Thin is about what a page answers, not about how many words it has. Erdington Electricals sells appliance spares from Birmingham and had 1,900 pages in January 2026, of which 41 of them had ever received a click. The rest were not badly written. Most of them were not written at all.
A page that exists because a template generated it is the commonest kind. Erdington's 1,900 contained 620 tag archives, 310 paginated list pages, 240 author archives and 190 empty category pages for brands the shop no longer stocks. An author archive with one author on it is a copy of the blog index with a different heading, and there were 240 of them because the software made one per month.
The five questions
Count what a page answers that no other page on the site answers. That is the test, and it is applied one page at a time in the fast cases and to a whole class of pages at once in the slow ones. The five questions Erdington used: does anything link to it, has it ever had an impression, does it answer a query on the map, is there another page answering the same thing, and would a customer looking for it be disappointed to find it gone.
The audit
Five questions asked of one class of page
The 620 tag archives failed four of the five, and the fifth is the one that saves a returns policy from the same fate.
620 tag archives, January 2026
- YesDoes anything link to itThe theme linked every archive from the bottom of every post.Yes
- NoHas it ever had an impressionNone of the 620 had a single one in twelve months.No
- NoDoes it answer a query on the mapNo row on the map had a tag archive in the right hand column.No
- YesIs another page answering the same thingThe category page, with more products on it and a description.Yes
- Not checkedWould a customer miss itNobody asked. They went to the category pages, and no complaint arrived.Not checked
The first row is a yes and it is not a reason to keep anything: the links came from a template rather than from anybody's judgement, which is what generated them in the first place.
Keep a page a customer needs even if search will never want it. The returns policy has never had an impression and it is not going anywhere; the last question exists so that this audit does not turn into a traffic exercise that removes the pages holding a business together.
What was done with each class
Merge, improve or remove, and removing is a redirect rather than a deletion. The 620 tag archives were removed and redirected to their category pages. The 240 author archives were removed and redirected to the blog index. The 190 empty brand categories were removed and redirected to the nearest brand that is stocked, which is a customer decision as much as a search one, and 90 product pages with forty words on them were improved rather than removed because the products still sell.
Triage
How each thin page class was redirected or improved
Template archives with no unique role were redirected to active hubs, while short pages for selling spares were expanded.
- Redirect to parent hub2
- Redirect620 tag archives to category pages
- Redirect240 author archives to blog index
- Redirect to stocked substitute1
- Redirect190 empty brand categories to nearest stocked brand
- Improve in place1
- Keep90 product pages expanded past forty words
After 1,140 pages were removed, surviving pages gained 4 places on average to June, though two pages fell.
1,140 pages removed and the remaining ones gained 4 places on average, over the four months to June. That average hides a wide spread and two pages fell, and it is the shape of what happens rather than a promise about a number.
Where this stalls: waiting for the traffic that was never there
The position of the surviving pages is what moves, not the traffic of the removed ones. This is the sentence that makes the work possible to defend, because the first question anybody asks is what happens to the traffic from 1,140 pages, and the answer is that there was none: 41 pages had ever had a click and none of them was removed.
Resistance
Why cutting zero-click pages stalls for eleven months
Pruning looks like deleting traffic on paper even when the removed URLs never earned a visit.
- Why teams hesitateVisible drop before gains appear
- Feels like losing 1,140 pages
- Index counts dip before rank rises
- Six-year-old pages have owners
- Gains are slow and indirect
- What actually happensSurviving pages gain room
- Only 41 pages had ever had a click
- Zero of those 41 were removed
- Surviving pages move up in rank
- No real search traffic is lost
Pruning trade-off
Defending the cleanup starts with showing that none of the pages ever earning clicks are on the removal list.
The stall is that removal feels like loss and the gain is indirect, slow and hard to attribute. Nobody gets credit for it, the report goes down before it goes up, and a page that has been on a site for six years has an owner somewhere who will defend it. Erdington's audit sat in a document for eleven months before anybody did it.
The generated ones will come back
Removing 620 tag archives does nothing if the software makes a new one every time somebody tags a post. The settings that stop them being generated, or stop them being listed, are part of the same afternoon, and they are what to let bots see, and what to keep out.
Prevention
Stopping template archives from growing back after the cleanup
Redirects fail as a one-off fix unless archive-type controls and indexing checks block new bloat from entering the index.
- Archive type exclusions+New tag and author archives from being indexed as posts are tagged→Hub pages and active product listings
- Sitemap exclusion rules+Generated thin archives from being submitted to crawlers→Surviving hubs and 310 managed list pages
- Search Console indexing monitor+Unnoticed drops in valid page counts out of 1,900 total→Shrinking crawled-not-indexed counts while the 41 click-earners hold
As of 2026, setting exclusions per archive type in Yoast SEO takes ten minutes instead of editing template files.
Yoast SEO can set an archive type to be excluded and to be left out of the sitemap, free in the version most WordPress sites run, and it replaced editing template files to suppress archives by hand. As of 2026 the settings that matter are per archive type rather than per page, which is what makes this a ten minute change rather than a monthly chore.
Search Console's page indexing report is where the 1,900 and the 41 both came from, free, replacing counting pages by hand and believing the total. The line to watch after a removal is crawled and not indexed, which should shrink, and the line to watch for accidents is the count of valid pages, which should not.
Which of the survivors should be a hub with the others under it is a hub, its supporting pages, and the links between, the 310 list pages are long lists, without splitting your own signals, and every removal needs an address to point at, which is moving a page without losing what it earned. All of it is how pages relate, which decides what a site is about.