Which pages should you prune, and how do you check that pruning helped?

Pages worth reviewing for pruning are those a long Search Console export, set against your sitemap, shows with no impressions or the fewest clicks, once canonical, noindex and age are checked. To check that pruning helped, compare the pages that took over their job with untouched pages over the same days; a gap is timing evidence, not proof.

• Google's core updates page calls deleting content a last resort, to be considered only if you think the content cannot be salvaged (Google, Search Central documentation, read 30 October 2026).
• Google says a Search Console export is truncated to 1,000 rows of representative examples of your data (Google, Search Console Help, read 30 October 2026).
• Penlid's content pruning tool lists addresses with no row only for an unfiltered export that Google did not cut, meaning fewer than 1,000 rows or a BigQuery bulk export table (Penlid, content pruning tool, read 30 October 2026).
• Penlid's methodology page gives a page with fewer than 10 clicks in the 28 days before a change a direction and no size, and a page with none no measurement (Penlid, methodology page, read 30 October 2026).

Which pages are worth reviewing for pruning?

The pages worth reviewing are the addresses that earned no impressions and the pages that earned the least, over the longest range you can export. The free content pruning review builds both lists in your browser from a Pages export and a list of the addresses you want in search, usually a sitemap, printing the no-impressions list only when the export is complete enough to say so. Each is a page to review, never one to remove: A page on either list may need improving, merging into a stronger page, a redirect or a noindex rule, or nothing.

Export the Pages table from Performance, then Search results, with the longest range you can choose, the Web search type and no filter, and upload the whole ZIP so the tool can read its Filters file. Search Console keeps data for the last 16 months, and the guide to keeping Search Console data older than 16 months covers building a longer history. A page with no impressions over a few weeks may simply earn in another season, so the tool warns when the range is shorter than 365 days, a judgment call and not a Google number.

Before acting on any address, check the three things the tool names first:

• The canonical. Google says Most performance data in this report is assigned to the page's canonical URL, so, as the tool's page notes, an address can have no row because its traffic was credited to another address.
• A noindex rule. A noindex page is meant to have no impressions, so having none is not a finding.
• How new the page is. A page published inside the range has had fewer days than the range implies.

Pages losing ordinary search traffic rather than earning none need a different review, covered in the guide to telling content decay from a site-wide dip.

Why does the 1,000-row limit matter so much for this review?

The limit matters because the rows an export cuts are, on Penlid's reasoning, the ones this review needs. Google says Data is truncated to 1,000 rows of representative examples of your data, and its page on data discrepancies says the table shows at most 1,000 rows, so some rare or long-tail rows might be omitted from the table. The tool's page states the consequence: The rows that may be cut are rare or long-tail ones, which is exactly where the pages you are looking for are. That those rows are the pages this review needs is Penlid's reasoning, not Google's.

So the tool prints its list of addresses with no row only for an unfiltered export of fewer than 1,000 rows or a BigQuery bulk export table. The guide to getting more than 1,000 rows from a Search Console export covers the routes around the cap.

A BigQuery bulk export is the exception: Google describes it as holding the property's performance data with the exception of anonymized queries, so the tool does not hold it to the 1,000-row rule. Its history starts late, because Google says The first export includes data for the day of the export and points to the API or the reports for earlier days.

What does Google say about deleting pages?

Google's core updates page says Deleting content is a last resort and, in the same sentence, limits that step to cases where you think the content cannot be salvaged. The only supportive line we found is conditional: for sections of a site created for search engines first, Google says deleting the unhelpful content can help the good content on your site perform better. Google's guidance on people-first content also says that removing a lot of older content mainly to make a site seem fresh, in the hope of better rankings, will not help.

How you remove a page changes what Google does with the address:

How do you check that pruning helped?

You check that pruning helped by comparing the pages that took over the removed pages' job with pages you left alone, over the same days, and reading the gap with its interval. Google's Search Console page of common tasks, in its section on measuring the impact of page changes, suggests you confirm whether your improvement in Search maps closely in timing to your site changes.

Penlid's free Did it work? tool does this with the method on the methodology page. Four points matter:

1. Record what you changed and when, and which pages you left alone; the untouched pages are the one fact that cannot be rebuilt afterward.
2. Measure the page that took over, not the page you removed. A removed page's own clicks are meant to end, so measuring it records the removal, not a result (our reasoning). After a merge or a redirect, measure the target.
3. Compare 28 days with 28 days. The change is measured over the 28 days before it and the 28 days after it, against untouched pages; the guide to how long to wait after a content update covers the timing.
4. Read the interval, not the headline alone. With fewer than three untouched pages left, the result is "not measurable" rather than a number.

Starting traffic, counted on the page you changed in the 28 days before the change, decides what the check can say. With no clicks there is no baseline, so there is no lift to measure; with 1 to 9 clicks you get a direction and no size; with 10 or more you get a size, its 95% interval and a p-value. Pruning candidates are low-traffic by definition, so Penlid's simulations matter here (4 clicks a day is about 112 clicks in 28 days):

Drift between pages costs a lot: on six comparison pages, the 4-clicks-a-day, +80% case falls from 92.01% with no drift to 59.65% with 15% drift, a figure from 2,000 simulated experiments that can be a point or so off. A real change smaller than the table's lifts will often come back "we cannot tell", and the result then says how big a lift it could have seen.