Campaign Data
The initial 140 items were in English and evenly split between anti-Muslim hate and antisemitism. A further 21 items (almost all on antisemitism) in languages other than English were then added. These items were mostly from South America.
The data is fairly evenly split between Instagram (58 items), X / Twitter (57 items), and YouTube (46 items). 90 of the items relate to Antisemitism and 71 relate to anti-Muslim hate. So far only one Instagram item and 2 YouTube items have been removed, no items have been removed from X but 4 items that are very clearly hate speech have had their reach limited because the “might be” hate.
A rmeinder that you can see some examples of the hate as well a statements from some of those who did the reporting.
This data was last updated at 5pm AEST / 9am CET, 18 June 2025.
Transparency reporting analysis
Our results are unsuprising in light of our past detailed works on antisemitism and anti-Muslim hate since October 2023. They also reflect a trend we are seeing reflected in the transparency reports from the companies themselves.
The three companies examined in this campaign all produce transparency reports to show how they are doing in addressing a range of harm to the public, including hate speech. In this background briefing we share data and analysis from these reports.
- Meta’s latest transparency report shows a 29% reduction in the number of hate speech items removed on Instagram for the first three months of 2025 compared to the same period a year ago.
- It also shows a drop in the amount of hate removed proactively based on AI, down from 98.2% (start of 2024) to 97.4% (start of 2025). (The drop on Facebook is even sharper down from 94.7% of items at the start of 2024 to 88.6% of items at the start of 2025. )
- From the above we can see the number of hate speech items automatically removed on Instagram has gone from 8.45 million at the start of 2024 to just 5.94 million at the start of 2025. That means over 2.5 million fewer items of hate are being detected by AI now compared to a year ago, a 29% reduction in effectiveness when comparing these two quarters.
X(Twitter)
- While X releases transparency reports every half year, they obfuscate the data around hate speech.
- In the first half of 2024 they provide data for hate speech, but provide only a combined total for the number of tweets they removed and the number that were labelled. In the second half of 2024 they disaggregated the number removed from the number labelled, but merged the data for hate speech with the previously separately reported data for Abuse & Harassment. Both approaches make it impossible to determine how many items were actually removed for hate speech.
- According to Twitter’s transparency report for the second half of 2021 104,565 accounts were suspended for hateful conduct. In the most recent report for the second half of 2024 we can estimate the number of accounts suspended for hate speech has dropped to just 2009. That represents a 98.1% reduction compared to 2021.
- According to Twitter’s transparency report for the second half of 2021 1,293,178 items of content were removed for being hateful conduct. Based on the most recent two reports, we estimate about 362,116 to 546,354 items of hateful conduct were removed in the second half of 2024. This is a reduction in the removal rate of between 72% and 58% since the second half of 2021.
- In comparing the data with 2021 it is important to remember 2021 was during Covid when platforms were showing a significantly reduced capacity of dealing with online hate due to a surge in platform use, and reduced staffing capacity.
YouTube
- YouTube provides the most consistent and usable transparency reporting with a specific statistic for content removals and a category for Hateful or abusive content and the ability to compare them across different reporting time periods.
- In January to March 2025 the number of hateful or abusive items removed was 54,096. That is an increase compared to 37,473 items removed under this category in January to March 2024.
- For comparison with X, in the second half of 2024 YouTube removed 89,654 items, compared to 78,844 in the second half of 2021, an increase of 13.7%.
Conclusion
Addressing online hate speech has become more difficult due to decisions of some tech companies to alter their approach to tackling hate speech. X has dramatically reduced actions removing hate speech and closing accounts dedicated to hate speech since Elon Musk bought the company and has greatly reduced its transparency. Our estimates suggest a reduction of 98.1% in account closures and a reduction of between 72% and 58% in content removals for hate speech on X compared to some years ago. On Instagram Mark Zuckerberg’s efforts to recalibrate and allow more hate speech to remain online are starting to show with a 29% reduction in the automatic removal of hate speech compared to a year ago. YouTube continues to improve its efforts.
Our data shows the same trend, with YouTube being more proactive than other platforms. It is the only platfrom to immediately email users a supportive message after they make each report. Despite less content being reported to YouTube, in absolute terms it leading on content removals at the time of writing.