A paginated list buries everything after the second page. Padiham Prints sells 900 art prints from a shop in Lancashire, and until March 2026 the only route to most of them was a list of 45 pages, twenty prints each. 610 of 900 prints sat more than four clicks from the home page, and a crawl found that most of those had never been fetched at all.
The depths break down as 40 prints within two clicks, 250 at three or four, 400 at five to ten, and 210 further out than that, most of the last group having been fetched once in 2024 and not since.
Nothing was wrong with the pagination. It worked, it was numbered, a person could click through it, and it was the reason two thirds of the stock did not exist as far as search was concerned.
The markup is not the fix
Rel next and prev stopped being used in 2019, and a great deal of advice still recommends adding them. Adding them to Padiham's 45 pages would have taken an afternoon and changed nothing, because the problem was never that the pages were not declared as a sequence. It was that page 31 is thirty one clicks from the front door and a crawler stops long before that.
The fix is more entry points, not different pagination markup. Padiham built 33 category pages, by subject, by size and by printing method, each listing between 12 and 40 prints on one page with no pagination at all. After 33 category pages, 880 of the 900 sat within three clicks, and the list of 45 stayed exactly as it was for anybody who wanted to browse.
Where the categories come from
Link every item from a category small enough to list on one page. That is the working constraint and it decides the categories: if a category needs pagination, it is too broad and wants splitting, which is a more useful rule than any judgement about what customers want.
Sizing rule
How the one-page constraint decides every category boundary
If a category still needs page numbers, it is too broad and needs splitting until every item fits on a single page.
- Yes, small enough for one pageKeep as a single category hubEvery product gets a direct link on one flat URL.
- Links all prints without page two
- Replaces deep sequential browsing
- No, still spills past page oneSplit narrower or flag unfiled itemsNarrows broad groups by subject, size, or method.
- Forces smaller, clearer hubs
- Only 20 uncategorised prints stayed deep instead of 610
The 20 prints left deep are a filing gap the shop still needs to name, not a pagination flaw.
The 20 prints that still sit deep are the ones nobody could categorise, and they are deep because the shop has not decided what they are. That is honest and it is a smaller problem than the 610, and it is a hub, its supporting pages, and the links between arriving as a filing question.
The pagination nobody calls pagination
Filter pages are pagination with a different name and they multiply. Padiham's filters for colour, orientation and frame combine, so the shop generated about 2,100 filtered list addresses, every one of them a page of twenty prints and none of them a page anybody would land on deliberately.
Crawl control
Letting shoppers filter stock without letting bots crawl 2,100 combinations
Category pages exist to be found in search, while combined colour, orientation, and frame filters exist only to sort.
- Combined filter URLs+2,100 generated list addresses from inviting a bot fetch→Live shopper sorting by colour, orientation, and frame
- Numbered list pages+Automated crawling past the early list pages→People browsing page by page through the full catalogue
- Dedicated category hubs+Nothing blocked at the gate→Crawlers following direct single-page links to prints
Handling filter combinations with a directive keeps crawl budget off pages built only for interactive use.
Those addresses are a crawl problem rather than a structural one and they are handled with a directive rather than with a link, which is what to let bots see, and what to keep out. The distinction worth holding is that a category page is built to be found and a filter address is built to be used, and the second kind should never be linked in a way that invites a fetch.
Where this stalls: replacing the list with a scroll
Load more and infinite scroll leave nothing for a crawler to follow. Padiham considered exactly this in 2025, on the reasonable grounds that customers preferred it, and it would have removed the last route to the 610 rather than shortening it.
Dead end
Why replacing numbered pages with infinite scroll hides deeper stock
Removing page links makes browsing feel smoother while stripping away the HTML paths crawlers rely on.
- 45-page numbered listSlow for bots to reach page 45, yet HTML links still existed.
- Swap to load-more scrollLeaves no link in the delivered HTML for crawlers to follow.
- 610 deep prints orphanedRemoves the last crawl route instead of shortening click depth.
- Flat category hubsGives both shoppers and crawlers direct links without optimising page 2.
A list page like page 2 is not a landing page worth optimising; direct category hubs solve the depth problem cleanly.
The stall is that the change looks like an improvement to the same problem. A list of 45 pages is obviously bad, infinite scroll is obviously nicer, and the property that mattered was the one nobody was looking at: whether there is a link, in the delivered HTML, that leads to each item.
A list page is not a landing page and should not try to be. Padiham's page 2 has no title worth writing, no description worth spending twenty seconds on, and no reason to rank for anything; treating it as a page to optimise is a common way to spend a fortnight on the least valuable addresses on a site.
Reading the depth
Screaming Frog SEO Spider reports crawl depth per address, free up to a few hundred, and it replaced clicking through a site with a notepad. Its depth column is the whole diagnosis here and it takes about three minutes to read; anything at depth five or more on a shop this size is effectively unlisted.
Indexing shift
Discovered and not indexed addresses before and after category hubs
Search Console's indexing report confirmed from the crawl side what Screaming Frog's depth column predicted.
Addresses in Search Console 'Discovered - currently not indexed'
FebruaryJune
- Discovered, not indexed57461Fell after 33 category pages brought 610 buried prints closer to the home page.
Screaming Frog depth flags anything at depth five or more; Search Console tracks whether crawlers actually fetch it.
Search Console's page indexing report separates discovered from crawled, free, replacing counting pages by hand and believing the total. At Padiham the discovered and not indexed line held 574 addresses in February and 61 in June, which is the same story from the other end.
Whether the categories that fixed it are worth having as pages in their own right is pages that dilute the ones that matter asked in reverse, and where the 33 sit in the menu is menus as the clearest statement of your hierarchy. All of it is how pages relate, which decides what a site is about, and 610 prints is what the answer looked like when nobody had decided.