Maintaining G2 Review Quality and Authenticity Through Hybrid Moderation
Client
G2 โ Review and Software Discovery Platform
Service Manual Review Moderation
Industry Software Reviews and Technology
Overview
G2 is a review platform where users share their experiences and opinions about software products and services.
With a large volume of reviews submitted by users, maintaining the quality, relevance and authenticity of reviews is essential to building trust among buyers, businesses and the wider review community.
To support this requirement, Foiwe provided manual review moderation services, where trained moderators carefully reviewed submitted content and evaluated it against defined review-quality and authenticity guidelines.
The moderation process involved reviewing user profiles, review content, product relevance, company relationships, IP-related signals, submission patterns, wording similarities, competitor references and potential AI-generated content.
The Challenge
For a review platform, every published review can influence how users perceive a product or company.
The challenge was not simply identifying inappropriate language. Moderators needed to determine whether reviews appeared genuine, relevant and compliant with established review guidelines.
Some suspicious activity could only be identified by looking beyond an individual review and examining the surrounding context.
Key challenges included:
- Reviewing individual user profiles
- Checking whether reviews were relevant to the product
- Identifying multiple reviews associated with the same IP address
- Detecting bulk reviews from people associated with the same company
- Identifying suspicious review patterns
- Checking whether the same person or company was submitting multiple reviews
- Identifying competitor names or references
- Detecting multiple reviews submitted within a very short period
- Identifying reviews with highly similar wording
- Reviewing content for potential AI-generated text
Our Manual Review Moderation Approach
Foiwe’s trained moderation team followed a structured review process to evaluate submitted reviews.
Rather than relying on a single indicator, moderators considered multiple signals and the available context before making a moderation decision.
1. Reviewer Profile Check
Moderators first reviewed the available information associated with the reviewer.
The profile check helped moderators identify unusual or suspicious activity that could provide additional context about the review.
2. Review Content Check
Each review was manually examined for quality, relevance, and compliance with the applicable review guidelines.
Moderators looked for:
- Irrelevant information
- Promotional content
- Suspicious claims
- Policy violations
- Repetitive content
- Potentially manipulated content
3. Product Relevance Check
A genuine review should be relevant to the product being reviewed.
Moderators checked whether the reviewer was actually discussing the product, its features, usage, experience, or other relevant aspects.
Reviews that did not appear relevant could be flagged according to the applicable guidelines.
4. Same IP Address Review Check
Moderators also considered IP-related information where it was available within the moderation workflow.
Multiple reviews associated with the same IP address could be treated as a potential signal of coordinated activity.
However, an IP match was not considered sufficient on its own to determine that a review was fraudulent. Moderators considered the surrounding information and other signals before making a decision.
5. Bulk Reviews From the Same Company
Moderators checked for unusual volumes of reviews associated with the same company.
A large number of reviews connected to employees or individuals from one organization could indicate coordinated review activity.
These cases received additional attention during the moderation process.
6. Pattern-Based Review Analysis
Moderators looked beyond individual reviews to identify broader patterns.
The review process considered factors such as:
- Reviewer activity
- Company relationships
- Review frequency
- Submission timing
- Review wording
- Product relationships
- Repeated behaviors
This helped moderators identify activity that might appear normal when viewed individually but suspicious when considered as a pattern.
7. Same Company or Person Review Check
Moderators checked whether multiple reviews were connected to the same person or company.
Where platform guidelines restricted multiple reviews from the same individual or company, moderators identified potentially conflicting submissions and handled them according to the applicable rules.
8. Competitor Name Check
Moderators checked review content for competitor names and references.
Where competitor mentions were not permitted under the applicable review guidelines, the review could be flagged for further action.
This helped keep reviews focused on the product being evaluated.
9. Multiple Reviews Within a Short Time
Review submission timing was another important moderation signal.
Moderators checked whether multiple reviews had been submitted within an unusually short period.
A sudden increase in reviews could indicate coordinated activity and therefore required closer examination.
10. Similar Wording Detection
Moderators compared reviews to identify unusually similar wording.
They looked for:
- Repeated phrases
- Similar sentences
- Duplicate wording
- Similar review structures
- Closely related descriptions
Highly similar reviews could indicate duplicated or coordinated submissions and were reviewed accordingly.
11. Potential AI-Generated Content
As generative AI became more common, moderators also considered whether review content showed characteristics associated with AI-generated writing.
Potential AI-generated content was treated as a moderation signal rather than automatic proof of manipulation.
Moderators considered the overall context and applicable review guidelines before taking action.
Human Judgment at the Center
Manual moderation was particularly important for cases where individual signals could be misleading.
For example, multiple users may legitimately share an IP address because they work from the same office. Similarly, employees of the same company may have genuine experiences with a product.
Because of this, moderators did not rely on a single signal.
Instead, they evaluated the full context of the review and the available supporting information before making moderation decisions.
Moderation Workflow
The review moderation process followed a structured workflow:
Review Submitted
โ
Reviewer Profile Check
โ
Review Content Check
โ
Product Relevance Check
โ
Company & IP Signals Check
โ
Review Pattern & Timing Analysis
โ
Wording Similarity Check
โ
Competitor Reference Check
โ
Potential AI-Generated Content Review
โ
Moderator Decision
This approach helped create a consistent and systematic review moderation process.
Key Review Moderation Checks
| Moderation Check | What Moderators Reviewed |
| Reviewer Profile | Profile information and activity |
| Review Content | Quality, relevance, and guideline compliance |
| Product Relevance | Whether the review relates to the product |
| IP Address | Potential connections between reviews |
| Company Association | Multiple reviews linked to the same organization |
| Review Patterns | Unusual or coordinated activity |
| Person/Company Reviews | Multiple reviews from connected users |
| Competitor References | Mentions of competing products or companies |
| Submission Timing | Multiple reviews posted within a short period |
| Wording Similarity | Repeated or highly similar review content |
| AI-Generated Content | Potential signs of automated content generation |
Results
The manual moderation process provided a structured approach for reviewing submitted content and identifying reviews requiring additional attention.
The moderation team was able to assess reviews using multiple factors rather than relying solely on the content of an individual review.
This supported the identification of:
- Potentially suspicious reviews
- Irrelevant reviews
- Duplicate or similar reviews
- Coordinated review activity
- Unusual company-level review patterns
- Reviews associated with suspicious IP activity
- Rapid review submission patterns
- Competitor references
- Potentially AI-generated content
Conclusion
Maintaining trust on a review platform requires careful evaluation of both what a review says and the context surrounding its submission.
For G2, Foiwe’s manual review moderation process used trained human moderators to examine profiles, review content, product relevance, company relationships, IP-related signals, timing, wording, competitor references, and potential AI-generated content.
By combining these checks with human judgment, the moderation process supported a more thorough and consistent approach to maintaining review quality and authenticity.