New York – Millions of anti-Semitic messages on Twitter have spread negative stereotypes and conspiracy theories about Jews across the social media platform, according to a report Monday by the Anti-Defamation League.

ADL national director and CEO Jonathan Greenblatt said the data showed many used Twitter as a “megaphone to harass and intimidate Jews.”

An earlier report from the Jewish civil rights group said anti-Semitic incidents in the U.S. last year had reached the highest tally it has counted in more than two decades. That increase appeared to be fueled by emboldened far-right extremists as well as the “divisive state of our national discourse,” Greenblatt said in February.

“Of course, 4.2 million tweets is a very small number out of the trillions of tweets sent on the platform each year,” the report said. “But that does not negate the lived experience of Jews who have found Twitter to be a toxic environment.”

Twitter says it has made more than 30 changes to its platform, policies and operations in the past 16 months to protect its users from abuse and hateful images.

“We are an open platform and hold a mirror up to human behaviors, both the good and the bad,” the company said in a statement. “Everyone has a part to play in building a more compassionate and empathetic society, including Twitter.”

The New York City-based ADL said it used a complex query of code words and symbols, statistical methods and expert analysis to develop this first-ever “snapshot” of anti-Semitic trends and themes on Twitter. A human review of the messages weeded out sarcastic expressions or tweets using anti-Semitic language to condemn it, the report said.

The report’s definition of anti-Semitic content included criticism of Israel or Zionism “when such criticism makes use of classic anti-Semitic language or conspiracy theories, or when it ascribes evil motivations to significant numbers of Jews.”

The report included a set of policy recommendations for Twitter but said the company already made “real progress” in fighting online hate and harassment.

Although the study focused on Twitter, the report noted many tweets shared or discussed anti-Semitic content on other platforms, such as YouTube, Reddit and 4Chan.

“We hope this report will create a renewed sense of urgency among all social media providers that this problem is not going away and that they need to find innovative new ways to tamp down the spread of hatred online,” Greenblatt said in a statement.

The ADL counted a weekly average of 81,400 anti-Semitic tweets.

It found eruptions of anti-Semitism around certain news events, including an October 2017 report by the New York Times on sexual harassment and assault allegations against movie producer Harvey Weinstein, who is Jewish.

“The story touches on multiple long-standing anti-Semitic tropes, including Jewish control of Hollywood and the media, and Jewish sexual degeneracy and perversion,” the report said.

In August, Twitter “exploded” with posts about anti-Semitic and racist displays at the white nationalist rally in Charlottesville, Virginia, the report said. But a human analysis showed less than 9 percent of them actually promoted anti-Semitism.

Holocaust deniers promoted then-White House spokesman Sean Spicer’s remark last year that Adolf Hitler “didn’t even sink to using chemical weapons,” a statement ignoring Nazi gas chambers, the report said.

“Although Spicer quickly clarified that he was not referring to the death camps in which Jews were killed, Holocaust deniers stuck to their own script,” the report says.

The report also cited conspiracy theory tweets blaming “the hidden hand of the Jews for many of the world’s worst tragedies and disasters.”

The ADL urged Twitter to expand content filters to let users screen out “hateful and extremist propaganda.”

“While Twitter’s offensive content filter is a start, it typically applies only to violent imagery,” the report said.

The group also recommended Twitter use “artificial intelligence” to flag hate-filled content.

“Machine learning can allow AI to integrate insights and develop more capable understanding of how to sort out truly hateful content,” it said.