"It's just a GPT wrapper" is the standard dismissal in AI circles. It means: this product is a pretty interface over ChatGPT or another model accessed through an API, adding little of its own. The implication is that anyone could rebuild it in a weekend, and the model provider could make it obsolete with one feature release.
The sneer is sometimes deserved and often lazy. Technically, almost every AI product is a wrapper, since almost nobody trains their own frontier model. The real question is how much value the layer adds: the workflow, the context about your situation, the integrations, the design, the harness around the model. Some of the most successful AI products are, strictly speaking, wrappers with exceptional layers.
Why you keep hearing it
Because it is the go-to insult in startup and investor conversations, and because the fear behind it is real: products that add too little do get steamrolled whenever the model companies ship a new feature.
What it means for you
When evaluating an AI tool, ask the wrapper question constructively: what does this add beyond the chat window I already pay for? If the honest answer is a nicer font, save the subscription.