Whether you’re looking to crack down on fake reviews, safeguard your users from scams or combat counterfeit on your platform, you’re ultimately talking about weeding out fake accounts.

Fake accounts are at the heart of many forms of platform abuse affecting multiple industries and undermining the integrity of online interactions. Let’s take a deeper look at what we mean by this, starting with fake reviews.

Fake reviews

Content is not a reliable signal for authenticity. With the increasing popularity and advancement of AI tools, genuine-looking content can be easily automated, in most languages, at scale. This means that, to be effective in detecting fake reviews, you need to look beyond the content for more meaningful indicators of suspicion. Behavioural data signals are much more reliable.

In the case of reviews, these behavioural data signals are numerous, from IP addresses and email structure, to geographical distribution of reviews and unusual spikes in frequency. Analysing this data allows us to answer a single key question:

“Can we trust this person is real and that their actions are genuine?”

This applies when looking for all types of fake reviews - from professional review sellers to individual businesses looking to boost their profiles by getting friends and family to write positive feedback for them.

Whether driven by bots or by paid-for humans, you don’t have a fake reviews problem, you’ve got a fake accounts problem.

OK, that’s reviews covered, but what about everything else?

Scams

Well, scams begin with a lie - the promise of a friendship, relationship, investment opportunity, job, delivery, holiday accommodation… the list goes on. Scams are about duping and exploiting unsuspecting victims, usually for financial gain, and are rarely a singular incident.

In the first quarter of 2023, 40 million adults in the UK were targeted by scammers. In the US, $485.6 billion was lost to scams in 2023. Scams are an entire industry comprising bad actors with deadlines and targets, playbooks and handbooks, just like legitimate businesses. The only way for scammers to do this effectively is to use fake accounts and fake content - basically, they’re cheap and easy to scale.

And sadly, they have just been supercharged.

Generative AI has opened the door for scammers to add inexpensive, high-quality content to their activities. Poorly translated, badly written, low quality content is a thing of the past. And voice cloning and deepfake AI looks set to amplify scammers’ arsenals even further.

So, how can I trust what I am seeing?

A layered approach - understand who, what, where, and why

To really get under the hood of the challenge and, ultimately, know who to trust, you need to follow a methodical, layered approach:

• Who is this person and can we trust them? Not who they were before they came on to your platform but who they are NOW. Are they acting like a genuine user? Are their interactions following a natural pattern? Do their connections to other users seem authentic?

• What are they doing right NOW? Does their activity differ from previous behaviour? Is there evidence that this account has been taken over? Does their behavior connect them to other accounts that are acting in a similar pattern? Do their interactions seem suspicious?
  
  
• Where is this happening on your platform? In the reviews? The listings? The comments? The DMs? Where it happens gives us the context to classify the behavior - whether it’s a fake review designed to boost or attack a business or individual, or a fake listing designed to sell counterfeit goods or fraudulent services.
  
  
• All of this leads to the why. It’s not always clear at first. It can take time for the patterns to emerge and tell the stories. But, it’s crucial not to act too quickly - to see the patterns and  to understand why. The whack-a-mole approach of taking down anything you‘re uncertain about immediately may seem satisfying at first - especially the metrics! But ultimately, these are vanity metrics, an endless task and ineffective approach. The why is often hidden but it is the key to finding the networks of connected accounts and hardening your platform against the biggest threats.

How can I solve this?

Well, luckily, what this all boils down to is you don’t have a fake content problem… or a scam problem…or a spam problem…or a bypass problem…or a fake review problem…

You have a fake accounts problem.

And there is a solution for that.

Detecting fake accounts and non-genuine behaviours

Digging into the ‘Who’

If you have fraudsters on your platform, we’re already seeing them elsewhere too.

Pasabi identifies fraudsters in your data against our repository of known bad actors, empowering you to act swiftly with confidence.

Analysing the ‘What’ & ‘Where’

Pasabi analyses behavioural signals to find fake accounts as they emerge on your platform. We use natural language understanding to classify the behavior.

As we see AI increasingly driving fake accounts and fake content, a behavioural approach is the only way to fight this problem effectively.

Identifying the ‘Why’

Scammers can fake their identities, but not their behaviours.

Using cluster technology, we connect networks of bad actors by their behaviours to understand the ‘why’ and identify the biggest threats.

Our high accuracy levels enable you to automate with confidence at scale.

Themis' Fraud Monitoring Module supports you in the fight against fraud by freeing up your teams to focus their efforts where they’re most needed to have the biggest impact.

Winning the war against the scammers

Fake accounts are cheap and fake content is becoming cheaper. If you want to win the war, you must tackle the who, the what and the where before you can understand the why.

Then, you can take the strategic action you need to strengthen your defences to deter the most ardent scammers, making your platform a safe and authentic place for your customers and users alike.