How do you detect comments' authenticity?

We take into account multiple factors while checking Comments' Authenticity. They are the following: an overall quality of an account who commented on a post as well as its comments' content authenticity.

A particular comment is marked as inauthentic if it corresponds to several negative factors at once.

Having on average 1-3 suspicious comments per post among 20 in total doesn't affect Comments' Authenticity level. If 50% of your comments per post are marked as inauthentic, it's time to ring the bell!

What type of comments are getting marked as suspicious?

Keep in mind that at a first glance, even suspicious comments are not sorted into an inauthentic group, unless an account who left those comments is getting caught by performing some suspicious activity. We do take into account multiple factors* to detect the suspicious type of comments:*

• comments that consist of emojis only or words like: wow, cool, fantastic etc.

Username1 Awesome 👌 Username2 Nice pic! 😍 Username3 Cool 👍👍👍

Most of them are monosyllabic and don't correspond to the post's content.

• comments that consist of another account's mention only, for example, comments to giveaways and contests.

How do we check whether the comment comes from a Comment Pod?

Engagement pods (comment pods, Instagram pods, boost groups) are created by bloggers who search for some extra exposure within social media space, mainly, to boost their profile activity. They create specific private group chats on Facebook, Telegram or by using other platforms.

A typical engagement pod participant makes a post and throws a link to a group chat with a comment such as: “likes, comments (3 words and more), saved.” Right after they scroll the chat up to see the last 10 post links from other bloggers to complete those. This type of suspicious activity is definitely hard to spot without the implementation of AI since even quality accounts are getting suspended by leaving tonnes of quite extended comments on a daily basis.

At SOMO we implemented a machine learning mechanism in order to detect Engagement Pods' type of comments. We are feeding our algorithm with thousands of comments examples found in different Engagement Pods. The comments are getting analyzed and sorted out into different groups according to the found and formed patterns. We check all the comments and accounts that correspond to these patterns, thus the error rate is extremely low.

The high percentage of suspicious comments among your posts is an indication of low Comments Authenticity. The low percentage of suspicious comments among your posts, on the opposite, raises up your Comments Authenticity level.