collection date and the metric name next to each figure. Volume estimates, difficulty scores, advertising competition and the clicks your own site receives measure different things, and none of them is a forecast for your site. They organise the research; they do not decide the assignment.
Then apply one grouping test: could one finished answer let both groups of searchers finish without searching again? Shared wording is a hint. The reader's next step and the proof needed to get there usually settle it. Split when the evidence or the follow-up differs, and keep the split pages linked with ordinary crawlable links and descriptive anchor text. This is where keyword intent mapping becomes concrete: an intent label earns its place only when it changes which page holds the answer, and the label comes from the reader's next step rather than from the wording of the query.
The result is one working sheet rather than a keyword list. Each row in the guide's format records the intended reader, main task, query group, market, data source and date, destination URL, questions, required evidence, owner and decision.
| Reader question (hypothetical example) | Next action | Page that owns it | Evidence on hand | Evidence missing | Action |
|---|---|---|---|---|---|
| Can members move between our chat tools without losing history? | Confirm it fits | Product page | The sync behaviour described on the product overview | Which plan carries it, and what it means at our member volume | Add plan and volume conditions to the product page |
| Our workspace shows the same member twice after sync — what do we check first? | Complete a task | Documentation page | The existing troubleshooting steps | The product version those steps were written for | Add version scope, and link to it from the product page |
| Has anyone run this with a community our size, and what did they change? | Decide whether to proceed | Community page | Peer discussion | A repeatable recommendation, and the official page that states it | Keep the thread as peer context and link the official page |
Those three rows are a teaching example for an invented community operations product. They carry no traffic, ranking or conversion figures. The point of the sheet is in the last two columns: each row names the page to fix and what to ask for next, which is often a fact sitting in a file nobody has published.
Two pages, one query: rewrite, split, or merge
Two URLs appearing for one query is a signal to examine, not a diagnosis. Published guidance on keyword overlap is explicit that one query producing impressions for multiple URLs does not by itself establish cannibalisation. Before changing anything, fix the conditions: one date range, one search type, one country, one device, and keep them with the export. Then check which URL actually earns the impressions, and for which query. Much of Search Console's reporting rolls up to the canonical URL, so the page list is not a list of every address that can appear.
Then decide which of three situations you are in.
- Different tasks, overlapping wording. Keep both pages. State each page's purpose in its title and opening, remove the paragraph that takes over the other page's answer, and add explicit cross-links so a reader can move from one to the other. Wording overlap alone does not put two pages on the same job.
- Same task, and the intended page is missing the proof. Strengthen the intended page: the specifications, limits, conditions or links needed to finish the task. This is a rewrite, not a new URL.
- Same task, near-duplicate content. Choose the survivor by information value and by who can maintain it. Migrate the answers that are still valid, set a permanent redirect from the old address to the equivalent content, and update the internal links and sitemap entries that still point at the old URL.
Two cautions while you do this. A canonical annotation expresses a preference for duplicate or very similar URLs; it is not an instruction to rank a chosen destination, and pointing every page at the home page is not a solution. Google's own URL guidance asks you to use the same URL in internal links, in the sitemap file and in the canonical tag, so keep those signals aligned instead of using canonical to tidy up a page-role decision. And do not retire a page only because it has had no clicks for a while, and do not use noindex or a robots rule to avoid a content judgement you have not finished.
Check the assignment against what the site actually shows
Open each page as a visitor who is not signed in. Check the response status, the main content, the links, the images, the metadata, and any login wall or error you did not intend. A URL that returns a login screen has not delivered a page.
Then, in Search Console, filter to the query group with stable filters and compare the pages receiving impressions against the URL you intended. If an unexpected page appears, three hypotheses are worth testing before you touch anything: it answers the task better than the page you meant to promote; your intended page lacks the information, links or access needed to compete; or your intended page is hard to reach, because of a robots rule for that path and user agent, a canonical pointing elsewhere, a redirect chain, or a missing sitemap entry. The third case is technical, and it needs a technical check rather than a copy change.
Two status labels need care, because they are records rather than diagnoses. A page reported as crawled but not indexed tells you what the search engine saw, not that the content is weak. A page reported as discovered but not crawled does not by itself mean you need a paid crawl service: read the robots rule that matches that specific path and user agent, the canonical target, the redirect hops and the sitemap entry first. Also keep in mind that a noindex directive is only visible to a crawler that can fetch the page, so a page blocked from crawling may never have its index control read at all.
Record the date of every change and any parallel work, and keep impressions, clicks, indexing status and qualified enquiries as four separate records. One of them moving does not prove the others moved.
Where Guangsuan fits
If your sheet shows unclear page roles, thin answers or a technical block you cannot resolve, that part can be handed over. The on-site technical optimisation engagement described on the Guangsuan on-site technical optimisation service page treats crawl and indexing as one group — access status, robots, noindex, canonical, XML sitemaps, redirects and dead links — and separates page and content work (matching query intent to page topic, titles, descriptions, heading levels, the order of existing content, image alt text and any necessary structured data) from link and access work (categories, breadcrumbs, related content, internal links, mobile reading and key image loading).
Within that scope, key pages are handled one by one on topic, title, existing content and internal links, which is not the same as rewriting every page from scratch; the remaining pages in scope receive template and rule adjustments with sampled re-checks. The handover is meant to be usable by whoever inherits the site: the optimised URL and key-page list, the problem baseline from before the change, change records for titles, body copy, internal links and configuration, backup and rollback notes, the re-check results, and the unfinished items with maintenance advice. A revisit and correction of the same implementation is included within 30 days.
The boundaries are stated on the same page. The engagement does not include new articles, a whole-site rebuild, server migration, link building or ongoing ranking operations, and Guangsuan does not promise a fixed ranking, a number of indexed pages or a number of enquiries. Final scope, fee and period are settled through the project list and the contract.
What to bring to the first conversation: the URL list with a page type and one-sentence purpose for each, the query and page export with its date range, country, device and search type recorded, the single fact each key page has to carry, who can approve an edit, the routes that must keep working, and the one conversion you care about together with how you will confirm it. If you want a second opinion on the mapping itself, a technical team can take the diagnosis and the verification as one engagement. The decision about which page should carry which question still depends on what you sell and the facts you can prove.
Frequently asked questions
Does every keyword need its own page?
No. Keep wordings together when one finished answer closes the task, and add a page only when the task, the required evidence and the follow-up all differ, and you hold the material to answer it completely. A page created to hold a phrase, with no task behind it, is a thin page that competes with other thin pages.
My product page and my documentation both mention the same feature. Is that cannibalisation?
Not on wording alone. If the reader's next step and the proof you need are different, the two pages have different jobs and can both stay, with clear links between them. Decide the assignment with the task and the evidence in hand rather than with a similarity score.
Two of my pages are showing for the same question. Should I merge them?
Not automatically. Check which URL actually earns the impressions, and for which query, with one date range, one search type, one country and one device. If the tasks differ, keep both and clarify the links. If the tasks are the same, pick the page that holds the proof, move what is still valid, redirect the old address and repoint the internal links.
Is a community page worth building for search?
It depends on what it carries. A page that restates the same policy line across every category is a convenience for members, not a search destination. A community page that answers a real question with the conditions attached is content, and Google's helpful-content guidance treats forum discussions, customer reviews, comments and member-uploaded media as main content when they serve the page's purpose directly. Decide the crawl and index handling for each area of the community by the rules that apply to that path, and keep threads with real answers apart from the empty archive.
Do I need a paid keyword tool to run this?
No. Support questions, on-site searches, your existing pages and your own Search Console rows are enough to do the assignment. A tool adds demand estimates and a view of competing pages; it is not the input that tells you which page should answer which question.
If you do only one thing: write the purpose sentence for every URL, open the ten that matter most as a signed-out visitor, and fix the one whose sentence is wrong or whose page cannot be fetched. Reachability and the job come before copy, and a rewrite on a page nobody can fetch produces nothing you can measure.