Corrections and adjustments

We discovered a bug in the query used to calculate prevalence of Fake Accounts on Facebook in Q1 2024 and have resolved it. As a result we slightly over estimated prevalence in Q1 2024.

Prevalence of violating content is estimated using samples of content views from across Facebook or Instagram. We fixed a bug where a small amount of irrelevant data was sampled, which has been resolved and did not result in any significant change to prevalence.

Proactive rate increased for multiple violation types as we updated our methodology. Prior to this change, when a post was reported by users we also considered the comments on the post as user reported. As of Q3 2023, we have changed the methodology to only consider comments as user reported if they were reported directly by users.

In addition, metrics for Phishing will no longer be counted as part of Spam in Community Standards Enforcement Reports as of Q3 2023 to align with a previous change in Meta’s policy.

Per the European Union’s Digital Services Act (DSA), appeals need to be accessible to users for 6 months after an enforcement action is taken, thus increasing the appeal window. This means that to meaningfully represent the user experience in our fake account metrics, we have updated our accounting of fake accounts to align with this new time period. The increase we observed between Q1 and Q2 in Fake Accounts removals is due to this accounting change.

As part of our work to constantly refine and improve the metrics we share in this report, we updated our proactive rate methodology starting in Q4 2022 to only count enforcement actions as “proactive” only if we find and action the violating content before users report it to us. The old methodology would count the actions as “proactive” if proactive detection happened first in scenarios where both detection causes were present (i.e., content was both proactively detected and reported to us by users). The new methodology takes the existence of user reports into account and counts these instances as “reactive” now, instead of “proactive.” While this change does not materially change the metrics, it contributed to minor quarter over quarter differences in the proactive rate metric. The table in this document provides a comparison of the metrics as measured with both the old and new methodology for the Q3/Q4 2022 reporting periods.

Starting in Q1 2020, due to a temporary reduction in our review capacity as a result of COVID-19, we could not always offer people the option to appeal but still gave people the option to tell us they disagreed with our decision. As reflected in our data between Q1 2020 and Q1 2022, we did not count these instances in our appeals metric definition, because they provide valuable user feedback but do not qualify as an appeal without the opportunity for review. Over the last year, we've been improving and developing these appeal experiences, and as our operations have stabilized we now review many of these instances.

As part of our work to constantly refine and improve the metrics we share in this report, we updated our appeals methodology starting in Q2 2022 to account for all instances where content was submitted for additional review, including when people told us they disagreed with our decision. We are still excluding instances where content is not submitted for additional review, even if people told us they disagreed with our decision such as in many cases of spam.

In Q4, we identified and reclassified actions we had taken on terrorist content on Facebook and Instagram. This impacted the numbers we had previously shared for content actioned, proactive rate, appealed content and restored content for Q3 2021, and we’ve adjusted the numbers accordingly.

In Q2 2020, some of the content we actioned against our policy for violent and graphic content was later found to be in violation of our specific policy for suicide and self-injury. We reclassified this content accordingly, which impacted numbers we previously shared for content actioned on Instagram in Q2 2020. Additionally we made slight adjustments to our restored content numbers for child nudity and sexual exploitation on Instagram in Q3 2020, due to previously mis-categorized entities. We will continue to update historical numbers as we update our policies and continue to improve our systems and accounting.

This quarter we made refinements to our metrics for spam and suicide and self-injury, which led to minimal changes from previously reported numbers. We also made methodology adjustments that led to some small shifts in proactive rates for bullying and harassment and suicide and self-injury. Finally, we reclassified some content actioned for spam under adult nudity and sexual activity, which affected metrics for content that we restored.

In 2020, some of the content we actioned against our policy for violent and graphic content was later found to be violating our policy for suicide and self-injury. We reclassified this content accordingly, which impacted numbers we previously shared for content actioned on Facebook in 2020.

In Q4, we introduced clarifications on certain classes of images for our policy on adult nudity and sexual activity on Facebook. We restored some previously actioned content based on the latest policy, which impacted the numbers we previously shared for restored content on Facebook in Q3.

For violent and graphic content on Facebook, prevalence was previously reported in the Community Standards Enforcement Report for November 2020 as between 0.05% and 0.06% of views. In the February 2021 report, we updated prevalence for violent and graphic content to about 0.07% of views in Q3.

In Q2, some of the content we actioned against our policy for violent and graphic content was later found to be in violation of our specific policy for suicide and self-injury, after we regained some manual review capacity in early September. We reclassified this content accordingly, which impacted numbers we previously shared for content actioned on Facebook in Q3.

Additionally, we adjusted our restored content numbers for Q1 and Q2 on Instagram to account for previously unreported comments we restored. This has resulted in minimal changes across most policy areas on Instagram, and we adjusted previously shared data accordingly. We will continue to update historical numbers as we update our policies and continue to improve our systems and accounting.

In Q3, we made an update that recategorized previously actioned cruel and insensitive content so it is no longer considered hate speech. This update impacted the numbers we had previously shared for content actioned, proactive rate, appealed content and restored content for Q4 2019, Q1 2020 and Q2 2020, and we’ve adjusted the numbers accordingly. We also updated our policy to remove more types of graphic suicide and self-injury content, and recategorized some violent and graphic content we had previously marked as disturbing in Q2.

Additionally, we adjusted our restored content numbers for Q1 and Q2 on Instagram to account for previously unreported comments we restored, in addition to an issue with our data source for the August 2020 report. This has resulted in minimal changes across most policy areas on Facebook and Instagram, and we are adjusting previously shared data accordingly. We will continue to update historical numbers as we update our policies, and continue to improve our systems and accounting.

In Q1 2020, we identified and corrected an issue with the accounting of actions taken by our proactive detection technology for violent and graphic content on Instagram, and we were able to update our full reporting systems in Q2. For violent and graphic content on Instagram, content actioned in Q1 2020 was previously reported in the May 2020 report as 2.3 million pieces of content, and has been updated to 2.8 million in the August 2020 report.

At the time of our last update in November 2019, we made a number of improvements to our systems and accounting. These improvements allowed us to estimate largest impacts while still adjusting our metrics at that time. Following the November 2019 report, we further refined these improvements.

Because of this work, in the fifth edition of the Community Standards Enforcement Report for May 2020, we are adjusting previously shared data. Most categories for 2019 are only minimally impacted, and any adjustments to data amount to no more than a 3% change in content actioned. We will continue to update historical numbers as we reclassify previously removed content for different violations based on existing and changing protocols, and continue to improve our systems and accounting.

At Meta, different systems take action on different types of content to improve efficiency and reliability for the billions of actions happening every quarter. One of these systems, which acts mainly on content with links, did not log our actions for certain content that was removed if no one tried to view it within seven days of it being created, even if this content was removed from the platform.

While we know this undercounts the true number of content containing external links, mainly affecting our spam metrics for content containing malicious links, we are not currently able to retrospectively size this undercounting. As such, the numbers currently reflected in the community Standards Enforcement Report represent a minimum estimate of both content actioned and proactive rate for the impacted period. Updates about this issue will be posted here when available.

When we shared the second edition of the Community Standards Enforcement Report in November 2018, we updated our method for counting how we take actions on content. We did this so that the metrics better reflected what happens on Facebook when we take action on content for violating our Community Standards. For example, if we find that a post containing one photo violates our policies, we want our metric to reflect that we took action on one piece of content – not two separate actions for removing the photo and the post.

However, in July 2019, we found that the systems logging and counting these actions did not correctly log the actions taken. This was largely due to needing to count multiple actions that take place within a few milliseconds and not miss, or overstate, any of the individual actions taken. Because our logging system for measurement purposes is distinct from our operations to enforce our policies, the issue with our accounting did not impact how we enforced our policies or how we informed people about those actions; it only impacted how we counted the actions we took. As soon as we discovered this issue, we worked to fix it, identify any incorrect metrics previously shared and establish a more robust set of checks in our processes to ensure the accuracy of our accounting. In total, we found that this issue impacted the numbers we had previously shared for content actioned, proactive rate, appealed content, and restored content for Q3 2018, Q4 2018, and Q1 2019.

The fourth edition of the Community Standards Enforcement Report includes the correct metrics for the affected quarters and the table linked above provides the previously reported metrics and their corrections.