AI Slop Is Destroying Online Reviews, and We Need to Fix It

Last week, I was trying to order a portable power bank for a quick trip up to Jos. I jumped on one of our popular local e-commerce apps, found a device with a shiny 4.8-star rating, and started scrolling through customer comments before hitting checkout.

Something felt dead wrong.

Every single 5-star review looked like it was spat out by the exact same ChatGPT prompt: "Excellent build quality! Outstanding performance and seamless integration!"

Not a single customer mentioned whether the power bank could actually survive a sudden light switch-off in Gbagada or how many hours it took to charge a basic smartphone. Just generic, soulless praise.

The 5-Minute Python Script Poisoning Trust

As someone who writes code for a living, I know how frighteningly easy this is to execute. You don't need a massive operational budget to fake a brand reputation anymore.

A vendor in Onitsha or Lagos can write a quick Python script using an OpenAI API wrapper, feed it a prompt like "generate 200 enthusiastic reviews for a Bluetooth speaker," and hit run. Throw in a basic cron job, route the requests through a handful of cheap residential proxies, and suddenly a completely mediocre product looks like the best thing since sliced bread.

The FTC in the US recently finalized rules banning AI fake reviews, but let's be real—foreign regulations won't clean up our local marketplaces. We have to deal with the reality on the ground here.

E-commerce in Nigeria Can't Afford More Trust Issues

We already operate in an environment where trust is extremely hard to build. People are naturally cautious because almost everyone has a horror story about "What I ordered vs. What I got" from a random Instagram page.

That skepticism is why Pay-On-Delivery held on for so long, and why actual word-of-mouth recommendations are worth more than any fancy landing page copy.

When users realize that star ratings are stuffed with AI slop, they stop trusting digital feedback altogether. If a shopper in Akure or Owerri gets burned once by an item backed by 500 fake 5-star reviews, they won't just hold a grudge against that specific vendor—they might delete the entire shopping app and go back to physical markets for good.

How to Spot the Bots (and Fix the Code)

If you're trying to shop online without getting scammed, look for these telltale signs:

1. Zero specific context: If a review for a backup power system doesn't mention setup time, battery degradation under heavy load, or noise levels, be suspicious.
2. Repeated template language: Generative models love phrases like "Game changer," "Outstanding performance," and "Highly recommended." When five consecutive reviews share the exact same rhythmic structure, move on.
3. Unnatural review spikes: 80 overly detailed reviews dropped on a random Tuesday between 2 AM and 4 AM is a classic bot run.

For those of us actually building products and platforms, we can't just tell users to "be smart." We have to engineer friction against bot networks directly into our backends:

• Strict Verified Purchase Locks: Block the write-review endpoint completely unless the transaction status in your database reads DELIVERED.
• Heuristic Rate Limiting: Flag or auto-quarantine accounts that drop multiple reviews across different product categories within short timeframes.
• Vector Embeddings for Deduplication: Run simple semantic similarity checks on incoming reviews using vector embeddings. If a new review is 90% semantically identical to 30 existing posts across the site, flag it for manual review.

If we don't start baking these defensive checks directly into our user experience and platform pipelines, we're going to destroy the hard-earned trust that took years to build across the Nigerian tech space.