In brief
- 150 articles of which 12 bring anything in: that is not a volume problem but a sorting problem. Producing more fixes nothing.
- The inventory comes down to three exports: the published URLs, Search Console rankings over 16 months, conversions by page.
- Five possible decisions: keep, update, merge, de-index, delete with a permanent redirect. The signal observed dictates which.
- Pruning often lifts the rest of the site, because Google assesses quality at the level of the domain, not only page by page.
A small business publishes two articles a month for six years: about 140 pages. When we open its Search Console, the distribution is almost always the same. Fewer than 10% of URLs account for more than 80% of organic clicks, and half the site has had no impressions in twelve months.
The usual reflex is to order one more article. It is the worst possible moment: as long as the hundred dead pages stay online, every new text lands on a site that engines crawl badly and assess badly.
Auditing existing content is the least glamorous job in search and often the most profitable. It needs no production budget, it needs a decision per page. Building organic visibility is another subject, covered in our SEO and content marketing method.
The inventory: three exports, one sheet
The XML sitemap (usually /sitemap_index.xml) gives all the indexable URLs. On WordPress, the CSV export of posts also does the job.
Second export: Search Console, Performance report, over 16 months (the maximum depth Google keeps). You get impressions, clicks, average position and click-through rate per page. The interface caps at 1,000 rows, which is enough here.
Third export: your conversions by landing page. Watch the retention period: in Google Analytics 4 it is 2 months by default and can be set to 14 months. If nobody has changed it, your history is shorter than you think.
Finally, add two manual columns: internal links to the page, and whether external links exist. These are what stop you deleting a page that still has value.
The decision grid, page by page
Each row gets one decision, and only one. The aim is to eliminate debate: the signal observed determines the action.
| Signal observed | Decision | Reason |
|---|---|---|
| Position 1 to 5, stable clicks, accurate information | Keep as is | You do not touch what works |
| Position 6 to 20, current topic, dated or too short content | Update | The potential is already proven by the impressions |
| Two or three pages on the same query, none in the top 10 | Merge | They compete with each other and share the links |
| Zero impressions over 12 months, page useful to customers | De-index, keep online | Real use, without weighing on the domain |
| Zero impressions, zero inbound links, dead topic or off-business | Delete with a 301 redirect | Will bring nothing and dilutes how the site is perceived |
| Decent traffic, zero conversions, non-commercial query | Keep if it attracts links, otherwise rewrite | An off-target audience is not an audience |
Two rules. Never decide on a single metric: a page with no traffic but three good external links is kept. And cap each wave of deletions at 10% of URLs, to observe the effect before continuing.
Why removing pages lifts the others
This is the hardest point to get across to a manager who paid for each of these articles. Deleting seems to destroy value. But the value had already gone, and keeping these pages costs the rest of the site something.
With the September 2023 helpful content update, Google explained that removing unhelpful content can help a site’s other pages rank better: the assessment is partly made at the level of the domain. In March 2024, that system joined the core ranking algorithms.
Add to that the crawl budget. A robot that spends its time on pages that are never served visits the ones that matter less often, and your updates take longer to be picked up.
The third effect is internal. Every link from your menu or your related-articles blocks passes on authority, and a hundred worthless destinations dilute what the ten pages that sell receive. We detail this hierarchy in our SEO, search ads and social expertise.
Merging, in practice
Merging means producing a page better than both and making the other disappear cleanly. The order matters.
First choose the surviving URL. The criterion is not the quality of the text but the history: keep the one with the most age, the most external links and the best position. Content can be rewritten; a link history cannot be rebuilt.
Then bring over the best of the other page: its unique passages, its examples, the questions it covered. Set up a permanent (301) redirect to the survivor, fix the internal links, and check that the absorbed page really leaves the index.
The mistake not to make Deleting without redirecting. A deleted URL returns a 404: the external links pointing to it pass on nothing and the history is lost. The same damage happens if you redirect everything to the home page, because Google sees a redirect with no semantic relation as a soft error and treats it as a 404. The rule: each deletion points to the closest page by topic. If none exists, the page needed updating, not deleting.
Updating often pays more than writing
A page between 6th and 20th place has already cleared the costly stages: indexed, judged relevant, it receives impressions. Moving it up takes a few hours, against months to get the same volume with a new article.
The trap is believing that an update means changing the date. An identical text republished is spotted, and the effect drops off within a few weeks. Here is what really moves a page.
| What you change | Why it moves the page |
|---|---|
| The queries actually served, taken from Search Console | The page answers questions the author had not anticipated: you write what is missing |
| The first 150 words | The direct answer at the top is what engines and AI assistants quote |
| The subheadings, reworded as readers’ questions | The page becomes addressable question by question |
| Figures, screenshots and tool names | A figure from 2021 signals abandoned content |
| Internal links from your strong pages | The only signal you control entirely |
Five pages updated a month beat two new articles, as long as the stock is not exhausted. Production then resumes, with a structured editorial method so as not to rebuild the same dead stock.
The case of mass-produced content
Many businesses had 50 to 300 pages generated in a few weeks, often in 2023 and 2024, on close keyword variations. The result is recognisable: texts that look correct, with no verifiable example and no identifiable author.
Since March 2024, Google classifies this practice under its scaled content abuse policy. The criterion is not the use of AI, it is producing pages in volume primarily for ranking rather than for the reader. An AI-assisted page that is reviewed, documented and signed does not fall into it.
The treatment depends on the data. Pages that receive impressions are rewritten with material from the field: your customer cases, your prices, your lead times. The others, the majority, are deleted in waves with a redirect to the matching category. That is what we point out regarding artificial intelligence in marketing: the tool speeds up writing, it does not replace what only you know.
Your audit in an afternoon, under 100 pages
Under a hundred pages, the exercise takes four hours with a spreadsheet. The order that works:
- 30 minutes. Export the sitemap, Search Console over 16 months and your conversions. One row per URL.
- 20 minutes. Add the internal links and external links columns. A free crawl or a
site:search is enough here. - 60 minutes. Sort by main query and highlight the URLs that appear twice: your merge candidates.
- 60 minutes. Apply the grid row by row, without going back. One decision per page, in a single column.
- 30 minutes. Write the redirect plan, old URL to new, to be approved by whoever manages the site.
- 40 minutes. Carry out the first wave: obvious deletions and de-indexing. Merges and updates go into the schedule.
What to remember Before carrying anything out, back up the site and export a snapshot of the day’s rankings. Without that starting point, you will not be able to say in three months whether the pruning had an effect. It is the step nobody takes and everybody regrets.
Frequently asked questions
Can deleting articles really improve my search rankings?
Yes, indirectly. Google has said that removing content that is unhelpful to visitors can help a site’s other pages rank better, part of the assessment being made at the level of the domain. Deletion does not create traffic: it removes a weight and concentrates crawling and internal links on the pages that convert.
How long before you see an effect after pruning?
Allow 4 to 12 weeks for Google to reprocess the URLs concerned, longer on a little-visited site. Merges give a faster signal, because a consolidated page gains positions on queries it was already serving. Change nothing else during that period, otherwise you will no longer know what to attribute to what.
Does a 301 redirect pass on all of the old page’s ranking?
It passes on most of it, provided the destination covers a close topic. If it sends to an unrelated page or to the home page, Google sees a soft error and treats it as a page not found: the history is lost. Semantic closeness matters more than the technical mechanics.
Should you delete all articles written by AI?
No, the criterion is not the tool but the usefulness of the page. Keep and enrich those that receive impressions and cover a real subject of your business. Delete, with a redirect, the pages produced in volume on keyword variants that are never served in the results.
Sort before producing again
An audit does not replace an editorial strategy, it makes it possible. As long as the dead stock stays online, every new article starts with a handicap that budget does not make up for.
The order does not change: take stock, decide page by page, carry out in waves, measure, then resume production. On an online shop, this trade-off combines with the other levers, as we detail in how SEO, content and automation fit together in e-commerce.
Unsure what your old articles are worth? In a first meeting, we open your Search Console over 16 months, sort your URLs into five piles and hand you the decision sheet, whether or not you then work with us. Let’s talk.