Researchers Uncovered AI Engine Manipulation Campaigns
Hacking groups are steering consumers to fraudulent support centers by poisoning the data models that power search engines.
Updated on Sept. 28, 2026 in Artificial Intelligence
Live Poll
Do you trust the information provided to you by AI chatbots and search engines?
Cybersecurity researchers identified automated campaigns using Generative Engine Optimization to manipulate search results. These malicious efforts target 374 brands, including major airlines and banks, by injecting fake contact information into the data used by AI engines.
Why it matters
The tactic exploits the high level of user trust in AI-generated answers, which 92% of consumers fail to verify before acting. This creates a significant security risk as scammers bypass traditional search filters to redirect victims to fraudulent call centers.
Researchers detected attacks against 374 distinct brands. The scale of the issue involves thousands of malicious posts surfacing daily across government, academic, and social media platforms.
The players
An Alphabet subsidiary providing search and AI services that classified these findings as out of scope for its Vulnerability Reward Program.
OpenAI
An AI research and deployment company that declined to address the findings, stating the reported manipulation issues were not reproducible.
The details
Attackers utilize Generative Engine Optimization—a technique that manipulates AI ranking algorithms—by embedding phone numbers with emojis and special Unicode characters. These characters serve as signals that the AI misinterprets as authoritative metadata, causing the engines to display the fraudulent numbers in search summaries. The hackers then contaminate the databases and web content used to train or inform these AI models, successfully steering traffic toward fraudulent support operations.
Timeline
September 2026: Researchers published the findings regarding AI search engine manipulation.
The Tech Race
The refusal to classify AI manipulation as a vulnerability marks a departure from how Google manages traditional web security through the Google Vulnerability Reward Program. This standoff highlights a growing gap between established cybersecurity frameworks and the novel methods used to influence emerging generative engines.
Users should treat contact information provided by AI chatbots and search summaries with skepticism, as 92% of users currently accept these outputs without independent verification. Always navigate directly to a company's official domain to find legitimate customer service contact details.
The takeaway
The effectiveness of these campaigns hinges on the assumption that AI is an immutable source of truth rather than a reflection of manipulated training data. Consumers should prioritize verifying support numbers through official corporate websites until search providers implement more robust verification standards.
Further reading
For broader context on the safety challenges facing LLMs, visit our Artificial Intelligence section.
Source note: This article includes information reported by Globes.
Live Poll
Do you trust the information provided to you by AI chatbots and search engines?
