It’s clear that AI is right here to remain, with many people (myself included) utilizing AI to assist us improve the pace and thoroughness of content material creation. Leaving evaluations on the services and products we use is one space the place AI could possibly be utilized for this function.
A part of my Product R&D crew right here at G2 leads evaluate moderation, verifying each single person evaluate to make sure belief and authenticity of all evaluations in our market. So whereas we in fact acknowledge that AI could make important and constructive contributions to the world of evaluations in several methods, inauthentic evaluations of any form should not accepted on G2.
Prioritizing genuine suggestions: evaluate moderation at G2
With over 3 million accepted evaluations from actual customers, G2 is the world’s largest and most trusted software program market. Getting thus far required onerous work and dedication throughout all of our groups, together with Product Administration, Consumer Lifecycle, Trusty & Security, Engineering, and Strategic Overview Consultants. But it surely wouldn’t be attainable with out respecting the ability of unbiased, genuine evaluations.
On the coronary heart of all of it, we all know that customers belief the experiences of others, which is why we prioritize the high quality of evaluations over amount. Patrons have come to belief G2 for researching and evaluating software program as a result of we do every little thing we are able to to ensure authenticity, similar to:
- Verifying the identification of each person on G2
- Leveraging over 43 knowledge factors to evaluate authenticity
- Vetting each evaluate with a constant and strict guideline
G2 takes an unbiased strategy to average evaluations. It’s our purpose to make sure each evaluate is the sincere opinion of an actual person and their very own expertise with a services or products. We additionally actively monitor for patterns that point out fraudulent conduct. When such situations happen, we carry out an intensive audit, take down the evaluations in query, and implement changes to our product or working procedures to stop recurrence.
G2’s stance on generative AI and software program evaluations
As we proceed to embrace AI to drive extra innovation, whereas making certain authenticity and high quality, G2 now permits AI-assisted and generated evaluations—offered they meet verification requirements for evaluations collected by way of Vendor Verified, LinkedIn Verified, or in-app submissions.
Finally, solely reviewers who meet one in every of these three verification requirements can use generative AI to generate evaluations. This marks an replace to how we take into consideration AI in evaluations, reflecting how reviewers naturally share suggestions at this time—usually utilizing AI to refine and specific their ideas.
Figuring out how a person is utilizing AI is extremely tough. Due to that, we’re opting to concentrate on validation of genuine customers’ and actual utilization and spend money on adopting AI instruments straight, so we all know how they use them. Take, for instance, G2’s current acquisition of unSurvey to deliver conversational AI to G2 evaluations with voice evaluate assortment and AI transcript-generated evaluations. In all of those situations, what customers can count on of G2 is that we’ll stand dedicated to making sure G2 stays a platform of actual content material.
Right here’s a bit extra context on how we take into consideration AI-assisted versus AI-generated evaluations:
AI-assisted: These are evaluations by which an actual person offers the firsthand draft of their evaluate (their expertise, insights, rankings, and so forth) after which makes use of AI instruments solely to shine or make clear their very own phrases. AI-assisted evaluations are all the time allowed on G2 when they’re genuine and observe our pointers. Typical AI-assisted duties would possibly embody:
- Grammar, spelling, or punctuation fixes
- Rephrasing sentences for readability or circulation
- Translating from one other language
- Suggesting synonyms or tightening up the content material construction
Key traits embody:
- Human-first: The core ideas, opinions, and examples come straight from the reviewer.
- AI as helper: The AI by no means invents new experiences or particulars; it merely refines what the person has already written.
AI-generated: These are evaluations the place AI is used to provide the majority of the written content material on behalf of the reviewer, successfully drafting full paragraphs or full narratives based mostly on prompts. These evaluations are solely accepted when they’re genuine from the person’s voice and meet one of many three requirements famous above for evaluate assortment by way of Vendor Verified, LinkedIn Verified, or in-app submissions. G2 is making certain that in these situations, we all know the person is genuine and trusting them to make use of AI to spotlight their distinctive expertise.
The important thing attribute of those evaluations is that they’re ‘AI-first,’ which means the AI writes most or all the evaluate textual content.
How G2 evaluate moderation detects AI-generated content material
Our crew has a rigorous vetting and auditing course of for all new evaluate submissions. Due to the large variety of evaluations we acquire on any given day, we should use a considerate mix of best-in-class expertise and well-trained crew members with a discerning eye.
G2 makes use of main third-party instruments and providers to detect AI-generated content material in evaluate submissions. Given our place as an unbiased market, we gained’t tip our hand on the particular applied sciences leveraged.
The instruments at G2’s disposal are enterprise-grade content material detection options that scan and flag AI-generated content material and conduct. After flagging this content material, our in-house crew of moderators evaluations these situations together with different elements to validate.
Whereas AI content material detectors and our machine studying mannequin assist flag faux evaluations at scale, we additionally layer within the specialised experience of our moderators to audit evaluations that require extra appreciable analysis. Our crew is expert at figuring out patterns generally related to fraudulent exercise and creating efficient fraud prevention practices.
Evolving our strategy with business developments and requirements in thoughts
To be probably the most trusted knowledge supply within the age of AI for informing software program shopping for selections and go-to-market methods, we’re dedicated to evolving our evaluate moderation strategy. This implies holding updated on market developments, speaking with communities that assist us keep knowledgeable on how different organizations deter faux evaluations, and extra.
One main manner the combat in opposition to faux evaluations and testimonials has progressed is with a current ruling by the Federal Commerce Fee (FTC). On August 14, 2024, the Federal Commerce Fee introduced a remaining rule to fight faux evaluations and testimonials by prohibiting their sale or buy and permitting the company to hunt civil penalties in opposition to realizing violators.
Inside this remaining rule is an announcement concerning faux or false buyer evaluations, buyer testimonials, and celeb testimonials. The rule states:
The ultimate rule addresses evaluations and testimonials that misrepresent that they’re by somebody who doesn’t exist, similar to AI-generated faux evaluations, or who didn’t have precise expertise with the enterprise or its services or products, or that misrepresent the expertise of the individual giving it. It prohibits companies from creating or promoting such evaluations or testimonials. It additionally prohibits them from shopping for such evaluations, procuring them from firm insiders, or disseminating such testimonials, when the enterprise knew or ought to have identified that the evaluations or testimonials have been faux or false.
Relating to this variation, FTC Chair Lina M. Khan shared, “Faux evaluations not solely waste individuals’s money and time, but in addition pollute {the marketplace} and divert enterprise away from sincere rivals. By strengthening the FTC’s toolkit to combat misleading promoting, the ultimate rule will defend People from getting cheated, put companies that unlawfully recreation the system on discover, and promote markets which are honest, sincere, and aggressive.”
Along with the replace concerning faux evaluations, the rule additionally introduced updates concerning shopping for constructive or damaging evaluations, company-controlled evaluate websites, insider evaluations and buyer testimonials, evaluate suppression, and the misuse of pretend social media indicators.
Critiques you’ll be able to proceed to belief within the age of AI
At G2, we proceed to embrace AI by discovering methods to work smarter and bringing new improvements to the marketplace for software program patrons and sellers. By means of this continued age of innovation and evolution, we stay staunchly dedicated to blocking faux or fraudulent evaluations.
AI is right here to assist real customers share their actual software program experiences extra simply, to not bypass our verification requirements. As a result of we imagine within the energy of genuine evaluations, you’ll be able to all the time depend on G2 to champion the voice of the shopper.
Take a look at G2’s neighborhood pointers and evaluate validity strategy to learn the way we forestall faux evaluations from our platform.
