Internal Linking SEO: One Category Tree, Links to What Sells
Help Google reach the pages that bring revenue: an example shop moves its best-selling plate from six clicks to one and uses the same branch in its feed.
Put every page you want indexed on one tree of plain HTML links: menu, category, subcategory, product. Google can infer a page’s importance from how many internal links point to it and how many clicks away it sits. So give the pages that bring revenue extra links from the home page, the top of category pages and related-product blocks, and refresh them monthly because bestsellers change. Write the same tree into your feed’s product type, so the site, Merchant Center and your campaigns share it.
How does Google read the structure of an online store?
Google works out how a store is organised mainly from its links. Links are also how Google finds the vast majority of new pages, according to its SEO Starter Guide.
Google’s ecommerce guide gives two examples of what Google can use to judge a page’s importance relative to the rest of the site (Help Google understand your ecommerce website structure). One is the number of links Google must follow to reach the page. The other is the number of links pointing to it. The same page says Google generally doesn’t look at URL structure to work out site structure.
That makes internal linking, the links between pages of your own site, one of the most direct ways to tell Google which pages matter. It is one part of ecommerce SEO, and the part where your sales data helps most.
Three rules from Google’s documentation decide whether Google can reach your pages through links at all:
- Use
<a href>links. Google can generally crawl a link only if it is an<a>element with anhrefattribute (Link best practices for Google). Google can’t reliably extract a URL from a menu or product tile that relies on a script event to open the page. - Make products reachable without site search or “load more”. The ecommerce guide says Googlebot generally doesn’t submit searches into a search box. Google’s crawlers also don’t click buttons and generally don’t trigger JavaScript that needs a user action (Pagination, incremental page loading, and their impact on Google Search).
- Link to every product you want indexed. Where that isn’t possible, Google suggests a sitemap or a Google Merchant Center feed, because both can list pages a crawler would otherwise miss.
The last rule connects SEO and ads. The feed you send to Merchant Center for Shopping ads is also one of the sources Google names for finding product pages that your links miss.
What should the category tree look like?
Google’s ecommerce guide recommends one path for links: from menus to category pages, from category pages to subcategory pages, and from subcategory pages to all product pages. A 2008 post that Google still marks as relevant states the depth rule plainly: “Keep important pages within several clicks from the home page” (Importance of link architecture). Google gives no exact number. Count the clicks to your best-selling pages yourself, and if one sits much deeper than its neighbours, give it a link from higher up.
Every link on a store page sits in one of a few places. Each place does a different job, and each tends to fail in its own way:
| Where the link sits | What it gives Google | What it gives the shopper | Where it usually goes wrong |
|---|---|---|---|
| Main menu | top categories one click from every page | the range at a glance | menu built without <a href> links |
| Category and subcategory grids | a path to every product | browsing and comparing | products reachable only through filters or “load more” |
| Breadcrumbs | a link up the branch from every page | where they are, one level up at a time | trail copies the URL instead of the shopper’s path |
| Pagination | page 2, 3 and onwards of long grids | shorter pages | page numbers kept after a # |
| Home page blocks | extra links to the pages you choose | shortcuts to popular products | set once and never updated |
| Related products | sideways links between products | alternatives and add-ons | filled with random or newest products |
| Buying guides and blog posts | links inside text, with descriptive anchor text | answers before the purchase | links to filtered URLs instead of the category |
Google’s documentation backs the technical part of this table. Its breadcrumb guide recommends a trail that represents a typical user path to a page, not a copy of the URL (Breadcrumb (BreadcrumbList) structured data). Its pagination guide asks for an <a href> link from each page to the next one. It also warns that Google ignores the part of a URL after #, so it may not follow a next-page link that differs only after the #.
Filters and sort options, known as faceted navigation, need a decision of their own: they can produce a huge number of URL combinations. Google warns that crawling them wastes resources and slows down how quickly it finds new, useful pages (Managing crawling of faceted navigation URLs). Which filter pages to keep in the index is the subject of faceted navigation SEO. Category page SEO covers what a category page itself needs to rank.
One more rule matters when a product sits in several categories. Link to its canonical URL every time, not to one of its duplicate addresses. Google says consistent links to the canonical version help it understand which URL you prefer (How to specify a canonical URL with rel=“canonical” and other methods).
Can one category tree serve the site, the feed and the campaigns?
Yes, and the feed already has a field for it. The product type attribute holds your own categories. The Google product category attribute is different: it takes values from Google’s predefined list (Product type [product_type], Merchant Center Help).
Google’s formatting advice for product type reads like a description of site navigation. It says to separate levels with ” > ”, include every level and use the full breadcrumb path, for example Home > Women > Dresses > Maxi Dresses.
In campaigns, product type does two jobs:
- Standard Shopping can organise bidding and reporting by it. You can submit up to 5 values, but Google uses only the first one for that. So if you organise campaigns this way, Google recommends a single value.
- Performance Max accepts product type as one of the dimensions for listing groups. Listing groups decide which products each asset group includes or excludes. Unlike in Standard Shopping, they don’t set bids (Listing groups for retail, Google Ads API).
Here is one branch of a tableware shop, written once and used in five places.
Example store, not client data.
| Where | How the branch appears |
|---|---|
| Menu and breadcrumbs | Home › Tableware › Plates › Dinner plates |
| Category URL | /tableware/plates/dinner-plates |
| Feed, product type | Tableware > Plates > Dinner plates |
| Performance Max listing groups | split by product type at the “Plates” level |
| Reports | ad spend and revenue for the “Plates” branch |
The payoff is a shared vocabulary. Your analytics show organic revenue per category page, and your ad reports show spend and revenue per product type. With one tree, both line up row by row. You can then compare each branch’s share of the ad budget with its share of revenue and see where the two disagree.
The portal’s Assortment section lines up the biggest product groups by budget share and shows each group’s return next to it. You can group products by product type, category or brand.
The trees can differ in three ways:
- A product can sit in several site categories, such as “Plates” and “Gifts”. The feed takes one main branch, because Google uses only the first product type value for bidding and reporting.
- Campaigns often need a second split that ignores categories: how each product performs. Custom labels handle that split, as shown in labelling products by performance.
- Promotional pages, such as a gift guide, live on the site only.
Product type in Google Shopping covers how many levels to use and what product type controls that Google product category doesn’t. How taxonomy affects ad performance explains why accurate categories in the feed matter for ads in the first place.
Sales sit in a small part of the catalogue, and that part keeps moving
Our data comes from Google Ads accounts, not from organic search. It shows product by product where shoppers’ money goes, and the split is lopsided. These are observations, not an experiment.
- Few products sell. In our study of 1.4 million products (1,404,808 products, 130+ stores, 13 months), 7.9% of products sold at least once, and they took 78.1% of ad spend. The top 10 products brought in an average of 31.3% of revenue.
- The top changes. In the same study, 71% of bestsellers stayed in the top for 3 months or fewer out of 13. The median stable winner, a product with sales in at least two months, lasted 3 months.
- Products with the most orders change slowest. In the median store-month, 12.5% of products with 6 or more orders that month had no sales the month before. Among products with 1–2 orders that month, the share was 75.0% (GetProfit data, 116 stores, June 2025 – June 2026).
For an internal linking strategy, this gives a simple split. Categories and the steady head of the catalogue can have permanent links. Current bestsellers need slots that you refill, because a link you place once points at last quarter’s bestseller within months. The long tail renews too fast for hand-placed links, so let category grids and pagination carry it.
A product can sell well through Shopping ads while its page ranks for nothing, and the reverse. Use these numbers to decide how often to review your links, and your own landing-page report to decide which pages get them.
Which pages should get the extra links?
Google’s ecommerce guide states the principle. If you have a best-selling product, consider linking to it from the home page or from other content, such as blog posts or newsletters. The work is in deciding which pages count as best-selling, and in keeping that list current.
Build the list from two sources:
- Organic revenue by landing page. In your analytics, export revenue from organic search by landing page for the last 12 months. Sort the pages by revenue and keep a running total of their share. The pages that make up the first 80% are your organic revenue core. Mark each one as a category page or a product page.
- Products that sell in ads. They show demand even before a page ranks. In the portal, two of the eight labels form this list. Bestsellers sell in at least 6 months out of 12 and hold the target ROAS. Strong products sell well and hold ROAS, but their sales history is still short. Labels come from each product’s behaviour in ads over 12 months, and any group exports to CSV (Products and labels).
Then give each kind of page its own kind of link:
| Page | Where its extra links come from | How often to review |
|---|---|---|
| Categories in the organic revenue core | main menu, home page, buying guides | when the range changes |
| Steady sellers (Bestsellers) | top of their category grid, a home page block, related-product blocks on similar products | monthly |
| Rising products (Strong) | top slots in their category, related-product blocks | monthly |
| Long tail | category grid and pagination only | nothing to refresh |
| Out of stock or discontinued | take them out of home page and related-product blocks | when stock changes |
How should the links themselves be written?
The anchor text, the clickable words of a link, tells shoppers and Google something about the page it leads to, as the SEO Starter Guide puts it. Google’s link guide asks for anchor text that is descriptive, reasonably concise and relevant to both pages. Its test: read the anchor text on its own and check that it still makes sense.
“Dinner plates” passes that test for a category link. “Click here” or “Shop now” makes sense only to someone who can already see the product.
The same guide warns against cramming every related keyword into the anchor text. It also advises against chaining links next to each other, because each link loses the text around it. Product tiles need one extra check: for a linked image, Google uses the image’s alt attribute as anchor text, so describe the product there.
Step by step: build the tree and point links at revenue
- Find your organic revenue core. Export organic revenue by landing page for 12 months and mark the pages that make up 80% of it. Check: every page on the list is marked as a category or a product.
- Write down the tree. Start from your current menu: categories, subcategories, products. Check: every product has one main branch.
- Copy the main branch into product type in your feed, with ” > ” between levels and one value per product. Check: open a few products in Merchant Center and compare their product type with the breadcrumb on the product page.
- Make every core page reachable by
<a href>links from its category, without site search, filters or “load more”. Check: crawl the site from the home page with an SEO crawler and note how many clicks each core page sits from the home page. - Add the extra links. The home page links to core categories and current bestsellers. Each category links to its bestsellers at the top of the grid. Each product links to related products from the same subcategory. Buying guides link to categories. Check: every current bestseller has at least one link from outside its category grid.
- Check what Google sees. In Search Console, open the Links report and the table of top internally linked pages. Your core pages should be near the top. The report shows a sample of links, and its tables stop at 1,000 rows (Links report, Search Console Help).
- Refresh the product slots every month. Swap out products that have left the Bestsellers and Strong lists. Review categories when your range changes. Check: no home page or related-product slot points to a product that is out of stock or has left both lists.
Example: moving a bestseller from six clicks to one
Example store, not client data.
A tableware shop with 3,000 products sorts its category grids newest first, 24 products per page, with a “next page” link at the bottom. Its best-selling dinner plate is the 60th product in “Dinner plates”, so it sits on page 3 of the grid. From the home page the path is Tableware, Plates, Dinner plates, page 2, page 3, then the product: 6 clicks. Only one page on the site links to it: page 3 of the grid.
The shop makes three changes. It adds a block of eight current bestsellers to the home page, a row of four bestsellers to the top of “Plates”, and a related-product block to three similar plates. The plate is now 1 click from the home page through the block, and 3 clicks through the tree: Tableware, Plates, then the product. Six pages link to it instead of one: the grid page, the home page, the “Plates” category and three similar plates.
Its product type in the feed reads Tableware > Plates > Dinner plates, the same branch as its breadcrumb. When the plate drops out of the Bestsellers list, it goes back to its place in the grid, and the home page slot goes to the next bestseller.
See how your catalogue is labelled. Sign in with Google in one click. The portal changes nothing without your consent.
Sources
- Help Google understand your ecommerce website structure — Google Search Central: importance inferred from links to a page and clicks to reach it; URL structure generally not used to work out site structure; menu → category → subcategory → product links; Googlebot doesn’t use search boxes; sitemap or Merchant Center feed for unlinked pages; link best-selling products from the home page, blog posts or newsletters. Checked 2 October 2026.
- SEO Starter Guide — Google Search Central: the vast majority of new pages are found through links; link text tells users and Google about the target page. Checked 2 October 2026.
- Link best practices for Google — Google Search Central: crawlable
<a href>links; descriptive, concise anchors; no keyword cramming or chained links;alttext used as anchor text for linked images. Checked 2 October 2026. - Importance of link architecture — Google Search Central Blog, 7 October 2008, marked as still relevant: important pages within several clicks from the home page. Checked 2 October 2026.
- Pagination, incremental page loading, and their impact on Google Search — Google Search Central: crawlers don’t click buttons; link each page to the next with
<a href>; Google ignores URL fragments for page numbers. Checked 2 October 2026. - Breadcrumb (BreadcrumbList) structured data — Google Search Central: breadcrumbs should represent a typical user path, not mirror the URL. Checked 2 October 2026.
- Managing crawling of faceted navigation URLs — Google crawling documentation: filter URLs cause overcrawling and slower discovery of new URLs. Checked 2 October 2026.
- How to specify a canonical URL with rel=“canonical” and other methods — Google Search Central: link to the canonical URL within your site. Checked 2 October 2026.
- Product type [product_type] — Merchant Center Help: your own categorisation, unlike Google’s predefined categories; ” > ” between levels, all levels, full breadcrumb; up to 5 values, only the first used for bidding and reporting in Shopping campaigns. Checked 2 October 2026.
- Listing groups for retail — Google Ads API: product type among Performance Max listing group dimensions; listing groups include or exclude products and don’t set bids. Checked 2 October 2026.
- Links report — Search Console Help: top internally linked pages; a sample of links; tables limited to 1,000 rows. Checked 2 October 2026.
- GetProfit study of 1,404,808 products, 130+ stores, 13 months — share of products that sold and their share of spend; revenue share of the top 10 products; how long bestsellers stay in the top.
- GetProfit data: 116 stores, June 2025 – June 2026 — month-to-month renewal of products with 6+ orders and with 1–2 orders a month.
SEO and PPC: Are Your Shopping Ads Paying for Organic Sales?
See when Shopping and Performance Max ads pay for sales organic search would bring anyway, and how to check a category's organic orders before you cut it.
Ecommerce SEO for Stores That Already Run Shopping Ads
Find which store pages earn organic revenue, what to fix first, and how one product record feeds both your Shopping ads and Google's unpaid results.
Selling on Marketplaces or Your Own Store: When Each Pays
See which channel brings the cheaper order at your stage, using what Google Ads costs per 100 of ad revenue in 114 stores grouped by ad revenue.
Google Free Listings: How to Get Products Shown for Free
Get your products into Google's unpaid results: what Merchant Center needs first, and how one store lost free clicks as Google showed fewer products.
ChatGPT Shopping: How Products Get In and What Data to Send
See how ChatGPT picks the products it shows, which nine fields OpenAI's product feed requires, and how they map to your Merchant Center feed.
Ecommerce Advertising Channels to Add After Google Shopping
See which paid channel to try next, in an order set by what each reuses from your setup. Demand Gen, for one, needs no new feed, tag or account.