Short answer? No. Slightly longer answer? You’ll struggle to do so today.
Anyone telling you they can is probably the same person who claims they can identify Bigfoot from blurry photos. It has been a few years since this experiment in large language modelling went public.
No one consented to this. We just sort of started engaging, and boom, we might now be in an AI bubble. We live in a time where a large chunk of the internet’s seemingly personal opinions about blenders, hotels, beauty creams, restaurants, and other businesses are entirely made up by software that has never eaten, slept, or blended anything.
AI reviews seem like an inevitable outcome, but what’s most disturbing about them is that they have gotten good enough that spotting the fakes is no longer something most people can do.
We can barely keep up with AI’s exponential improvement
When AI hit the market for the first time, we played with its functions, and the most discerning among us quickly realized it had tells. They built tools and an entire cottage industry based on sniffing out the clankers.
You know to watch out for generic phrasing, reviews posted in suspicious clusters, oddly specific superlatives, and so on. That advice has not lost all its value, but it’s marginal at this point. Now, the bots know how to vary sentence length, mimic different styles, intentionally misspell things, and complain about seemingly normal issues like shipping times.
The new reviews we get now sound like a mildly tired human. They include specific details and opinions about packaging or sustainability. Some have even specialized in complaining, because complaint language and generation turned out to be really easy and very effective at building trust.
Things tend to read as more real when you have a complaint thrown into what is an otherwise positive review.
Meanwhile, actual humans, bless them, have started writing more like bots without AI even getting involved. We interact with them so much that we have started matching their weird cadence, word choice, rhythm, and so on.
What you end up with is AI that has run out of data, training on what is mostly data it generated itself (snake eating its own tail vibes), and humans who have learned what search-engine-optimized language looks like (increasingly LLM-style), imitating AIs.
We are converging on the same-ish bland middle. Picture it as a hall of mirrors with a five-star rating attached.
Why this matters more than it sounds like it should
It can be easy to look at all this and just not care. However, you’ll care when you have to buy something and can’t tell what’s real and what’s not. The authenticity of reviews, for the most part, before the bots got here, was something we all depended on.
At first, the LLMs’ infiltration targeted seemingly random and low-stakes products like toasters, flashlights, and other knick-knacks on online marketplaces. We complained a little but then quickly lost interest because, who cares.
Then, they expanded into places that carry real weight. Medical practices, contractors, hotels, online gambling, and so many others are now under fire.
Casino and betting sites are a genuinely instructive case study here, because the stakes of getting a fake review wrong are higher than buying a mediocre blender. A review convincing someone that a shady operator is worth checking out leads down a path of lost funds and nowhere to complain.
That’s part of why the more credible operators in the space lean so heavily on structured, criteria-based verification instead of a sea of five-stars that could be from anyone. This trusted source is worth pointing out as an example of how to make this work.
Licensing checks, payout data, complaint resolution records, verifications – these may be duller reads than a subjective review, but they tell any serious player everything they need to know about a casino.
What even works anymore?
AI may have gotten really good at sounding like people, but there are still a few things that could give it away, even if the writing seems perfectly normal.
Look for friction instead of relentless positivity. Genuine reviewers tend to mention the boring, specific inconvenience, such as a rude customer service rep. The faker stuff may try to copy this style, but they still sound frictionless and performative in a way that feels like it was reverse-engineered from a marketing brief.
Check the paper trail. This is your classic due diligence advice, where you dig beyond the surface. A reviewer’s writing voice can be faked in a breath now, but their history of other reviews, purchase verification, and other supporting material are decidedly harder/more expensive to fabricate at scale, though not impossible.
Prefer sources that show their homework. When on the assessment end of things, you have to watch out for vibes and star counts. Have a list of documented, checkable things that are hard to fake convincingly, not just the writing style.
Look beyond the writing
We are past the point where you can read something, see a clear tell, and dismiss it as AI. Now, the prose is too good. The writing itself is no longer reliable evidence of a clanker at the helm. What you now need is a slower, less-satisfying skill: checking who is accountable for a claim and how well-phrased it is.
It is less fun than playing sentence-structure detective, but it also happens to be what still works. Stay sharp!
