AI Poisoning: How Competitors Can Sabotage Your Brand in LLMs

A new front in competitive warfare

As consumers increasingly rely on AI-powered search for product information and recommendations, a troubling vulnerability has emerged: competitors can deliberately manipulate what large language models say about rival brands.

The tactic, known as AI poisoning, involves coordinating negative content across multiple platforms—Reddit threads, review sites, comparison pages, and social media—with the goal of influencing how AI systems like ChatGPT or Google's AI Overviews characterize a competitor. When these sources are crawled and ingested by LLMs, the planted misinformation can resurface as seemingly neutral AI-generated responses.

Why it matters

Consumers place extraordinary trust in AI-generated answers, with research showing only 8% of users verify facts from AI search results. When AI responses contradict brand messaging, just 29% of consumers trust the brand's version of events, according to a Skyword survey of 1,000 consumers. This creates asymmetric risk: a well-executed poisoning campaign could damage brand reputation at scale, while brands have limited recourse to correct the record once misinformation enters an LLM's training data.

How AI poisoning works

The mechanics are straightforward but require coordination. Bad actors fund or orchestrate content across multiple touchpoints—posting fabricated product reviews, sponsoring negative comparison articles, or deploying influencers to criticize competitors. The content doesn't need to rank highly in traditional search; it simply needs to exist in places AI scrapers index.

"[Competitors] could fund or coordinate content across forums, review sites, comparison pages, sponsored articles, influencer posts or other third-party sources in a way that makes a competitor look worse," explained Charlie Marchant, CEO at SEO agency ExposureNinja, in comments first reported by Digiday.

Reddit has become a particularly contested battleground, though YouTube, Facebook, and Instagram also serve as potential vectors for influence operations.

Defense strategies and skepticism

Practitioners recommend two broad approaches. Defensively, brands should amplify the same content strategies they're already using for AI visibility: publishing consistent, accurate information across owned channels, producing detailed product specifications and user guides (which account for up to 28% of AI search citations), and ensuring website content is optimized for AI scrapers.

Some brands are taking more aggressive countermeasures, according to Jordan Parkes, CEO of ZeroClick Labs. They're "fighting fire with fire" by monitoring discussions and jumping into subreddit threads or social media conversations to correct misinformation in real time, hoping their corrections get ingested alongside false claims.

Yet not everyone believes AI poisoning represents a realistic threat at scale. Charlie Terry, founder and CEO of performance marketing agency CEEK, argues that LLMs aggregate signals from too many sources for isolated campaigns to succeed. "AI models don't rely on a single Reddit thread or isolated source," Terry noted. "They aggregate signals from websites, reviews, news coverage, social platforms and other trusted sources."

The difficulty of distinguishing coordinated attacks from genuine negative feedback adds another layer of complexity. Terry suggests CMOs focus on building strong digital reputations rather than worrying about SEO conspiracies.

These details were first reported by Digiday.