March 29, 2023 - 7:00am

Since Elon Musk took over Twitter in October last year, there have been numerous articles claiming that “hate speech” on the platform has “surged”. To make this point, various kinds of evidence are cited.

The Center for Countering Digital Hate tracked the frequency of four slurs and their plurals, finding that all had increased substantially since Musk’s takeover. What’s more, engagement with tweets mentioning the particular slurs was also up.

Tufts University’s Digital Planet group examined the 20 most popular tweets containing racist, antisemitic or anti-LGBT content both before and after Musk’s takeover. They found that beforehand almost all the top 20 in each category were either quoting others or using words in a non-hostile way. Yet after the takeover, several in each category were genuinely hostile.

The Anti-Defamation League regularly reports antisemitic tweets to Twitter and keeps track of whether any action is taken. According to the New York Times, Twitter has gone from taking action on 60% of such tweets to taking action on only 30%.

Amnesty International carried out a survey in which they asked “LGBTQ+ activists” whether they had experienced an increase in abusive speech on Twitter since Musk took over. Among respondents, 60% said they had experienced an increase, with the remaining 40% saying they had experienced the same level as before; none had experienced a decrease.

Twitter is even facing a lawsuit in Germany over its failure to remove antisemitic content from the platform. In one case, the claimants argue that a request to remove a post denying the Holocaust was “explicitly rejected” (Holocaust denial being illegal in Germany).

All this would seem to constitute prima facie evidence that there has been an increase in “hate speech” since Musk took over Twitter. But of what magnitude and for how long? To try and answer these questions, I turned to StoryWrangler — a tool developed by researchers at the University of Vermont which allows you to track the frequency of words and phrases used on Twitter over time.

The chart below plots the frequency of the four slurs analysed by the Center for Countering Digital Hate (and their plurals). Each series was standardised to remove differences in overall usage. The red line is the average of the eight standardised series, while the dashed vertical line marks the date of Musk’s takeover.

On this timescale, there is little evidence of a major uptick in hate speech: slurs were used much more in earlier years than they are today. But perhaps there was a sharp increase around 27th October, one that will snowball with time? To check this, I did the same thing as before while limiting the analysis to the period comprising one month either side of Musk’s takeover.

There was a sharp increase, as can be seen above, but it only lasted for a few days. By 1st November, average usage of slurs was back to where it was on 26th October.

As Yoel Roth, Twitter’s former head of “Trust and Safety”, notes, a trolling campaign began shortly after the takeover was announced, whereby various accounts posted irreverent tweets to test whether they could use racial slurs. Many of these accounts were removed, which may help to explain why there was no sustained rise in the usage of such words.

I also checked the period comprising one month either side of 13th December – the date Musk dissolved Twitter’s “Trust and Safety Council”. But again, there was no sustained rise. There has been a moderate increase since the start of February, driven largely by the word ‘c*nt’. Its cause is unclear.

Tracking the frequency of slurs is not the only way of measuring “hate speech”, so my analysis cannot rule out that other kinds of hostile content have increased (e.g., image-based tweets). However, cases of in-your-face bigotry may be rarer than recent headlines suggest.


Noah Carl is an independent researcher and writer.

NoahCarl90