ChatGPT loses hype, drops traffic for three months in a row

Rate this post

Sam Altman, CEO of OpenAI.
Kevin Dietz/Getty Images

  • The number of visitors to ChatGPT’s website has declined for three months in a row, per SimilarWeb.
  • Although it appears to be leveling off, and could be linked to students being off for the summer.
  • Yet academics have found that chatbots have become less accurate over time.

ChatGPT took the world by storm when it was released last November, but it looks like it’s losing momentum.

According to data from analytics firm SimilarWeb, August marked the third month in a row that the number of monthly visits to ChatGPT’s website worldwide declined.

And according to the same web data shared by Reuters, the average time spent on OpenAI’s chatbot has been falling from 8.7 minutes per month since March to 7 minutes in August.

Although Similarweb reported that the total number of visits fell by about 10% in June and July, it fell slightly in August with a 3% drop. In the US alone, ChatGPT website visits increased by 0.4% in August.

The number of unique visitors last month also rose slightly – about 0.3% – from 180 million to 180.5 million, per Reuters.

“One theory as to why ChatGPT’s web traffic dropped over the summer was the school holidays, which could help explain why the traffic trend stabilized in August as large numbers of US schoolchildren returned to class at the end of the month,” David F. Carr, senior insights manager at Similarweb, said in the report. wrote

Before Meta’s Threads usurped the title in July, ChatGPT was the fastest-growing app of all time, reaching 100 million users in two months.

Some of it was promoted by students, leading professors to find ways to combat ChatGPT plagiarism, and a Princeton student launched GPTZero to find out if an AI had written an essay.

But it’s also being used in the workplace, with employees using ChatGPT to write code, conduct research, and improve time management.

In July, users of OpenAI’s latest model, GPT-4, began to complain of reduced chatbot performance. A paper by Stanford and Berkeley academics found GPT-4 to be less accurate in several tasks – such as 2.4% accuracy in recognizing prime numbers compared to 97.6% accuracy three months earlier.

OpenAI did not immediately respond to Insider’s request for comment.

Leave a Comment