AI washing is the practice of overstating a product’s or a company’s artificial intelligence: presenting a roadmap item, a thin feature, or ordinary automation as advanced AI. The claim runs ahead of what the software actually does.
AI washing is what happens when the marketing describes more artificial intelligence than the product runs on. A roadmap item gets presented as a shipping feature. A rules-based workflow gets relabeled as an intelligent agent. A model that autocompletes text gets sold as autonomous decision-making. The label moves faster than the engineering.
The distance between claim and capability
The tell is the gap between what the demo implies and what the license delivers once it is configured. Some AI washing is deliberate positioning, and some is a product team believing its own roadmap. Either way the buyer inherits the difference. The pattern sits next to agent washing, the narrower version aimed at dressing up automation as agentic AI. AI washing is the wider category across any AI claim.
Why it lands on the buyer
A buyer cannot verify an AI claim from a sales deck, so the cost of finding out gets pushed down the table. It shows up as longer evaluations, pilots that exist mainly to test whether the claim holds, and purchases that underdeliver against the pitch. The defense is a specific question: what does the system do without a human in the loop, on which task, and how is the result measured? A vendor who can answer that is describing capability. A vendor who answers with vision is selling the wash.