Trustpilot's use of AI
July 2025
At Trustpilot, our mission is to become the universal symbol of trust. This mission is critically important as AI changes the way people discover and evaluate businesses. We strongly believe that using AI responsibly is essential to achieving our mission and we’re evolving our platform to make sure both consumers and businesses benefit from AI advancements.
This page explains our approach to AI and identifies which Trustpilot product features are based on AI. You’ll learn how we’ve designed AI-powered tools to be safe, trustworthy, and easy to use.
What do we mean by AI?
We use the term “AI” to mean: artificial intelligence techniques and software that allow machines to apply learning and reasoning to make decisions that improve their ability to achieve specified objectives.
Trustpilot employs several types of AI technology to enhance our platform and services:
-
Machine Learning (ML): A subset of AI where systems are trained from sample data. We use ML extensively in our fraud detection systems and content moderation processes.
-
Natural Language Processing (NLP): Technology that helps machines process and understand human language. We use NLP to extract insights and sentiment from customer reviews and to power our semantic analysis capabilities.
-
Generative AI: AI systems that create new content based on patterns learned from training data. We use generative AI selectively for specific features such as AI-Assisted Review Responses and in our content moderation activities.
How are we using AI?
AI-Powered Fraud Detection
At the heart of our platform is a commitment to authenticity. Trustpilot uses a variety of machine learning techniques to detect and remove fraudulent content at scale. These systems analyze patterns across millions of reviews, looking for suspicious activity—such as unnatural language, review spikes, and other anomalies—that may signal fake or incentivized reviews.
Our models continuously learn and adapt, identifying new tactics used to manipulate our trust ratings. Today, more than 90% of fake reviews are removed automatically, without manual intervention. This allows us to act quickly while staying ahead of emerging fraud trends.
Generative AI for Moderation and Policy Enforcement
In 2024, we expanded our AI capabilities by deploying generative AI to help enforce platform policies —such as impersonation, abuse, or misinformation. This technology helps us identify violations of our guidelines, and flag potentially harmful or misleading content for further review.
By automating the detection of policy breaches we’re able to process and review more content, faster and more consistently.
Human Review for Accountability
AI is powerful, but at Trustpilot it doesn’t operate alone or unsupervised. Trustpilot complements automation with a team of trained human moderators who review flagged content, verify complex cases, and ensure decisions align with our standards and values.
This human-in-the-loop approach helps avoid overreliance on automation and ensures fairness in how reviews are evaluated. It’s especially important when context or nuance is critical to interpreting user intent.
Helping Consumers Make Choices with AI-Powered Features
At Trustpilot we use AI to help consumers make the right choice. These tools include:
-
AI Summaries on Company Profiles: This feature helps consumers quickly understand what people are saying about a business. AI Summaries analyzes hundreds or thousands of reviews and distills the most recent and relevant themes into a short, easy-to-read summary, giving users a fast, trustworthy snapshot of public sentiment.
-
Topics and top mentions: our AI models allow thematic grouping of reviews, instantly organizing vast amounts of feedback into meaningful themes, like "Customer Service" or "Payment". This helps consumers to find information that matters and pinpoint relevant feedback, ultimately leading to more confident decisions.
-
Search engine: The smart, semantic, language-aware search engine makes it possible for millions of users to instantly navigate our universe of over one million businesses and over 3,000 business categories.
Helping Businesses Grow with AI-Powered Features
At Trustpilot, we are also using AI to create tools that help businesses engage with reviewers faster and uncover important insights to improve their own service offerings. These tools include:
-
AI-Assisted Review Responses: AI doesn’t just protect the platform—it helps businesses engage with reviews more efficiently and at scale. Our smart response tools can automatically generate personalized replies to customer reviews - but only where a business wants to, and always giving the business editorial control. These tools reduce the manual work for customer service teams while maintaining a consistent, on-brand tone. Read more about AI assisted review responses in our help center.
-
Sentiment Analysis, Topics, and Review Spotlight: We’re also using AI to help businesses make sense of their feedback faster. AI highlights and identifies positive, negative, or neutral sentiments in review content and categorizes them into thematic topics. Our review spotlight tool summarizes large volumes of reviews into clear themes, helping teams spot trends, respond to issues proactively, and extract insights that can inform product, service, and operational improvements.
-
Review Highlights: AI also makes showcasing review content easier and smarter. Our new review highlights tool automatically displays review quotes grouped by AI-driven topics that can be curated automatically to show relevant reviews on landing or product pages. Read more about Review Highlights in our help center.
-
Market Insights: Market Insights utilizes advanced language models to analyze review texts and business profiles in order to identify and group similar businesses into relevant peer cohorts. For these defined cohorts, the feature provides objective, data-driven benchmarks, comparative statistics and illustrates performance trends over time. This feature offers businesses an objective basis for comparison and a perspective on their specific market landscape. Read more about Market Insights in our help center.
How do we ensure the safe use of AI?
We have a comprehensive AI governance framework in place, supported by clear policies and procedures. This helps ensure everyone can use our platform and AI-powered features with confidence, knowing they’re designed with safety and responsibility in mind.
Our framework is built on the following principles:
-
Content you can trust: we deploy industry-leading techniques to root out fake reviews, identify platform misuse, and ensure the trustworthiness of our content.
-
Humans-in-the-loop: our expert data scientists are involved in the deployment, ongoing maintenance, and review of our AI systems. They pay particular attention to any technology that could have a significant impact on consumers or businesses. To ensure fairness and transparency, we always offer the option to appeal content moderation decisions to a human.
-
Designing for data ethics: our internal Data Ethics & AI policy requires us to build ethical considerations into our systems from the start and is supported by testing, validation, monitoring and human oversight, helping us prevent harm or unfair treatment of any individual or group.
-
Privacy & Security: we have obtained a SOC 2 Type 2 Attestation Report and deploy strict security protocols and privacy controls throughout our organisation to ensure consumer and business data is handled in line with industry standards and our own Privacy Policy. For more information please see here.
-
Accountability and transparency: our ML and AI teams develop and deploy models with strong oversight and clear accountability to our leadership team. We ensure internal transparency by documenting how data is used, how systems are designed to operate, and the outputs or decisions they produce.
Frequently asked questions about Trustpilot's use of AI technology
No. Use of this service is optional and you will always have editorial control and the final choice before publishing anything using the AI-assisted review response service.
Yes. We maintain a Decision Dispute Process (DDP) available to all reviewers and businesses. This allows any decision about a review - whether assisted by AI or made manually - to be reviewed by a human agent in our Content Integrity team. Please see here for more information.
Advanced AI has been the foundation of our product suite for many years, delivering critical insights to consumers and businesses. In that time we have developed sophisticated monitoring and quality assurance processes including:
-
An AI Research lab where new approaches are peer-reviewed by the Trustpilot data science community.
-
A requirement for high performance and accuracy, with models going through a rigorous quality assurance (QA) framework that defines and tracks model performance against a set of criteria specifically chosen for the task being performed.
-
Explainability tools to support investigators with model spot checks.
-
Ongoing trend analysis of review and business data.
-
Best practice engineering processes applied across the machine learning and AI development lifecycle.
-
Partnerships with global leaders in artificial intelligence, leveraging their state-of-art AI models and secure, scalable infrastructure that adheres to best practice for safe and scalable AI.
-
SOC2 & GDPR compliance.
We implement a combination of testing, validation, monitoring, and human oversight throughout the lifecycle of our core AI systems to detect and protect against any unintended, systematically harmful, or otherwise unjustifiable bias. As a preventative measure, we ensure as a rule that our AI systems do not ingest any data points which could lead to bias resulting in a significant negative impact, such as gender, sexuality, ethnicity, physical, mental or financial health.
We use multiple systems to identify fake reviews. Our systems evaluate the reviewer, the business, and the review itself. Over time, our detection tools learn from the millions of authentic reviews on Trustpilot what real reviews look like and how to detect suspicious reviews - even those created by AI. By comparing reviews, reviewers, and businesses across our platform, we can identify and remove fake content - including fake content developed by AI.
No. Any personal data relating to a business’ own customers that is provided to us by the business as part of the review invitation services (such as names, emails and order reference numbers, used to issue review invitation emails) is never used in connection with our AI services. This data is processed strictly according to our Data Processing Agreement to provide review invitation services.
We never use any personal data relating to a business’ own customers that is provided to us by that business to train AI models. However, we may use personal data provided to us directly by reviewers, or the replies which business users of our platform post in response to reviews, to train certain models which underpin our services where permitted to do so pursuant to our Privacy Policy. Where this is the case, we rely on legitimate business interests as the lawful basis for this processing.
In line with our internal governance procedures and following Privacy by Design methodology, the technology teams at Trustpilot are guided by data minimisation principles to ensure that only data that is needed for the defined purpose and outcomes is used for AI processing.
As with all personal data processed by our organisation, we implement technical and organisational measures designed to protect the information in line with applicable laws, including encryption, access controls, and regular security assessments as appropriate.
All personal data used in our AI systems is hosted in EU-based data centres. Please refer to our Privacy Policy for more information on where we store personal data.
In line with our internal data privacy and security governance framework, access to data used by AI systems is done using the principle of least privilege, and leveraging various security controls as appropriate - such as single sign-on (SSO) and multi-factor authentication (MFA).
More information on our Security and Privacy practices can be found here and here.
This document reflects Trustpilot's AI practices as of June 2025. We review and update our AI practices regularly to ensure they remain aligned with our principles, legal requirements, and technological developments. Significant changes to our AI practices will be communicated through appropriate channels.