Instagram's Algorithm Serves Antisemitic Content to Millions, CAM Finds

Instagram's recommendation algorithm actively delivers antisemitic content to millions of users, according to new research published by the Combat Antisemitism Movement (CAM) and distributed via GlobeNewswire.

The finding shifts the debate from moderation gaps — hateful content that slips through review — to a harder question: whether the platform's own discovery systems amplify that content to audiences who never searched for it. For practitioners, that distinction matters. Algorithmic amplification is an ad-adjacent and brand-safety problem in a way that buried policy violations are not.

CAM, an international organization tracking antisemitism trends, says its researchers documented a pattern in which Instagram's suggestion machinery pushed antisemitic material to large audiences. The phrase "feeds antisemitic content to millions" in the research announcement frames the scale of exposure in user-reach terms, not just in takedown counts — the metric platforms typically cite when defending their enforcement record.

Meta has faced this category of criticism before. Its own researchers concluded in 2020 that the Facebook recommendation engine had pushed users toward increasingly extreme content, and the company has since rebranded its AI-driven feed ranking around "meaningful social discovery." The CAM research suggests Instagram, Meta's flagship property for younger demographics, has not resolved the underlying dynamic when it comes to antisemitic material.

The implications for marketers and creators run in two directions.

First, brand safety. If recommendation surfaces — Reels, Explore, Suggested Posts — can carry antisemitic content to millions, advertisers buying placements adjacent to those surfaces inherit the risk. Brands that paused spending on X over similar amplification concerns now face a comparable question about Instagram inventory, which commands a larger share of most social budgets.

Second, organic reach integrity. Creators and publishers building audiences on Instagram depend on the same recommendation systems the research implicates. If those systems reward engagement patterns associated with extreme content, legitimate publishers compete inside a ranking environment that the research says boosts hateful material. That affects distribution strategy, content-mix decisions, and how teams benchmark reach performance against paid spend.

The research also arrives amid regulatory pressure on algorithmic amplification in the EU under the Digital Services Act, which requires very large platforms to assess and mitigate systemic risks in their recommendation systems. Findings from groups like CAM feed directly into that oversight process, and Meta risks enforcement action if regulators accept the exposure claims.

CAM has not said whether it shared the underlying dataset with Meta before publication, and Meta had not issued a public response in the initial announcement. Researchers and platform critics have repeatedly argued that Meta's published enforcement statistics understate amplification because they count removals, not impressions — a gap this research appears designed to expose.

Expect Meta to dispute the methodology, question the sample, or point to its investments in AI detection — the standard playbook for such reports. Until the company releases comparable impression-level data of its own, advertisers and creators will have to weigh CAM's reach claims against Meta's assurances. Watch for follow-up data from CAM and any EU regulatory response in the coming weeks.