A few months ago, I sat in on a pitch for a startup that claimed its new app was "revolutionizing the industry with proprietary AI." The founder wore a black turtleneck, used the word "synergy" unironically, and projected a graph that looked like a rocket ship. The product? A to-do list.
Not even a smart to-do list. Just a regular to-do list with a chatbot that said "Good job!" when you checked something off.
That’s when it hit me. We aren’t in a bubble. We’re in a fog of nonsense.
According to a recent analysis by the SEC, there has been a sharp uptick in enforcement actions and inquiries regarding companies overstating their use of artificial intelligence. While that’s the legal side, the cultural side is worse. A study from London’s Alan Turing Institute found that nearly 40% of startups claiming to be "AI-first" don't actually use AI in any meaningful way—they’re just running basic automation scripts or, worse, a bunch of "if-then" statements.
This phenomenon has a name: AI washing. And it’s everywhere.
The Grand Masquerade of the Modern Tech World
Let’s be honest: "AI washing" is just greenwashing with better marketing. It’s the practice of exaggerating, fabricating, or stretching the truth about how much artificial intelligence is actually in a product or service. The goal is simple—charge a premium, attract investors, and look cutting-edge without doing the hard engineering work.
Here’s what most people miss: AI washing isn't just about lying. It’s about ambiguity. If a company says "we use AI," they might be referring to a static decision tree that was coded in 2019 and hasn't been updated since. They aren't necessarily lying about the code; they're lying about the sophistication.
I’ve found that the worst offenders aren't the massive tech giants—they’re usually the mid-tier B2B SaaS companies and the "disruptive" startups. They know that saying "we have a machine learning model" sounds infinitely more attractive than "we have a database with a really aggressive search function."
The "Magic Box" Fallacy
So, how do you spot a fake? It starts with the vocabulary.
Real AI—specifically machine learning—involves training models on data, iterating, and improving over time. It’s messy. It requires massive compute power and constant tuning. Fake AI is static. It doesn't learn. It just executes.
Here are the three biggest red flags I look for when I smell the scent of AI washing:
- The "Everything" Claim: If a company says their AI "solves all your problems," run. Real AI is narrow. It does one thing well, like detecting fraud or recommending a song. It doesn't "optimize your entire business ecosystem."
- The Black Box Response: Ask them how the model works. If they say, "It’s proprietary, we can’t disclose the architecture," but the product feels simple, they’re hiding a script, not a secret sauce.
- The "Human-in-the-Loop" Excuse: This is the sneakiest one. Some companies literally have people in the Philippines manually performing the tasks the "AI" claims to do. They call it "human-in-the-loop AI," but really, it's just a call center with a fancy logo.
The High Cost of Cheap "Intelligence"
Why do I care so much? Because AI washing is killing the credibility of actual AI.
We’re seeing a classic economic phenomenon: the Gresham’s Law of tech. Bad AI drives out good AI. When every CRM tool claims to have an "AI copilot," consumers become numb. They stop believing the hype, and when a genuinely revolutionary product comes along—one that could actually save them hours per week—they dismiss it as another gimmick.
I remember reading about an actual medical imaging startup that spent years developing a model to detect early signs of cancer. They were struggling to raise a Series B because investors had been burned by so many "AI-powered diagnostics" that were just Excel spreadsheets with a nice UI. That’s the tragedy. The fakes aren't just scamming investors; they’re starving the real innovators of capital and trust.
How to Expose the Fraud (A Practical Guide)
If you’re a consumer, an investor, or just a curious geek, you need to cut through the noise. You don't need a PhD in computer science to spot the fakes. You just need to ask a few uncomfortable questions.
First, check for the "data trail." Real AI needs data. If a startup claims to have a predictive algorithm, ask them where they got the training data. If they’re using public datasets from Kaggle that have been around for years, they aren't innovating—they’re just running a tutorial.
Second, look at the output variability. Use the product. Ask it the same question in different ways. A rule-based system will often give the same robotic, stilted answer. A good LLM (Large Language Model) will adapt its tone and syntax. If it feels like you're talking to a customer service script from 2005, that’s because you are.
Third, and this is my favorite, check the "failure mode." Every AI system fails eventually. The difference is how it fails. A genuine ML model usually fails gracefully—it gives a low-confidence score or says "I don't know." A fake AI fails spectacularly, either by freezing up or giving you a completely irrelevant answer that shows it doesn't actually understand the context.
The "AI Washing" Paradox
Here’s the irony that makes this entire situation so fascinating. We are currently living in the most AI-forward era in human history. We have generative tools that can write code, paint masterpieces, and compose music. Yet, the market is flooded with products that don't use any of that.
Why? Because building a wrapper is easier than building a brain.
It takes a week to build an app that calls an API from OpenAI. It takes years to build a custom model that runs on a private server. Most "AI startups" are just wrappers. They rent the intelligence from someone else and slap their branding on it. There’s nothing inherently wrong with that—it’s a business model—but it’s dishonest to claim you’re building the intelligence when you’re just the glue.
I’ve got a friend who works at a "AI-driven recruitment" firm. I asked her how the AI selects candidates. She laughed and said, "We just filter by the keyword 'MBA' and let the human recruiters handle the rest."
That’s it. That’s the "AI."
The Silent Shift We Need
We need to stop rewarding the story and start rewarding the substance.
In the coming years, I predict we’ll see a massive shakeout. The SEC is already cracking down on fraudulent claims, and the venture capital market is getting tighter. The companies that survive won't be the ones with the flashiest demos; they’ll be the ones with the most robust infrastructure.
So, the next time you see a headline screaming about "Revolutionary AI Disruption," take a breath. Look at the team. Look at their research papers. Look at their patent filings. If they can't show you the math, they're probably showing you the smoke.
Here’s the call to action: Don't be a sucker. Be the person who asks the hard question in the boardroom or the comment section. Call out the hype. Because the more we tolerate the "washing," the longer we have to wait for the real magic.
The future isn't built on buzzwords; it's built on backbone. Let's demand better.
