A standard question I receive during consultations, mainly from SEO specialists, is whether to do content clean up. Indeed, there is no consensus in the industry regarding this approach, and I understand these dilemmas, especially since many respected SEO experts are opposed to removing content from websites. My experience with this type of operations is mainly based on large media websites, which doesn’t mean that such actions shouldn’t be undertaken in e-commerce – quite the contrary. I have always made decisions about content removal based on my own considerations and data, and after clearly defining what we want to achieve with such action and what the goal is. Below I have outlined how I approach this.

Should You Do Content Clean Up?

The answer is short – YES. This type of housekeeping on a website should be done cyclically, especially in times when Google is becoming increasingly selective about our content and rolls out updates like helpful content. If such actions have never been undertaken, this first time will always be a challenge and must be approached with particular caution, and often the removal process itself should be spread over several stages.

What Is the Goal of Content Clean UP?

Removing worthless and outdated articles requires arguments. You can’t do this just because it needs to be done; there will always be the question of what benefit it brings us. Below are several arguments that can help convince the unconvinced.

Indexing optimization – increasing the speed of indexing new and updated content by search engines, which will translate into faster visibility for readers. Relatively easy to measure – if articles on a media website take longer than 5 minutes to index, a warning light should already be flashing. This argument is also very effective for people obsessed with time to market. The warning light should also flash if the “Page crawled but not yet indexed” metric in GSC is growing for articles that had visibility – I covered this parameter in more detail in my post about Crawled – Currently Not Indexed in Google Search Console. The third metric is simply Googlebot requests.

Strengthening quality signals – raising the average quality of content in Google’s index strengthens signals at the entire website level and will improve the performance of all materials (current and future). Difficult to measure, but can always be presented as an argument for building reputation.

Improving user experience – providing recipients with access only to the highest quality content strengthens positive brand perception. Building a strong brand in the era of Google Discover and AI often emerges as a key argument in the SEO industry. And if you present such arguments to the marketing and PR department, they will jump for joy.

Reducing maintenance costs – decreasing resources needed to administer and update content by focusing on fewer but more valuable materials. Maintaining a large website is expensive, maintaining cache on this website is expensive, everything costs money, you need to think carefully about whether it makes sense to bear these costs just for Googlebot when users don’t use these resources. Reducing maintenance costs is easy to measure and very popular nowadays.

Content Clean UP Step by Step

What seems like a simple task in large organizations often turns into a six-month crusade. Content Clean UP is not only the work of the SEO team, but in media it’s primarily close cooperation with the editorial team (we don’t want to remove content that is important or image-related to the editorial strategy), the sales department (removing sponsored content can end unpleasantly), and the product team (redirects, 410 on removed content, etc. – sometimes complex systems that media websites operate on can surprise with unusual obstacles).

Step 1: Data

• List of canonical URLs. Ideally extracted directly from the database. In case of problems, you can use Google Analytics, Sitemap, or Google Search Console, but you would definitely prefer to avoid this type of exercise.
• Publication date. As above, best directly from CMS.
• Sessions – minimum in the last 12 months. In large organizations, data is always in internal monitoring systems, but data from Google Analytics is also perfectly fine.
• Title and SEO Title
• Publication type. Depending on the organization, naming may differ, so I’ll list three basic ones here: evergreen, news, sponsored. If you don’t have this marked in CMS, it can be challenging, but it can be handled. Evergreen articles written for SEO are created in a specific way, so you can try to identify them using a pattern (that’s why we also extract titles and SEO titles). Another way to identify evergreen articles is long-term traffic analysis. We also can’t forget about the blessing that AI gives us.
• Category. We should have control over whether we accidentally clean out some department too much; additionally, categories can serve as natural redirects for articles marked for removal.
• Author. Similar to categories, we want to control how much content from a given author will be removed. Additionally, there will be a lot of content without an author, which is also an important indication to get rid of such content.
• Word Count. Counting words is, in my opinion, a relic of SEO, but we all used to do it this way, commissioning content by character count, it’s worth using this metric if any doubts arise.
• External links. This element is still important and content with backlinks should not be removed or redirected.

Step 2. Combining Sessions

Many websites have canonical addresses that were created, for example, through parameters or amp. This data needs to be combined so that we have a complete and reliable picture of what the real traffic is on a given article.

Step 3. Fresh URLs

URLs that were published in the last twelve months are removed from our analysis list; their time will come during subsequent Content Clean UPs.

Step 4. Evergreen and Sponsored

Content that we marked as Evergreen is moved elsewhere; a separate analysis will be performed for them, and depending on its results, such content will need to be updated, or ultimately also removed.

Sponsored content that cannot be removed due to signed contracts is removed from the analysis list.

Step 5. Determine Traffic Threshold

Here everyone must determine their own “pain threshold” based on data. This can be done by extracting the average and everything that is 20% below average qualifies for redirection or removal, and such content must be marked accordingly.

If you’re doing this for the first time and the scale is large, you can focus on content without traffic in the first iteration.

Step 6. Backlinks

All articles that have backlinks should remain on the website.

Step 7. Impact on Categories and Author Page Lists

There should be no situation where, after going through point five, it turns out that some category will be left with very little content or will be completely cleaned out. This will negatively affect user experience, but also classification. In this case, it will be necessary to reconsider traffic thresholds or, after consultation with the editorial team, make decisions about closing the category if it aligns with the publishing strategy.

The situation with authors is similar, where we assume two options:

• The article author is a syndicated source, e.g., PAP, Associated Press, etc.
• The article author is a specific person

In case a), we mark such content for removal. In case b), we must determine the scale. If it’s not large, we can send the list of addresses to be removed to the author for consultation for peace of mind, or manually assess what we shouldn’t remove ourselves. The situation becomes slightly more complicated when there’s no contact with the given author or there’s too much content. As a rule, however, editors often don’t want to sign their names to news articles; they prefer to use acronyms or other forms that don’t directly indicate that they are the author of the content.

Step 8. Consultations with Editorial Team

When you have a list of content that should be removed, send it to the editorial team for control with full information about why such actions are being taken. In theory, the editorial team should review this content and confirm or indicate content that shouldn’t be removed. This is theory; practice says that reviewing sometimes several or even tens of thousands of articles is impossible for the editorial team, which is why there’s step 9.

Step 9. Batch SEO Content Analysis

This step can be performed earlier, e.g., before step 8 (or replace it – not everyone has an editorial team), so as not to flood the editorial team with a large amount of content. Different companies have different additional tools that assess, for example, content quality. These are often proprietary tools, but there are also many tools available on the market for such analysis. Use these tools. If you don’t have such tools, you can use artificial intelligence and build a tool that will assess, for example, based on helpful content update guidelines, the quality and usefulness of content you have marked for removal.

Step 10. Decision

Do you already have a list of content to remove? Great. Assess the scale. Don’t remove everything if it turns out that most content is suitable for this (it happens). Approach this task in stages. Remove some content and observe Googlebot behavior and metrics.

Step 11. Removing

Now the ordeal with IT begins; it won’t be so easy. After all, we won’t be putting 410 everywhere; that would be too beautiful. First, redirects should be implemented.

• You need to find content with high similarity scores. In the AI era, this is a task that shouldn’t be challenging. Here again, you must decide what similarity score will be your matching threshold based on scale.
• Content that doesn’t meet threshold requirements can be redirected to a category or the first selected tag in order.

What content do I mark as 410? Most often “one-liners” or sponsored content with zero traffic. As it turns out, there’s quite a lot of this.

What else to pay attention to regarding redirects:

• Make sure redirects also apply to non-canonical versions (amp, addresses with parameters)
• Make sure redirects aren’t in sitemaps
• Avoid redirect chains
• Remember to remove redirects after some time (6-12 months) and mark articles as 410.

Step 12. Impact Analysis

Depending on the website size, you must assess how much time you need to monitor the impact after implementation and compare it with the previous period. As always, I pay attention to the following metrics:

• Crawled – currently not indexed – expected decrease in this metric
• Googlebot requests
• Impressions – expected increase in metric
• Visits from Google (Search, Discover, News) – expected increase in metric

Assessing how much something should increase or decrease depends on the website size, so I don’t want to give specific numbers. I’ve seen cases where changes in metrics were double-digit, but there were also cases of single-digit changes, which doesn’t change the fact that they were less satisfying.

What’s important is to share the data with a broader audience in your organization. This will help understand the SEO team’s work, understand the purpose and validity of SEO actions, increase organizational engagement in SEO processes, and allow building PR for the SEO team, often perceived as a group of magicians working on something very secretive.

From my experience, but also from available literature, this type of website housekeeping has always brought positive effects when performed according to best practices. I believe that such a process should be repeated at least once a year.