Confluence Needs a Native "Content Health" Feature

[#1]

Confluence Needs a Native “Content Health” Feature

After working with Confluence in a large enterprise environment, I’ve noticed a recurring challenge that rarely gets enough attention. 

It’s oversized pages.

This was one of our major issue for us leading to performance issues, especially when a large team access it, they export and causing OOM, java heap issues. 

As knowledge bases grow, it’s common to find pages containing tens of thousands of words, dozens of tables, hundreds of images, numerous macros, and years of accumulated updates. These pages become difficult to edit, review, navigate, and maintain. In some cases, they can also contribute to slower editing experiences and increased strain on indexing or rendering.

The missing capability

Today, Confluence doesn’t provide:

• Warnings when a page becomes excessively large
• Configurable word or character thresholds
• Visibility into page complexity (tables, macros, images, attachments)
• A content quality or health indicator
• Recommendations to restructure large pages

I don’t think it has a mature solution that tells an author, in real time:

Content Health: 58/100

• 18,000 words
• 75 macros
• 66 tables
• 2.2 GB attachments
• Recommendation: Split this page into smaller child pages.

That kind of quality guidance is different from analytics or reporting. It helps prevent problems before they arise rather than identifying them after the fact.

A better approach

Rather than imposing hard limits, I’d love to see Confluence introduce a Content Health feature.

Imagine every page displaying a simple health summary:

• Word count
• Number of tables
• Number of macros
• Images and attachments
• Estimated page complexity
• Overall Content Health Score

Administrators could define recommended thresholds for their organisation. When a page exceeds those thresholds, Confluence could gently guide the author with suggestions such as:

“This page is becoming large and may be harder to maintain. Consider splitting it into child pages.”

This isn’t about preventing people from creating comprehensive documentation. There are many legitimate cases for large pages, such as technical documentation, architecture specifications, and operational runbooks.

Instead, it’s about encouraging sustainable knowledge management through guidance rather than enforcement.

FYI: Enterprise customers often don’t want to write and maintain custom scripts for capabilities that feel fundamental to knowledge governance. A native solution would be easier to adopt, supported by Atlassian, and available to customers. 

Why this matters

Confluence has evolved into a critical platform for documentation, governance, and collaboration. As organisations continue to grow their knowledge bases, helping teams create content that is maintainable, discoverable, and scalable becomes increasingly important.

A native Content Health feature could improve:

• User experience
• Knowledge quality
• Long-term maintainability
• Administrator visibility
• Overall platform performance

Sometimes the most valuable features aren’t the biggest ones, they’re the ones that quietly help users build better habits.

I’d be interested to hear how others manage oversized pages today. Do you rely on conventions, Marketplace apps, or manual governance? Would a native Content Health feature add value in your environment?

Love to hear thoughts, proposals and ideas. 

With curiosity,

Viswa

[Josh]

@Viswanathan Ramachandran I really like the idea you've proposed here. I could see orgs creating Rovo agents to perform this kind of assessment periodically.

Ideally, Atlassian would have the functionality built right into the page editor itself so the insights can be addressed at the time of authoring / updating content rather than after-the-fact.

Regarding the large attachments piece - that's something that can be configured today (at least on an individual file basis) to prevent overly large files from being attached. It'd be great to be able to set total attachment file size limits per page as well.

• Confluence as a file storage system is definitely an enterprise anti-pattern.

[Viswanathan Ramachandran]

Thank you @Josh for the response. I'm aware of the attachment file from general configuration. I was more interested in per page limit. 

I'm more positive that this is coming as Confluence evolved in years.

[Rob Hean]

Heya @Viswanathan Ramachandran  - It would be interesting to see what metrics would be useful.

Personally I don't think there should be "hard" limits to a Confluence Page as it's designed to let folks capture whatever they think is necessary. Are incredibly long documents with lots of tables/etc. generally a good idea? not really... but they do have a purpose.

That said any time I think about a technical a control, I try to think of why it's needed and why it's happening (it's very easy to think of the technical control as a solution to underlying challenges, when, in my experience, it's more of a bandaid...) For example:

1. Have people been given any type of training on what good/useful content looks like?
2. Are there guidelines/policies in place to keep people compliant?
3. Is anyone monitoring pages against those guidelines?
4. etc

Without work on the "why" behind the control I'm not sure displaying numbers will be as useful as we'd hope.

I'd also be curious to know what other processes are in place to help manage content. You mention pages with multiple revisions over a few years... is that content REALLY needed, or is it something legacy that should either be archived or rewritten?

I could also see a Rovo agent take some of this on. Give it instructions to review newly created pages (or whatever) against whatever guidance your company has on what good/useful content is, and then provide recommendations on improvement.

Something like:

You are a technical writer specializing in creating easily accessible and useful content for our company. Whenever a new page is created you should evaluate it against this criteria (either link to a Confluence page or add in the instructions).

Once you've analyze the page, append a comment to it detailing potential adjustments to make it more useable / easily to understand.

[Stavros_Rougas_EasyApps]

@Viswanathan Ramachandran this post is thought out and well written. You have articulated a pain point well clearly and some ideas to help solve it.

In some ways the problem is good as it means the organization is using a wiki (Confluence) a lot and not simply low-balling it (playing is safe) with silos that hide lots of good stuff as scared to show imperfection (I'm looking at you Microsoft Office).

@Rob Hean you said it well:

Without work on the "why" behind the control I'm not sure displaying numbers will be as useful as we'd hope.

I'd also be curious to know what other processes are in place to help manage content.

I build apps for Confluence and since I'm not a developer I come at it with less a belief that there is a 'magical' technical solution, Rob is saying something similar in another way.

@Josh I hear you about Rovo, but it is a support tool, if the underlying problem is not dealt with and understood no tool or AI can do a lot, worse it can be an excuse not to have those harder conversations (magical thinking, e.g. Brexit).

I have thought a lot about what data adds value, not simply noise. This is from our suite of content management tools called Space Content Manager, an example of some of the data we provide as a proxy for space health.

When I say health, knowing what you have is a step that @Viswanathan Ramachandran articulates well. Then having ways to discover and update, in our case in bulk, as a way to make it less of a mess (i.e. a piece of the solution).

@Josh part of the problem with Rovo/AI is the lack of preview, being able to account of anomalies. In theory (technically) it works but in reality have to review or the cleanup could make things worse.

The technical work is only part of the solution. I would argue the easier part. But alone it barely moves the goalposts.

[Viswanathan Ramachandran]

@Stavros_Rougas_EasyApps and @Rob Hean 

great to see your well put thoughts and response. You both indeed had put well. Yes, having technical control and managing the contents is extremely important and it comes via practice and discipline. 

It will be great if Rovo agent does this: for example, moving and splitting the contents of larger pages to new ones. 

On this note, having the statistics and visualisation of health metrics adds value. 

[Stavros_Rougas_EasyApps]

@Viswanathan Ramachandran I agree with your point about Rovo:

It will be great if Rovo agent does this: for example, moving and splitting the contents of larger pages to new ones.

But it's garbage in, garbage out. Splitting is only useful if done well, if pages are a mess than Rovo will not split them well. So as @Rob Hean said, we have to ask ourselves how do we reduce the 'garbage in'.

[Barbara Szczesniak]

I agree, @Stavros_Rougas_EasyApps. I recently went through an exercise where I wanted to split some very long pages into multiple pages.

Rather than having Rovo do it, I instead asked Rovo to analyze the long page and propose how it could be broken up into separate pages. Rovo gave me a proposed outline of pages, what content would go on each, and a reasoning why.

I considered Rovo's suggestions (some of which I would never have thought of), determined what I thought was good and bad, and then manually created the separate pages. If I just had Rovo do it for me, I wouldn't have been happy with the result, and I don't know if it would have been easy to revert the changes.

[Stavros_Rougas_EasyApps]

@Barbara Szczesniak sounds like you leveraged Rovo well.

For our bulk editing, clear previews as so important as users need to verify before changing. Hard to know if are exceptions without at least a quick review.

AI is great, but it always gives an answer or recommendation with total confidence. I used to work as a journalist, it reminds me of when would get an American male as a guest, were often 'good talkers'. That means say things with an convincing confidence, but that has a major downside to understanding because rarely was it definitive and obvious, if it was we would not be discussing it, so it could given listen a false sense of clarity.