One of the biggest complaints that ChatGPT users have is the lack of up-to-date information from the web in its models and chatbot. OpenAI’s announcement on Monday night of GPTBot, a web crawler that scrapes sites for data that may be used to improve its future models, could be the company’s answer.

It might not be a coincidence that the announcement follows recent news that the number of people using ChatGPT on the web has declined in the past two months. Data from Similarweb released late last week, which went unnoticed by most major publications, showed the number of monthly visits to ChatGPT had dropped 9.6% to 1.5 billion in July from 1.6 billion in June, which was in turn down from 1.8 billion in May. 

Meanwhile, ChatGPT’s Google competitor, Bard, had a strong month with monthly visits increasing 34.5% to 189 million in July following a weaker June. (Though to be fair, Bard’s usage trends may just be a few months behind ChatGPT’s.) OpenAI and Google did not respond to a request for comment.