Sam Makwana.
Is the drop in paid subscriptions a sign that ChatGPT is set to follow the path of other internet leaders-turned-relics, or can OpenAI avoid the fate of those that came before? Impressive’s Sam Makwana asks the question.
Last month, ChatGPT hit 1 billion monthly active global users, making it the dominant AI tool in a hotly contested market. But being the pioneer is a dangerous business in Silicon Valley. History is littered with the corpses of first-generation internet giants – like MSN Messenger and Yahoo.
Once the dominant messaging service, MSN Messenger cornered the market in the late 1990s and early 2000s. But it was usurped by WhatsApp and Apple’s built-in messaging platform when mobile phone adoption took off. Similarly, Yahoo was once the front page of the internet, but failed to evolve with changing technology. To rub salt into the wound, Yahoo famously passed on the opportunity to buy Google not once but twice.
The lesson for ChatGPT is clear: market leaders can’t afford to rest on their laurels. While ChatGPT still commands massive raw numbers, its dominance is fracturing. Nipping at its heels is Google Gemini, with 900 million monthly users and an additional 2 billion users interacting with Gemini via AI overviews in Google search.
Then there’s Claude, paling in comparison with 26 million monthly active viewers but with a fast-growing enterprise user base.
So is history set to repeat in the case of ChatGPT?
The financials
OpenAI is facing a severe cash-flow problem, with the company’s own forecasts predicting a $14 billion loss in 2026. This hasn’t been helped by the reported 80 per cent drop in ChatGPT Plus subscriptions, down from 44 million in 2025 to just 9 million in 2026.
Part of the reason for this drop off is the quality of ChatGPT’s output, which has deteriorated over the last couple of years. We’re seeing this play out in our own business. While we once relied on ChatGPT to create scripts to automate our workflows, increasingly, we found more issues with the output, and the tool was unable to provide an accurate solution to our problems.
Switching behaviour
Like many enterprises, we switched to Claude and saw a marked difference in quality. Recommendations are much more balanced and based on evidence. There are also very few cases of hallucinations, where the tool makes up new but false information, which has been a major problem with ChatGPT. To get an idea of the scale of the issue, a Deakin University study found that more than half of ChatGPT citations in mental health literature are either fake or contain errors.
Still, as it stands, ChatGPT retains around 82 per cent of all AI platform referral traffic to external websites, an important statistic for ecommerce businesses to note. You might not be using ChatGPT, but a large number of your customers are.
The people’s platform
Perhaps we can consider ChatGPT as the people’s AI. Given its biggest strength is breadth – image generation, real-time browsing and low-latency voice conversations – it is the natural answer if someone needs one AI that does a bit of everything.
Where the general public can rely on this platform for day-to-day usage, for more in-depth and research-driven work or for workflow automation, systems like Claude or Perplexity are more appropriate, and we can expect to see a rise in usage of these tools.
ChatGPT and its parent company, OpenAI, certainly have some challenges in the future, but they are not out of the game. If OpenAI can get past its financial woes and overcome its trailblazer structural vulnerabilities, it could avoid the same first-mover curse that turned MSN from an internet staple into a nostalgic relic. Only time will tell if it ends up in the history books next to those once dominant players.
Sam Makwana is the General Manager at integrated marketing agency Impressive.