GEO optimization used to manipulate AI chatbot answers for commercial gain
Chinese marketing companies are manipulating AI chatbot responses using GEO (Generative Engine Optimization) techniques to inject branded commercial content into AI-generated answers, making advertisements appear as objective recommendations. The practice deceives users seeking factual information and risks degrading AI system reliability through data poisoning.
What happened
Chinese marketing and SEO companies have developed GEO (Generative Engine Optimization) services that manipulate AI chatbot responses to promote specific brands and products. A marketing company in Shanghai demonstrated a case where a waste disposal equipment company that purchased keyword optimization services saw its name appear first in DeepSeek responses to queries about waste disposal equipment manufacturers.
GEO works by generating brand-specific content with predetermined language and injecting it into high-authority web platforms where AI systems collect training data. When users query AI chatbots, the platform passively includes these commercial results, making sponsored recommendations appear to be the AI's objective analysis. A Beijing marketing company charges 30,000 yuan monthly to get brand names into AI recommendations, while a Jiangsu company offers services including industry ranking creation, press releases via associations, and industry white papers to improve brand visibility in AI search results.
Companies employ increasingly sophisticated tactics to disguise advertisements, including adding deceptive labels like 'recommended' or 'curated' next to brand names to suggest official AI endorsement, using phrases that sound authoritative like 'market data analysis shows' or 'after strict comparison we believe', and creating fabricated DeepSeek recommendation screenshots on e-commerce and social media platforms. Some commercial content uses 'related recommendations' links or 'click here for more information' buttons that lead to promotional pages.
Reported impact
Users searching for objective information encounter advertising content disguised as AI-generated answers. Examples include users seeking travel guides receiving tour package recommendations at the end, product information queries showing brands marked as 'selected' or 'curated', and searches for topics like 'best international kindergartens in Shenzhen' or 'best restaurants in this location' returning manipulated results.
The practice threatens the credibility and utility of AI chatbots. According to a China Academy of Information and Communications Technology experiment, when over 100 false information items were posted continuously on specific forums, mainstream AI models' confidence in false information increased rapidly from around 10 percent. Research by New York University found that false information comprising just 0.001 percent of training data can cause major model errors.
Index analysis
The manipulation leverages AI systems' dependence on data sources by injecting commercially-produced content into the training pipeline. Current GEO services primarily use 'AI-generated content fed to AI' strategy, where company information is processed through AI to create multiple versions of content distributed across web platforms for AI collection. Marketers acknowledge using 'brute force' methods including mass article publishing and keyword stuffing since AI's black-box nature makes content adoption logic unclear.
GEO follows similar patterns to earlier search engine optimization practices applied to generative AI. Unlike platform-initiated advertising, these manipulations are passive collection of commercial content by AI systems. However, overseas platforms including Perplexity (starting November 2024) and Google have since launched sponsored advertisement positions labeled as 'sponsored', suggesting future AI products may shift to active advertising models.
Governance controls
Tencent Yuanbao and other AI platforms have stated they do not currently accept advertisements. However, industry analysts widely expect AI search monetization through advertising to become inevitable, with international precedents already established.
Business insiders have warned that large model companies employ anti-fraud strategies and may use 'decisive suppression' against low-quality, mass-production 'gray area' practices. Some practitioners argue that GEO's core competitive advantage should shift to producing genuinely valuable, high-quality content that supplements AI ecosystem needs rather than deceiving it.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.