This post is for the SEO lead, product owner or developer who is building or rescuing a doctor directory for a hospital, a hospital group or a referral platform. A directory is the part of a healthcare site where filtering and URLs collide. Patients want to filter by specialty, city, language, gender, availability and insurance. Each filter can generate a URL. Combine five filters across hundreds of doctors and you can produce millions of URLs, almost all of them useless to search engines and some of them harmful.
It answers a design question: which filter combinations should have their own crawlable, indexable URL, which should exist only for users, and what rules keep the two apart? The approach is deliberate URL design with a small set of indexable landing pages, canonical rules that match, and crawl control for everything else. It assumes you control the application code or can brief the developers who do. The examples are hypothetical and use no real clinicians.
The problem: faceted URL explosion
Google describes the issue in its guidance on managing crawling of faceted navigation: crawlers may access a very large number of faceted URLs before deciding they are not useful, which slows discovery of the new and important pages. For a hospital, the cost is visible: new doctor profiles take longer to be found, and crawl effort goes to pages like "female, Tamil-speaking, available today, accepts insurer X, sorted by experience, page 4".
The same guidance sets out the toolbox. If faceted URLs do not need to appear in search, block crawling with robots.txt, and allow a single listing page with no filters. If filtering uses URL fragments (after #), it has no effect on crawling either way. The nofollow attribute is weaker because every link to a filtered URL must carry it. Canonical tags pointing filtered URLs at the unfiltered page may over time reduce crawling of the non-canonical versions but are less effective than robots.txt. If you do want faceted URLs crawled and indexed, use the standard & to separate parameters rather than commas, semicolons or brackets, keep filter order consistent for path-based filters and avoid duplicate filters. Return an HTTP 404 for combinations that produce no results.
We use those rules as the backbone of the design below.
Decide the indexable set first
Start from demand, not from the filter list. A directory has three kinds of page:
- Profile pages: one per doctor. These are indexable. Each doctor has one canonical profile, and listings link to it.
- Landing listings: combinations that match a real search intent and have enough doctors to be useful. These are indexable.
- Filter states: everything else. Users can use them; search engines should not spend time on them.
Which combinations deserve landing pages? Use evidence from Search Console and keyword research. For each candidate (specialty, specialty plus city, specialty plus city plus language), check whether people search the phrase and whether you have enough listings to give a helpful page. A rule of thumb that you should tune: a combination becomes indexable when it has demand and at least a minimum number of doctors, say five, with real, published profiles. Below that, the page is thin and better served by the parent listing.
A typical resulting set:
| Combination | Indexable? | Example URL |
|---|---|---|
| All doctors | Yes, as the directory home | /doctors |
| Specialty | Yes | /doctors/cardiology |
| Specialty and city | Yes, if enough doctors | /doctors/cardiology/pune |
| Specialty, city, language | Only for high-demand languages with enough doctors | /doctors/cardiology/pune/marathi-speaking |
| City only | Optional, if a general city page makes sense | /doctors/in/pune |
| Gender, availability, insurance, sort, page size | No | filter state only |
"Language" needs care. Here it means languages a doctor speaks with patients, not the language of the website. A site that publishes pages in several languages is a separate multilingual setup with its own hreflang considerations; see our multilingual SEO service. Do not conflate the two when naming URL segments.
URL structure rules
Use paths for indexable combinations and query parameters for filter states. Path segments signal a stable landing page; parameters signal a state. This makes it easy to write rules: anything with a query string is a filter state.
Fix the segment order. /doctors/{specialty}/{city}/{language} always, never /doctors/{city}/{specialty}. Google's guidance asks for consistent filter order in path-based filters. Redirect any other order to the canonical order with a 301.
Use lowercase, hyphenated slugs from controlled vocabularies. Specialties, cities and languages come from reference lists, not free text. Keep slug aliases (for example "ent" and "ear-nose-throat") redirecting to one canonical slug.
Use standard parameter syntax for states. ?available=today&gender=female&sort=experience, with & separators and = between key and value. No comma-joined values, no brackets.
Do not allow duplicate or contradictory filters. ?gender=female&gender=male should return a clean response, either ignoring the duplicate or returning 404, and never create a new indexable URL.
Canonical rules
Canonicals should reinforce the design, not substitute for it.
| URL type | Canonical |
|---|---|
| Landing listing, first page | Self-referencing, absolute URL |
| Landing listing with filter parameters | The landing listing without parameters |
| Paginated landing listing, page 2 and beyond | Self-referencing (each page lists different doctors) |
| Non-qualifying combination (too few doctors) | Redirect or canonical to the parent listing |
| Profile page | Self-referencing at /doctors/{slug} |
Google's guidance on consolidating duplicate URLs calls rel="canonical" a strong signal and redirects a strong signal as well, advises absolute paths, and notes that Google makes the final choice by content similarity. So a canonical on a filter state is a request. If the filter state differs a lot from the landing page (it shows a different set of doctors), Google may ignore it. This is one reason to also stop crawling of filter states.
Do not point the canonical of page 2 at page 1. Page 2 lists different doctors; a canonical to page 1 hides those profiles from the crawl path. Give each page of a paginated listing its own canonical, a distinct title suffix and normal crawlable links to the next page.
Crawl control
Apply layers, in order of strength:
- Do not generate links you do not want followed. Filters should be implemented as buttons that update state, or as links with
rel="nofollow"if they must be links. Only landing listings should have plain<a href>links in navigation and facets. - robots.txt disallow for filter parameters. Disallow patterns such as the query parameters used for availability, gender, insurance and sort. Per Google's guidance, there is often no good reason to allow crawling of filtered items.
- Canonicals on any filter URL that still gets through.
- Noindex only for pages that must be crawled but not indexed. Remember a page blocked in robots.txt cannot be crawled, so a noindex on it will not be seen. Pick one mechanism per URL class.
Test the robots.txt rules before release with a crawler or the testing tools in Search Console, and re-test after every change to the parameter list.




