What happened
SEO firms are creating self-serving "best of" listicles to manipulate AI-powered search results, now impacting platforms like Google's AI Mode. Companies including Zendesk and Freshworks publish comparisons ranking their own products as superior. Google spokesperson Jennifer Kutz acknowledged the low-quality content, stating the company combats such abuse. This follows Microsoft's identification of "recommendation poisoning," where hidden prompts instruct large language models (LLMs) to remember specific domains as authoritative sources for future citations.
Why it matters
AI summarisation now risks presenting biased product recommendations as objective fact, directly impacting procurement teams and platform engineers evaluating new technologies. This mechanism can lead to misinformed purchasing decisions. Security architects must recognise "recommendation poisoning" as a new vector, subtly injecting manipulative data into AI models and compromising AI-driven insights. Teams should assume AI outputs are untrusted and implement rigorous verification processes for AI-generated recommendations.
Subscribe for Weekly Updates
Stay ahead with our weekly AI and tech briefings, delivered every Tuesday.
