The pages you should delete this quarter

Most established sites would perform better with 30% fewer pages, and the pages to remove are obvious once you stop measuring coverage and start measuring what each page contains.

This used to be an argument about crawl budget, which for a 400-page site was never a real constraint. The argument now is different and better. Retrieval selects passages, and it selects them from your site's whole pool of eligible text. Every thin page is a competing candidate that will never win and might be picked over a better one. Pruning does not reduce your surface area. It concentrates it.

There is also a compliance edge. Google's scaled content abuse policy targets many pages made mainly to manipulate rankings rather than help people, and human-written thin content is covered identically to generated content. A directory of 180 location pages built in 2021 by a previous agency is squarely in scope, whoever typed it.

The four page types to cut

Location pages with swapped suburb names. If the only difference between the Parramatta page and the Chatswood page is the word Parramatta, you have published one page 40 times. Keep the locations where you have something real to say, a job you did there, a council requirement, a travel or coverage fact, and delete the rest.

Answer pages that answer nothing. The definition explainer written for a keyword, containing what every other explainer contains. Google's generative AI guidance names this category directly, using something like 7 Tips for First-Time Homebuyers as its own example of commodity content.

Old campaign and event pages. The 2022 webinar registration, the promotion that ended, the announcement that you were attending something. These are not risky, they are just dead weight that dilutes the pool and makes your site look stale to anyone who lands on one.

Tag and filter pages producing near-duplicates. WordPress tag archives, faceted URLs, printer-friendly variants. Mostly a technical fix rather than a delete.

The two questions that decide each page

Run this over every URL. It is quicker than it sounds.

Does this page contain anything that could only come from us? A number from our own work, a photo of our own job, a rule we learned by getting it wrong, a price, a constraint. If no, it is commodity.

Would we publish this today if we were starting the site now? Not does it get any traffic. Would we choose to write it.

Two noes is a delete or a merge. Not a rewrite. Rewriting a page with nothing to say produces a longer page with nothing to say, and it costs more than deleting it.

Delete correctly, because the mechanics matter

Merge before you delete wherever there is real content. Six thin pages on adjacent topics become one page that answers all six sub-questions in self-contained sections. That page is now retrievable for six things instead of nothing, which is the entire point of the exercise.

Redirect to the closest equivalent page, not the homepage. Bulk redirects to the homepage get treated as soft 404s and you gain nothing.

Where there is no equivalent, return 410 rather than 404. It is a clearer signal and it drops out of the index faster.

Only use noindex when you need the page live for humans, a legacy landing page a client still links to, a form nobody wants broken, but do not want it in the pool.

Export everything before you touch it. Full HTML, not just a list of URLs. You will want it in three months.

Measure it honestly, because the effect is not guaranteed

Pruning is not a lever that reliably lifts rankings on its own, and anyone promising a percentage is guessing. What it reliably does is remove risk, remove maintenance, and improve the odds that the passage a retrieval system picks from your domain is a good one.

Set the measurement up before you cut. Take a baseline of impressions and clicks for the pages you are keeping, not the whole site, because total site impressions will fall when you remove 120 pages and that fall means nothing. Watch the kept pages over eight to twelve weeks. If the pages you kept are flat and the pages you removed were producing nothing, you succeeded: same output, smaller surface, less risk.

Give it a full quarter. Index changes on a few hundred URLs take weeks to settle, and reading the first fortnight will send you into a panic over noise.

The failure mode to avoid

Do not delete pages that convert quietly.

Every site has a page with 40 sessions a month that sits in three sales email templates. Content reports cannot see this, so the audit flags it for removal and someone approves the list without asking. Before you cut, send the delete list to sales and to whoever runs paid, and give them 48 hours to object. Two people will save two pages each, and those four pages will be worth more than the other 116 combined.

Also check backlinks before deleting. A dead 2019 page with three links from industry publications is an asset. Merge it and redirect, do not 410 it.

The order of work

Pull all URLs with a 16-month Search Console window and crawl the site so you catch pages that never got an impression. Tag every URL against the two questions. Group the commodity pages into merge candidates. Publish the merged pages first, let them index, then redirect the originals. Delete the rest in one batch with clean 410s. Document what you did and why, because in eighteen months someone will ask.

A tighter site is not a smaller asset. It is the same asset with the noise taken out, which is the version retrieval is picking from now.