1) Demo ≠ Product
Staged videos without a sandbox are meaningless. Ask for a trial with your data.
A flashy video demo is a great way to grab attention, but it's not a product. Many startups create highly polished, staged videos that showcase a product's full potential, but what you're seeing is often a crafted illusion. The real test is a hands-on experience. A legitimate company will be confident enough to let you test their product with your own data in a real-world environment. This isn't just about functionality; it's about seeing how the tool handles edge cases, deals with errors, and integrates into your existing workflow. A great follow-up question is, "Can you provide a sandboxed environment for me to try this with my own data, or at least a live, unscripted demo that I can interact with?"
2) “Our Own Model” (But It’s a Thin Wrapper)
It’s fine to use APIs; it’s not fine to misrepresent them. Ask about architecture and rate limits.
Building a large language model from scratch is a massive undertaking, requiring billions of dollars and years of research. So when a small startup claims to have "their own model," a healthy dose of suspicion is warranted. More often than not, they are using a thin wrapper—a simple interface built on top of powerful, established APIs from companies like OpenAI, Google, or Anthropic. While using these APIs is a smart and efficient way to build a product, misrepresenting it as proprietary technology is deceptive. Don't be afraid to ask direct questions about their technical architecture. Inquire about the underlying models they use and what they have added on top to create unique value. This also helps you understand potential rate limits and long-term costs that are tied to their API usage.
3) Lifetime Deals on Heavy Compute
If costs are recurring, revenue must be too. Lifetime offers often collapse or throttle later.
If an AI tool is doing something complex like generating images, running complex data analysis, or processing large amounts of text, it's using a lot of computing power. This heavy compute isn't free—it costs money on a per-use or subscription basis. When a company offers a "lifetime deal," it's a huge warning sign. It's a business model that is fundamentally unsustainable for products with recurring costs. These deals often lead to one of two outcomes: the company eventually collapses, leaving users with a broken tool, or they start to throttle and severely limit usage, making the "lifetime" access almost worthless. Sustainable businesses have recurring revenue to cover recurring costs.
4) No Customer Logos—Only Affiliates
Look for organic case studies and retained users, not just influencer threads.
Customer logos and organic case studies are the gold standard for social proof. They show that real people at real companies are finding value and are willing to stand behind the product. Be wary of startups whose marketing is dominated by affiliate links and paid influencer threads. While influencer marketing has its place, it's often more about generating quick sales than showcasing genuine, long-term user satisfaction. Do a little digging: search for independent reviews, look for the product on job boards (are companies hiring people to use this tool?), and see if you can find genuine testimonials from users who are not part of an affiliate program.
5) Science-y Jargon, No Measurable Outcomes
If they can’t define success metrics, you won’t either. Demand benchmarks or pilots.
A common tactic among unproven startups is to cloak a simple idea in complex, "science-y" jargon. They might talk about "proprietary neural nets," "advanced semantic clustering," or "generative pre-trained transformers" without ever explaining what the product actually does for you. Your job as a potential customer is to cut through the noise and demand measurable outcomes. If they can't clearly define how their tool will save you time, make you money, or solve a specific problem, they likely don't have a solution worth paying for. Ask for specific benchmarks or propose a small-scale pilot project with clear success metrics. If they can't agree to this, it's a sign they're not confident in their product's ability to deliver.