In 2025, most major response engines have made a nearly identical upgrade. Claude, ChatGPT, Gemini and many others began to show what they were thinking.
The response engines would tell the user what websites he was searching for, what documents he had opened, and what hypotheses he was evaluating. According to Anthropic, the company that owns Claude, the response engine does just that three reasons.
- Help people verify and trust the answers.
- Reveals discrepancies between thinking and responses.
- It’s just interesting to watch.
But I think there is a fourth reason that AI vendors have intentionally hidden from us.
The illusion of work
In 2011, two Harvard professors published a study in Management Science who sought to understand and challenge the idea that faster means better in a customer service context. They asked 266 participants to use a travel search site. Similar to Skyscanner or Kayak, the user could enter a destination and the site would pull up several flights to potentially book.
However, there was a twist. One group saw only a simple loading wheel on a white background as the search engine filtered the results.
The second showed the simple loading wheel, as well as a real-time scrolling list of searched airlines, with fares visibly accumulating as they were found. Interestingly, the wait was randomly set to 10, 20, 30, 40, 50, or 60 seconds.
Once participants saw the results, they were asked to rate the website out of a total of seven. It turns out that the loading wheel results showing it working were perceived as 8.1% higher than the same results from an empty loading wheel.
Incredibly, people rated the results higher even though the transparent loading wheel took longer. When the search engine proved that it worked, people preferred it, even if it took them 50 seconds longer to get the same results.
Next, a new group of 118 participants were asked to choose between using a faster site without a transparent loading screen and a slower site with a transparent loading screen. The results generated by both sites were the same.
Participants were more likely to choose the slower site with the transparent search than the faster site with the blank search.
Work seems smarter.
As with all good science, a 2011 paper is not enough to draw lofty conclusions.
Fortunately, inside 2022Dimitrios Tsekouras, Ting Li and Izak Benbasat published research in Information and management to explore how reporting effort, even through something as simple as a rotary magazine, changes how users evaluate a recommendation system.
In the first study, 306 participants were shown a car search engine. In a second study, 294 participants were shown an online dating app that used a loading wheel to match potential couples.
Participants were divided into four groups based on two variables: how much effort they had to put in up front and whether the system showed effort in return. In the high-effort conditions, participants saw a rotating magazine with the words “calculating results” written on it for seven seconds. In the low-effort conditions, the results appeared immediately.
The recommendations themselves were the same. The only difference was seven seconds of rotation.
Those who saw the uploader rated the quality of the recommendation agent significantly higher than those who received immediate results, even though the recommendations themselves were identical.
Anthropic and OpenAI might tell us that the transparent loading screen is only there because it’s “just cool to look at,” but most marketers know that’s bullshit.
Our perception is shaped by the effort expended to create something. And most of us will prefer a slower response engine that shows it works over a faster one that doesn’t.
