AI Propaganda in 2026: How to Spot AI-Generated Fake Content

What is AI propaganda?

AI propaganda is the use of artificial intelligence to create, scale and disguise persuasive or false content meant to shape public opinion. Think fake articles, invented personas, bot networks and AI-written posts made to look independent.

The UK government has called it "an industrial-scale assault by hostile actors on the information environment", according to the RMIT Information Integrity Hub's September 2026 edition of The Repost.

How are states and companies using AI for influence operations?

Anthropic's latest threat report describes nine influence operations built with its Claude chatbot. Three stand out:

• Iran. Government accounts ran "cognitive warfare" campaigns. They attributed false claims to Western research institutions and turned intelligence bulletins into posts from seemingly independent accounts.
• Russia-linked user. Daily news stories with pro-Russia, anti-France messaging went through a local radio station in the Central African Republic and onto the national broadcaster.
• A French advertising firm. It ran around 70 fake news websites and hundreds of inauthentic X accounts, backing different political sides depending on who was paying.

Operators also used AI to write training manuals, personas and opposition dossiers, translate content, and strip signs of state origin from republished material. That kind of fake article is exactly what AI text detection is built to flag.

What are AI swarms?

Swarms are groups of autonomous AI agents that open their own social media accounts and post without human oversight. The New York Times reported that Iran, China and groups in Israel released hundreds of them, filling accounts with false posts about politics and current events.

Open-source models can be downloaded and modified, so safeguards can simply be switched off. Experts caught these agents easily. The next wave may be harder to spot.

Can AI chatbots spread propaganda?

Yes. It's the risk most people miss.

• pattrn.ai found most chatbots can be co-opted for information operations. Integrity scores ranged from 93% down to 12%, and Claude was the most likely to refuse.
• Reporters Without Borders found leading chatbots quoting sanctioned Russian state media when asked.
• NewsGuard found 15% of chatbot answers to false-claim prompts contained falsehoods, with higher rates for claims about the US-Iran war.
• A 2026 study in Nature found that state control of media leads to more government-friendly chatbot answers.

Russia-linked "Pravda" aggregator sites push Kremlin talking points into the sources chatbots cite. China is also reported to be using AI-simulated audiences to test fabricated, BBC-style coverage before release, according to ASPI analyst Fergus Ryan.

What is "slop jihad"?

The Institute for Strategic Dialogue uses the term for AI-generated memes, fan edits and cartoons that repackage extremist ideology for younger audiences. Trolls and genuine extremists both make them.

Wired reported that AI tools let a wider, more decentralised set of users remix extremist imagery, and that engagement can push algorithms toward more overt propaganda. Spotting these fakes starts with AI image detection.

What else happened in September 2026?

• Meta began testing Community Notes in 16 Spanish-speaking Latin American countries. Fact checkers rely on Meta for roughly 35% of their funding on average.
• Australia released a draft "digital duty of care" law that would let users opt out of algorithmic recommendations.
• BBC Verify reported that Iran's supreme leader's social accounts again used a profile image very likely generated or manipulated by AI, per Google's SynthID detector.
• The Guardian found 39 submissions to Australian parliamentary inquiries with hallucinated references.
• Coordinated campaigns made up 61% of the ABC's content complaints last financial year.

How do you check if content is AI-generated?

Gut instinct won't hold up. Even journalists and parliamentary inquiries are being fooled or overwhelmed. Manual checking can't match content produced at machine speed.

Five habits help:

• Check who owns the outlet.
• Look for identical wording across "independent" accounts.
• Distrust viral posts that provoke outrage without a named source.
• Cross-check with professional fact checkers.
• Run text, images and video through a detector. The UncovAI browser extension does it while you scroll.

FAQ