What it means
As AI became a funding and procurement advantage, describing a product as AI-powered became close to mandatory. AI washing is what happens when the description outruns the technology: conventional rules-based automation relabeled, a thin model call bolted onto an unchanged product, or genuine machine learning doing something peripheral while the marketing implies it drives the core value.
Two questions separate substance from packaging. First, is the learned component load-bearing — remove the AI and does the product still work? Second, does the claim survive where the vendor is legally accountable? Marketing copy is cheap; securities filings and product documentation are not, and the language often changes noticeably between them.
This has moved from a credibility problem to a legal one, with securities regulators bringing enforcement actions against companies that overstated AI capabilities to investors.
Why it matters
Buyers pay a premium for AI capability and make architectural commitments on the strength of it. Washing distorts that market, and it also poisons the well for vendors doing real work — when every product claims the same thing, the claim stops carrying information.
In practice
Ask what specifically the model does, what happens if it is switched off, and what the product documentation says as opposed to the homepage. A vendor doing genuine work can answer precisely and usually enjoys the question.