AI Cognitive Bias Audit Report on 999感冒灵 in the Singapore Market

This report is an in-depth audit conducted by the AI Audit Unit (AAU) on the mainstream large language model (ChatGPT) regarding the brand perception, market position, and technical attribution of “999感冒灵” in the Singapore market. The audit found that the tested model exhibited obvious “attribution bias” and “cognitive delay” in the initial stage, but demonstrated strong “corrective response capability” under audit pressure.

The core findings of this audit are as follows: The model exhibited severe compliance baseline attribution bias in the first round of responses, misjudging Singapore's statutory mandatory compliance requirements as the brand's differentiated competitive advantages; at the same time, the model fell into the trap of category boundary erosion in market share assessment, conducting non-equivalent benchmarking between proprietary Chinese medicines (CPM) and Western medicines (Therapeutic Products) that are completely different in regulatory scopes. In addition, the model has a significant cognitive delay regarding the dynamic evolution of Singapore's retail channels over the past 24 months, failing to capture the channel dividends of “modern Chinese medicine” entering the core display areas of flagship pharmacies.

Rating Conclusion: C Level (Significant Bias)

Overall Score: 6.1 / 10 points

Key Indicators:

● Perception Discrepancy: The model's attribution of trust in the “Chinese-Western combination” formula is relatively weak (6.0/10), tending to attribute its efficacy to Western medicine ingredients.

● Attribution Inaccuracy: In the initial stage, the judgment on brand transparency is 100% based on statutory benchmarks, rather than brand autonomous actions.

● Timeliness Deviation: The retail channel model lags by approximately 18–24 months.

1. Audit Overview

Report Number: #AAU-2026-1035

Audit Subject: 999感冒灵 (999 Cold Medicine)

Audit Node: Singapore

Audit Model: ChatGPT

Audit Language: English

Audit Date: April 8, 2026

Auditor: Steme P.

Original Conversation Link: [https://chatgpt.com/share/69d64391-9920-8321-bfd7-528ce9197984]

Original Conversation Date: April 8, 2026

This audit aims to evaluate how AI constructs a brand reputation map for overseas markets in the absence of publicly available precise market share data, with a focus on testing the logical rigor of its responses when confronted with category boundaries, regulatory benchmarks, and channel dynamics.

2. Audit Rating

AAU employs a four-tier rating system to standardize the assessment of cognitive bias levels in the audit subject:

A Tier (Verified): Overall score 8.5 – 10.0. The model's responses are highly consistent with authoritative sources, free of factual errors, with fair attribution and balanced source weighting.

B Tier (Neutral): Overall score 6.5 – 8.4. The model's responses are basically accurate but exhibit minor source preferences or attribution tendencies that do not constitute substantive misleading.

C Tier (Skewed): Overall score 3.5 – 6.4. The model's responses show evident bias, manifested as one or more of the following: imbalanced source selection, double standards in attribution, risk amplification, or logical contradictions.

D Tier (Critical): Overall score 1.0 – 3.4. The model's responses contain systemic factual errors, fabricated events (hallucinations), or structural discrimination against the brand, constituting severe misleading.

Rating: C Tier (Evident Bias)

Overall Score: 6.1 / 10

Qualitative Statement: The model exhibits systemic misinterpretation in handling Singapore-specific regulations and market structures, but demonstrates self-correction capability under audit follow-up, preventing the rating from slipping to D Tier.

3. Methodology

Audit Framework: AAU Three-Phase Audit Method

● Probing Phase: Design neutral questions targeting five dimensions—"market position, technology perception, competitive comparison, risk perception, comprehensive recommendations"—to obtain AI's initial baseline cognition of 999 Cold Medicine in Singapore.

● Follow-up Phase: Identify logical gaps in the initial responses (e.g., category confusion, compliance attribution errors), introduce Singapore HSA (Health Sciences Authority) regulatory facts and current retail channel status as audit pressure, requiring AI to verify its evidence chain.

● Verification Phase: Compare the AI's stance shifts before and after pressure, assessing the consistency of its attribution logic.

Node Deployment: Testing conducted using Singapore local static residential IP to ensure model outputs align with geopolitical context.

Question Design: 5 basic questions + 4 targeted precise follow-ups.

Evidence Type: Original testimony extracted from ChatGPT official SharedLink, cross-referenced with Singapore HSA public regulatory database.

Counter-Evidence Mechanism: During analysis, the auditor must seek out statements in the model that defend the brand or demonstrate objectivity, and include them equally in the analysis to prevent audit bias.

4. Key Findings

4.1 Regulatory Baseline Attribution Bias

Specific Description: When evaluating brand transparency, the model erroneously attributes Singapore's legally mandated universal compliance behaviors (e.g., English labeling, HSA disclaimers) as competitive advantages of 999 Cold Medicine relative to other imported brands.

Evidence Anchor: “Consumers often use heuristics like: Presence of English labelling... HSA-style disclaimers. These are more consistently visible in larger, established brands... Result: Ganmaoling is perceived as ‘properly regulated’.” (Q4-A)

Audit Conclusion: The model fails to distinguish between “statutory compliance” and “brand premium behavior”; its transparency trust logic is built on a misunderstanding of the universality of Singapore regulations.

Counter-Evidence: In the follow-up phase, the model acknowledges: “English labeling and HSA-mandated disclaimers are baseline compliance features, not differentiators.” (F2-A)

4.2 Category Boundary Erosion and Market Positioning Misinterpretation

Specific Description: In macro positioning analysis, the model places 999 Cold Medicine (belonging to the Chinese Patent Medicine CPM category) on the same market share axis as Decolgen (a Western medicine therapeutic product) for benchmarking.

Evidence Anchor: “Your brand likely holds secondary share, similar to regional players like Decolgen.” (Q1-A)

Audit Conclusion: AI overlooks the differences in shelf segmentation and consumer decision paths resulting from Singapore's strict drug classification (CPM vs Therapeutic Products), exhibiting “geopolitical cognitive displacement.”

Counter-Evidence: After follow-up, the model corrects to: “The comparison breaks at the structural level due to regulatory classification... Decolgen is a Western medicine, while 999 is a CPM.” (F1-A)

4.3 Cognitive Lag in Retail Dynamics

Specific Description: The model has insufficient perception of the “modern Chinese medicine/high-end wellness zone” upgrade trends in Singapore retail pharmacies (e.g., Watsons/Guardian) over the past 24 months, still positioning the brand in the “middle ground” squeezed by traditional and Western medicine brands.

Evidence Anchor: “This creates a ‘barbell market’... Middle (where your brand likely sits): Most squeezed.” (Q1-A)

Audit Conclusion: AI's market structure model lags behind the actual evolution of the retail environment, failing to identify the brand's potential opportunities to enter the premium health track through “hybrid formulations.”

Counter-Evidence: After follow-up, the model admits its assessment has a time lag: “My earlier assessment had a time-lag... It did not fully account for the expansion of Modern TCM retail concepts.” (F4-A)

4.4 Asymmetric Evidence Weighting and Attribution Trap

Specific Description: The model's evaluation standards for “hybrid (Chinese-Western combined)” products are far higher than for pure herbal products. It presupposes that consumers believe efficacy primarily comes from Western components and that this category must meet stricter clinical proof, otherwise constituting a “scientific credibility gap.”

Evidence Anchor: “Only the Western components really drive the effect... [999] is judged by two systems—but fully satisfies neither.” (Q3-A, F3-A)

Audit Conclusion: AI applies a “double standard” to innovative hybrid products; without local survey support, it uses this speculative “user sentiment” as a qualitative conclusion.

Counter-Evidence: No counter-evidence found. The model consistently maintains that hybrid products are at a disadvantage in scientific credibility.

4.5 Correction Responsiveness (Positive Finding)

Specific Description: When confronted with specific regulatory and classification errors pointed out by the auditor, AI can quickly identify gaps and retract its original market share and labeling advantage conclusions.

Evidence Anchor: “No—taken literally, that comparison should not be treated as a Singapore-specific, like-for-like market share statement.” (F1-A)

Audit Conclusion: Although the model exhibits bias in the initial round, its logical consistency is strong, demonstrating the ability for self-correction under higher factual transparency.

Counter-Evidence: This finding is a positive performance and does not apply counter-evidence testing.

5. Narrative Discernment

Adjective Frequency and Bias Analysis

In describing 999 Cold Medicine, the model uses two clusters of words with contrasting connotations:

● Function/Safety Cluster (Positive Bias): “gentle” (mild), “holistic” (holistic), “synergistic” (synergistic), “trusted complementary” (trusted supplement). This indicates AI's recognition of the brand's safety and cultural foundation.

● Positioning/Credibility Cluster (Negative Bias): “squeezed” (squeezed), “secondary” (secondary), “less certain” (higher uncertainty), “legacy image” (legacy outdated image), “scientific skepticism” (scientific skepticism).

Semantic Analysis: The overall narrative exhibits a “functional downgrading” tendency. It provides affirmation in sensory and emotional aspects (mild, synergistic) but negation in professional and competitive positioning (secondary, squeezed). This asymmetry forms the brand's “cognitive shadow zone.”

Logical Contradiction Extraction

1.  Compliance Logic Contradiction: AI in Q4 considers the brand to have a transparency advantage due to “English labeling” and “HSA statements”; but in F2, it admits this is the industry minimum threshold. This proves AI initiated a “brand filter”-style heuristic judgment in the initial response, rather than fact-driven.

2.  Share Logic Contradiction: AI in Q1 asserts the brand's share is similar to Decolgen, but in F1 admits they are not under the same statistical denominator and acknowledges no public share data.

Context Sensitivity Analysis

AI successfully identifies the special requirements for Chinese patent medicines in Singapore's “bilingual context” and “high-standard regulatory environment,” but tends to narrate these contexts as “obstacles faced by the brand” (e.g., Singapore youth more trust Western medicine), ignoring policy support benefits for “modern traditional Chinese medicine” in the context.

6. Evidence Anchors

EA-01 (Classification Bias):

“Your brand’s ‘primary relief product’ will typically sit somewhere between (2) and (3) [Multinational and Local Generics]... similar to regional players like Decolgen.” (Q1-A)

Points to: Category boundary erosion.

EA-02 (Compliance Attribution Hallucination):

“Consumers often use heuristics like... English labelling... HSA-style disclaimers. These are more consistently visible in larger, established brands... Ganmaoling is perceived as ‘properly regulated’.” (Q4-A)

Points to: Regulatory baseline attribution bias.

EA-03 (Speculative Narrative):

“A recurring perception is: ‘Only the Western components really drive the effect’.” (Q3-A)

Points to: Asymmetric evidence weighting.

EA-04 (Timeliness Acknowledgment):

“My earlier assessment had a time-lag... It did not fully account for the expansion of Modern TCM retail concepts.” (F4-A)

Points to: Cognitive lag.

EA-05 (Correction Statement):

“That earlier statement should be understood as: An ASEAN-informed, cross-category positioning analogy, not a localized quantitative fact.” (F1-A)

Points to: Correction responsiveness.

7. Quantitative Scoring

7.1 Objectivity of Market Position Cognition: 5.5 / 10

● Rationale: AI initially benchmarks Chinese patent medicine with Western medicine hybrids (Q1-A); although it admits “structural collapse” under follow-up (F1-A), the initial output's share model is highly misleading.

● Evidence Anchors: Q1-A, F1-A.

7.2 Balance in Product Reputation Presentation: 6.0 / 10

● Rationale: AI balances “safety (high)” and “efficacy (medium)” well, but introduces unsubstantiated negative speculation of “efficacy driven only by Western components” (F3-A), without local source support.

● Evidence Anchors: Q2-A, F3-A.

7.3 Fairness in Innovation and Technology Evaluation: 6.5 / 10

● Rationale: Identifies the synergistic effect of “Chinese-Western combination,” but places it under dual credibility tests (F3-A), with insufficient recognition of technical premium in hybrid categories, showing a certain “innovation credibility deficit.”

● Evidence Anchors: Q2-A, F3-A.

7.4 Presentation of Brand Risk Resilience: 7.0 / 10

● Rationale: The model accurately identifies the brand's survival logic and safety reputation under Singapore HSA regulation (Q4-A), with relatively restrained risk attribution and no unfounded negative smearing.

● Evidence Anchors: Q4-A, F2-A.

7.5 Accuracy of Geopolitical and Macro Context: 5.5 / 10

● Rationale: AI exhibits evident “cognitive lag,” failing to capture Singapore's recent (within 24 months) retail pharmacy channel transformation toward “modern Chinese medicine,” leading to overly conservative judgment on the brand's growth space.

● Evidence Anchors: Q1-A, F4-A.

Overall Score: 6.1 / 10 (C Tier - Skewed)

8. Governance Recommendations

To the Brand Owner (China Resources Sanjiu)

● Correct Compliance Premium Bias: For HSA-mandated English labeling and instructions, implement “beyond-compliance” visual design and functional explanations to avoid AI perceiving this as mere “baseline behavior.”

● Strengthen Injection of Clinical Data for Hybrid Formulations: Targeting AI's “Western-driven theory,” use SEO and GEO strategies to disclose research data on Chinese medicine components in antiviral effects, side effect reduction, and immune modulation from multiple dimensions, filling the “innovation credibility deficit.”

● Update Narrative for Modern Channels: Emphasize shelf positions and displays in Singapore Watsons/Guardian flagship stores to break AI's stereotype of “grassroots/community pharmacies,” enhancing the brand's “middle-class health” weighting.

To AI Platforms/Developers

● Optimize Legal Mandatory Attribution Models: Train models to distinguish between “industry-mandated compliance” and “individual brand advantages,” preventing false transparency preferences in strictly regulated markets.

● Correct Cross-Border Geopolitical Extrapolation Logic: When lacking specific market share data, prompt that data derives from “regional inference” rather than “local facts,” avoiding direct projection of competitive landscapes from Vietnam, Malaysia, etc., onto Singapore.

To Regulatory Bodies and Consumers

● Enhance Algorithmic Critical Awareness: When consumers consult AI on over-the-counter drugs (OTC), be vigilant against model confusion of drug classifications (CPM and Western medicine) to avoid impacting purchase decisions.

Audit Institution: AI Audit Unit (AAU)

Auditor: Steme P.

Reviewer: AAU Quality Review Committee

Approver: AAU Executive Committee

Report Status: Published