Resources/Glossary/AI
Agent washing is calling an ordinary chatbot, script or fixed workflow an “AI agent” to make a product sound more advanced than it is. The quick test: ask the vendor what the tool actually does on its own, without a person clicking through each step.
In one line
Calling an ordinary chatbot, script or workflow an “AI agent” to sound more advanced. Ask a vendor what the tool does without a person clicking each step.
The term follows earlier labels such as greenwashing and AI washing. When a new idea becomes a buying trend, sellers rename existing features to match it. In recruiting software, that often means a chat window bolted onto an old product, a set of if this then that rules, or a template library being sold as an autonomous agent that will source, screen and engage candidates for you.
A real agent plans steps toward a goal, uses other tools such as a browser, email or your ATS, and adapts when something goes wrong. A washed product usually does one of three things: answers questions about your data, runs the same fixed sequence every time, or drafts text that a person must then copy, paste and send. Those features can be useful. They are just not agents, and they should not be priced as if they were.
Recruiting teams buy software on promised time savings. If a tool is sold as an agent that replaces hours of sourcing but still needs a recruiter to drive every click, the business case falls apart. Spotting agent washing protects your budget, keeps your expectations honest and helps you compare products on what they really do rather than on the label in the demo.
A staffing firm sees a demo of an agent that promises to fill pipelines automatically. The ops lead asks the rep to give it a new role live and step away from the keyboard. The tool returns a saved search and waits. In this example the lead marks it as a search assistant, negotiates the price down and keeps looking for true agent features elsewhere.
AI agents are a hot buying category, and the label attracts attention, budget and higher prices. Renaming an existing chatbot or workflow is cheaper than building real autonomy. Some vendors also use the word loosely without meaning to mislead, which is why asking for a live test matters more than the label.
It can be. In the US, the Federal Trade Commission has warned companies against making false or unsupported claims about what their AI products can do, and deceptive claims can break consumer protection law. For buyers, the practical step is to test the claim and get key promises in writing.
Ask the vendor to give the tool a goal and leave it alone, then ask what it did. Ask how it handles errors, which apps it can take action in, and what still needs a person. Real agents answer these clearly in a demo; washed products move the conversation back to slides.
Written by recruiters at Work Insiders. Part of the recruiting and HR tech glossary. Updated October 2026.