OpenAI, regarded as the frontrunner in the generative artificial intelligence (AI) market, is seeing its growth slow. After the release of "GPT-5," exaggerated performance promotion, a decline in users, and policy controversies followed, damaging trust in corporations. As the fever around the rapidly spreading ChatGPT cooled, growth decelerated, and in the United States, a key market, user time on site fell by more than 20%.
According to the app analytics platform Apptopia on Oct. 20, the rate of new downloads for the ChatGPT mobile app has shown a clear slowdown recently. The global monthly download growth rate in October was expected to fall 8.1% from the previous month. While simple cumulative installs are still in the millions, the pace of growth has clearly turned down.
In the United States, average time spent per daily active user (DAU) has dropped 22.5% since July, and the average number of sessions per user fell 20.7%. The rapid growth of rivals such as Google's "Gemini" and changes in GPT-5's model characteristics are seen as jointly driving the trend.
OpenAI rolled out GPT-5 to reverse the decline, but instead drew "public embarrassment" over allegations of performance exaggeration. Kevin Weil, OpenAI's vice president, said on social media (SNS) on Oct. 19 that "GPT-5 solved 10 'Erdős problems' that had gone unsolved for decades." The Erdős problems are open problems in combinatorics and number theory posed by Hungarian mathematician Paul Erdős, some of which have remained unsolved for decades.
But mathematician Thomas Bloom pushed back, saying "this is a dramatic misunderstanding." Bloom noted that items marked "open" on the website he maintains were simply ones for which he had not checked the papers, and explained that "GPT-5 did not discover new solutions but found existing papers."
The controversy spread quickly. Yann LeCun, Meta's chief AI scientist, criticized that they were "intoxicated by their own fantasy," and Demis Hassabis, CEO of Google DeepMind, publicly called it "embarrassing." In response, OpenAI researcher Sébastien Bubeck also acknowledged that they "only found solutions that exist in the literature," and GPT-5's claim of a "scientific discovery" effectively ended as a flap.
Earlier, OpenAI also faced noise during the GPT-5 announcement process. In a live broadcast in early August, a comparison chart intended to showcase GPT-5's performance was produced out of proportion to the actual figures, sparking controversy. Because some bar charts were displayed with exaggeration relative to the real numbers, criticism of "chart manipulation" poured in, and CEO Sam Altman stepped in to explain. He said it was "a simple mistake by an employee working late into the night," but criticism arose that it was intended to inflate results.
Since launch, user complaints about GPT-5 have also continued. As paying users were forcibly switched from the existing GPT-4o model to GPT-5, reactions followed that "answers are short and lack creativity" and that it "sounds robotic." Some noted that GPT-5 makes errors even in simple calculations or basic fact checks. As dissatisfaction spread, OpenAI reversed its policy to allow users to select the previous GPT-4o again.
Amid successive missteps and controversies, OpenAI pulled out a "forbidden choice." CEO Sam Altman announced that starting in December, it will allow the creation of "erotica (pornography)" content only for users who have completed age verification. He said, "Adult users should be treated as adults," adding it was "a decision that considers the balance between freedom of expression and mental health."
This move is seen as a kind of user lock-in strategy to revive the slowed growth. As safety filters and moderation rules led to user dissatisfaction, the view is that OpenAI is trying to pull that demand into its official service. In fact, circumvention prompts for moderation are already being shared in overseas communities, and some users have moved to alternative chatbots such as Character AI or Grok.
In the AI industry, reactions to OpenAI's moves are mixed. Some assess it as "an inevitable choice amid accumulated user fatigue," while others criticize that "it undermined its own ethical principles."
Choi Byung-ho, a professor at the Korea University Artificial Intelligence Research Institute, said, "The slowdown in ChatGPT's growth is largely due to a weakened user acquisition momentum," adding, "OpenAI's consideration of expanding adult services or low-priced service regions is also an attempt to break through this stagnation phase." He continued, "Google's 'Gemini' is rapidly expanding its reach across image, video, coding, and other end-to-end services, threatening OpenAI," and added, "The two companies' competition will extend to new fronts such as agents, enterprise, and robots."