Most exposure to antisemitism is in the online sphere, which calls for a different approach to confronting anti-Jewish hatred, a new report by the American Jewish Committee found.
Of the 45% of Americans who saw or heard antisemitism in 2025, about three-fourths saw or heard it online or on social media — by far the highest place.
Half of social media users experiencing antisemitism online do not report these incidents to social media platforms because they don’t think anything will be done.
“The State of Antisemitism in America 2024-2025: Findings and Recommendations for Major Digital Platforms,” released in late March in partnership with Israeli independent nonprofit organization CyberWell, is the first publication of its kind to explore American Jews’ experiences with antisemitism across social media over time, according to Holly Huffnagle, AJC’s director of antisemitism policy.
The findings of both AJC and CyberWell weren’t surprising to Huffnagle, who works out of AJC’s Washington, D.C., office.
“We’ve known about these trends for some time,” she told Washington Jewish Week, adding that the AJC has collected data on American Jews’ experiences with antisemitism in real life and online for the past seven years.
New over the past few years is an increase in antisemitism on social media. This new report is a “culmination of that research to this tipping point that we’re at now, where most exposure to antisemitism is in the online space,” Huffnagle said.
But just because it’s behind a screen doesn’t negate the danger.
One in five American Jews who have seen or heard antisemitism online indicated that the incident made them feel “physically threatened,” the report stated. More religiously affiliated Jews reported feeling more physically threatened by online antisemitic incidents than those who are more secular.
More American Jews are turning away from some social media platforms altogether due to the regularity of this hatred.
Many of the social media companies the AJC works with have taken steps to try to make their platform safer for users, especially those who are Jewish, but Huffnagle said there are still gaps.
“This report showcases those gaps,” she said.
“In our community, where international policy makers and government officials make critical decisions that impact the security of our community, we must make sure they consider the battlefield of the virtual world in the struggle against hatred and disinformation targeting Jews,” Alan Ronkin, AJC Washington, D.C.’s regional director, told Washington Jewish Week.
“We don’t protect people in the online space like we do in real life, and yet you see that link now more than we’ve really ever seen it before,” Huffnagle said.
Because online antisemitism has real-life consequences — many violent, even fatal, ethnic attacks in the United States began online with the dissemination of antisemitic or racist rhetoric, Huffnagle said — the AJC report puts the onus of user safety on social media and digital platforms.
AJC and CyberWell collaborated to share recommendations for companies based on their research, including establishing new, clear policy guidance, strengthening user safety protections and addressing evolving forms of antisemitism, such as artificial intelligence-driven antisemitic content.
“One of our asks is that every single [social media] company looks at each of these nine recommendations and say[s], ‘Are we doing this to the best of our ability? Are we protecting users to the best of our ability?’” Huffnagle said.
“Evolving forms of antisemitism” includes emojis, symbols or coded language to target Jews while evading moderation. One popular example is the use of a juice box emoji, which phonetically resembles the word “Jews.”
Because the juice box emoji, for instance, doesn’t automatically equate to antisemitism, the report emphasizes the importance of examining the connotation of its use.
Telling Jewish users to “take a shower” can be coded antisemitism, according to Huffnagle and the report.
“Most AI models wouldn’t know that ‘taking a shower’ is referencing the Holocaust and Jews being murdered,” she said. “But how do you start training a model not only to recognize that in English as an incitement to violence or as a threat, but also in other languages, which might be completely different? You have these terms or phrases that are referring to Jewish people that were just created in the last six years — how are models catching up to that?”
Another derogatory reference is “small hats” or “tiny hats,” alluding to the kippah. In a YouTube video, a speaker who accuses Jewish Israelis of staging the Oct. 7, 2023, attack says, “These small hats, they control everything. They run everything; they control the media.”
That October 2025 video, along with other antisemitic content, remained online despite escalation by CyberWell, according to the report.
“Antisemitism training for third-party human content moderators must include material and curriculum on how to recognize how emojis and coded words and phrases … are used to convey explicit and implicit antisemitic messages and circumvent moderation across cultures and languages,” the report stated.
Huffnagle recognized that the issues associated with short- and long-form video on YouTube largely differ from 15-second TikTok clips with overlaid text. The user bases are also very different in terms of age demographics.
“We realize that we need to be more targeted with our approaches,” Huffnagle said.
Much of this work is further complicated by the flood of anonymous social media users and fake accounts on platforms.
She and the AJC team work closely with CyberWell to provide training and discuss “creative solutions” to these issues.
Huffnagle sees this collaboration as especially important given the current congressional stalemate around social media regulation.
The report also suggests optimizing reporting systems and increasing transparency in content moderation and policy enforcement, with nine out of 10 American Jews indicating the importance of social media companies making it easy to report antisemitism.
“We need to start treating this online problem like an offline problem in ways that we really haven’t before,” Huffnagle said. “That’s [what] we want to convey.”
