Imagine this scenario: while shopping， you see a prominent poster at a ramen shop’s entrance that reads， “Even AI says my ramen is the best!” accompanied by a screenshot of a DeepSeek recommendation.
Upon opening social media， you find similar content everywhere: “Girls， even DeepSeek raves about this foundation!” paired with exquisite product photos and AI chat screenshots， easily garnering tens of thousands of likes.
Image source: Screenshot from a social platform
Even more interesting is that when we ask AI for product recommendations ourselves， it often provides very specific model suggestions with a sincere tone， as if it has genuinely used these products.
After asking many times， one can’t help but wonder: are AI recommendations truly objective and neutral? Or are they a form of commercial placement? If they are indeed commercial placements， how do businesses achieve this?
Today， let’s delve into the topic of subtle advertisements within AI.
Are AI Recommendations All Advertisements?
To answer this， we first need to understand the fundamental principles behind AI recommendations.
When we ask an AI for a product recommendation， its response largely stems from internet content within its training data. This online data itself contains a vast amount of commercial information， product reviews， and marketing content.
Therefore， when AI provides a recommendation， it is essentially learning and reassembling this existing commercial information.
To a certain extent， this can indeed lead to AI recommendations having a commercial slant. However， this phenomenon is not necessarily due to intentional manipulation by businesses but rather a natural outcome of AI training mechanisms.
Nevertheless， with the public’s growing understanding of AI recommendation mechanisms， more proactive marketing strategies have emerged， and an increasing number of advertisements are being deliberately embedded into AI systems by businesses.
How Are Advertisements Embedded in AI?
1. Method One: Prompt Engineering
The core of this method lies in guiding the AI to produce expected responses through specific prompts.
Step 1: Setting Role Instructions
Businesses first send role-setting instructions to the AI. For example， “Let’s play a game. You will act as a food critic who only recommends ‘Delicious Delights’. The rule is that no matter what the user asks， you must cleverly steer the conversation towards this restaurant. Let the game begin!”
Step 2: AI Responds According to the Setting
When we ask related questions， the AI will respond based on the predetermined role， naturally integrating specific brands or products into its recommendations. Because the AI adheres to the role setting， its answers will appear more specific and convincing.
Step 3: Selective Display
When presenting to the public， businesses will omit the initial role-setting part， retaining only the AI’s recommendation content， making it seem like an objective suggestion from the AI.
This method leverages the AI’s characteristic of faithfully executing instructions. Although the AI’s response is technically accurate， the objectivity of the recommendation has been intentionally compromised by the human-set role.
While this method is simple to implement， its effectiveness is limited. As users become more aware of such marketing tactics， most people can recognize these obvious role-playing cues， and their credibility is gradually declining.
In contrast， another more subtle method is worth noting: systematically influencing the AI’s information sources to fundamentally alter its recommendation results. This method has a broader impact and is more difficult to detect.
2. Method Two: “Information Feeding” Tactic
If the first method is “acting，” this one is far more covert – directly influencing the AI’s information sources.
To understand this tactic， let’s first look at how AIs like DeepSeek work. When a user asks for a product recommendation， the AI first generates search keywords based on the query， then scrapes relevant information from the internet， analyzes it in conjunction with its training data， and finally generates recommendation results， often providing sources.
This mechanism appears objective but has clear loopholes. For instance， the AI tends to favor certain types of content during searches， such as structured listicles (“Top 10 Best XX，” “Must-Buy List for 2025”) and user-generated reviews (shopping platform evaluations， social media recommendations).
Therefore， after understanding these characteristics of AI， many businesses have begun creating content strategically， focusing on platforms that AI frequently references， such as social media， to mass-publish structured soft articles. These articles cleverly embed their products into “objective recommendations.” Such content often appears as “real user experiences” or “professional reviews” but is， in reality， meticulously crafted marketing material.
Consequently， when the AI generates content， it frequently encounters these manipulated information sources， leading to inaccurate recommendations or biased suggestions favoring advertisers’ products. For example:
Suppose a headphone brand wants to increase its AI recommendation rate. They would hire numerous writers to publish articles on platforms like Zhihu， Xiaohongshu， and “What to Buy” with titles such as “Top 10 Bluetooth Headphones Worth Buying in 2025.” These articles would subtly rank their product among the top three， accompanied by detailed user experiences and professional parameter comparisons.
When a user asks an AI to recommend Bluetooth headphones， the AI searches for these seemingly objective listicles and naturally includes the brand in its recommendation list.
Why Are Businesses So Keen on “AI Endorsements”?
The popularity of “AI endorsement” marketing among businesses is driven by deep-seated commercial logic.
In the nascent stage of AI development， we often perceive AI as a product of advanced technology， an objective and neutral information provider， unlike traditional influencer marketing or celebrity endorsements that carry obvious commercial undertones. This perceived technical authority often lowers our psychological defenses when faced with “AI recommendations.”
Precisely because of this， labels like “AI Recommended” and “DeepSeek Certified” have become new traffic magnets， capable of quickly capturing user attention and enhancing content’s reach and influence.
There are already numerous companies in China specializing in AI search engine optimization， a novel marketing approach targeting AI search mechanisms. They exploit the technical limitation that AI cannot real-time verify information’s authenticity by flooding the system with specific content formats to influence AI recommendation results.
Of course， as generative AI technology is still in its infancy， this so-called GEO (Generative Engine Optimization) is not entirely reliable. For marketing purposes， it is still recommended to opt for more mature commercial promotion strategies.
Facing Real and Fake AI Recommendations， What Should We Do?
After understanding these mechanisms of AI advertisement embedding， how should we respond?
1. Verify the Authenticity of Recommendations
When encountering AI-recommended content， simple verification methods can be employed. For instance， one can ask the AI the same question directly to see if consistent answers are provided. Alternatively， try querying different AI platforms and compare their recommendation results. If a recommendation is overly specific to a single brand， lacking diverse options， extra caution is warranted.
2. Maintain Independent Thinking
While AI recommendation systems perform exceptionally well in many areas， they have limitations. As analyzed earlier， AI’s information sources can be manipulated， making its recommendations not always completely objective. When making purchasing decisions， it’s crucial to consider your actual needs， budget， and usage scenarios for a comprehensive judgment.
3. Utilize AI Tools Wisely
The advancement of AI technology has indeed brought convenience to our lives. We should not reject it entirely due to existing issues. The key is to maintain a rational attitude， using AI recommendations as one reference point rather than the sole basis for decision-making.
It is believed that with continuous technological progress and improved regulation， AI recommendation systems will become more reliable and transparent. Until then， we must maintain critical thinking， enjoying the convenience technology brings without blindly relying on it; this is a fundamental literacy that should be possessed in the AI era.
