Manipulation and Detection of Online Product Reviews

Online product reviews have become pivotal in shaping consumer decisions, yet they are vulnerable to manipulation by various actors seeking to skew perceptions of product quality and influence purchasing behaviour. Manipulation strategies range from paid or incentivised postings to coordinated campaigns exploiting platform anonymity, often emerging at the early stages of a product’s diffusion to inflate ratings and create artificial popularity. Such practices compromise the integrity of electronic word of mouth (eWOM) and pose risks to brand reputation, competitive fairness and consumer trust. In response, research has advanced a variety of detection methodologies drawing on behavioural and linguistic cues, network analysis and machine learning. These approaches include the examination of verbal and non-verbal features, metadata patterns such as IP addresses, generative language models used to simulate fraudulent reviews and the development of explainable indices that combine structured and unstructured data. Together, these efforts aim to safeguard marketplace transparency, support regulatory measures and equip platforms and consumers with tools to discern authenticity in review ecosystems.

Research from Nature Portfolio

Recent studies have delved into the linguistic characteristics of deceptive reviews through the lens of cognitive and perceptual cues. By applying reality monitoring theory to large-scale online review corpora, researchers have identified distinct patterns: emotion-driven reviews exhibit heightened affective language, while perfunctory postings lack detailed narrative markers such as adjectives and prepositions. Deceptive contributions often omit perceptual cues regardless of motive and display variable cognitive indicators when produced by paid posters. These insights present a nuanced framework for integrating linguistic features into automated detection systems, enhancing the precision of review authenticity assessments and informing sustainable e-commerce practices.

Manipulation and Detection of Online Product Reviews publication trend

The graph below shows the total number of articles in manipulation and detection of online product reviews across all publications each year (not limited to Nature Index journals).

Technical terms

Topic modelling: A statistical method for uncovering abstract themes in large text collections by grouping words into topics based on their co-occurrence patterns.

Generative language model: A neural network trained to generate human-like text by predicting the probability of word sequences.

Reality monitoring theory: A cognitive framework distinguishing memories of real events from those that are imagined, applied to detect fabricated content through linguistic cues.

Explainable AI (XAI): Techniques that provide interpretable justifications for machine-learning predictions, enhancing transparency and trust in algorithmic decision-making.

References

1. Investigating reviewers' intentions to post fake vs. authentic reviews based on behavioral linguistic features. Technological Forecasting and Social Change (2024).
2. Chasing spammers: Using the Internet protocol address for detection. Psychology and Marketing (2024).
3. Towards the development of an explainable e-commerce fake review index: An attribute analytics approach. European Journal of Operational Research (2024).
4. What makes deceptive online reviews? A linguistic analysis perspective. Humanities and Social Sciences Communications (2023).