Glassdoor: Employee AI Sentiment Fell From 81% to 43%

81 percent in 2019. 43 percent in mid-2026. That is the trajectory of positive sentiment in employee reviews that mention AI, according to an analysis Glassdoor ran on its own platform, and it is one of the cleaner long-run measurements we have of how working people feel about this technology.

What was measured

Glassdoor's team scanned employee reviews posted between January 2019 and May 2026 for terms like AI, LLM, GPT and OpenAI, then classified the sentiment of those reviews. The positive share fell from 81 to 43 percent over seven years; the negative share climbed to 53 percent, meaning reviews that mention AI are now more likely to be sour than not. The window matters: it starts well before ChatGPT, when AI was an abstract headline for most workers, and ends with AI embedded in daily workflows.

Who soured first

The breakdowns are sharper than the average. Positive sentiment among Gen Z women sits at 21 percent, the most skeptical group in the data. Managers are the most optimistic, which fits a familiar pattern: the person who picks the tool and the person who uses it all day rarely feel the same way about it. By occupation, insurance claims adjusters, writers and customer service representatives lean heavily negative, the roles whose daily output overlaps most with what AI now does. The software split is the striking one: architects mention AI most often and most positively, while 57 percent of software engineers' mentions are negative.

How much weight the numbers carry

Glassdoor itself flags the caveat: its reviewers do not represent the US workforce, and unhappy employees are more likely to post. So treat 43 percent as a reading from a biased thermometer. The trend is the sturdier fact: the same bias existed in 2019, when the reading was 81. The instrument did not change; the direction did.

The employer takeaway

Set this next to the employment data and a gap opens up. Just this week we covered a survey putting actual AI-related job loss among US workers at 3 percent. Mass displacement is not showing up in the numbers yet; collapsing trust is. That order matters, because trust erodes long before jobs do, and a rollout that ignores it meets quiet resistance no dashboard will show you. Involve the people who will use the tool, say plainly what it does and does not monitor, and let the team keep a visible share of the productivity it generates. Then measure: run an anonymous pulse survey in six months and see whether your own 43 percent is moving.

Sources: The Decoder