Image by Getty / Futurism
A striking graph, which went viral on Reddit last week, shows the purported percentage of internet users visiting OpenAI's blockbuster AI chatbot ChatGPT overtaking Wikipedia over the past two years.
The data, compiled by UK-based market research company GWI, shows a steady decline in the proportion of users visiting Wikipedia worldwide, excluding China. In less than a year following its launch in late 2022, ChatGPT appears to have surpassed the online encyclopedia, in a striking reversal of fortunes.
If the data — which is based on survey responses and not site visits — is to be believed, it's a good reason to be concerned about the reliability of information people are seeking out online. While Wikipedia has never been known to be an infallible source free of bias or inaccuracies, generative AI has proven to be far more unreliable, thanks to widespread hallucinations and biases present in its training data. And while Wikipedia is built on prominently displayed citations, AI systems like ChatGPT often struggle to explain where their info is coming from, even cooking up fake references wholesale.
There are also serious questions of ethics and fair use. Wikipedia has an army of nearly 260,000 human editors who ensure accuracy across millions of articles worldwide. In contrast, it remains infamously unknown what exactly the large language models supporting ChatGPT were trained on — but it more than likely contains data pulled straight from Wikipedia, which OpenAI is now profiting off.
ChatGPT has seen a "36 percent rise in users from Q4 2023 to Q4 2024, as other online platforms remain the same or see a slight percentage increase or decrease," GWI senior data journalist Chris Beer told Futurism in an email. "It’s reaching more of the internet, more quickly, than almost any other platform in history."
Beer also pointed out the "massive adoption among university students," with 49 percent using ChatGPT, "compared to 53 percent who use Amazon!"
ChatGPT, which has been named as the fastest-growing app in the history of computing, has quickly burgeoned into a global phenomenon.
"I will say that based on our research, ChatGPT is actually most popular in the Global South," Beer said. "The markets where we see the highest adoption of it include Kenya, India, the UAE, and Brazil."
China was excluded since "Chinese authorities ban" many online platforms, making the data "unrepresentative," Beer said.
However, in the absence of a more detailed breakdown of actual usage numbers, it's hard to draw any definitive conclusions.
In a statement to Futurism, a spokesperson for the Wikimedia Foundation, the non-profit that operates Wikipedia, said the organization hadn't noticed any "significant drops in traffic on Wikimedia websites since early 2021" in terms of "pageviews and readership traffic."
"Our analytics show that Wikipedia’s pageviews are approximately 15 billion (B) per month (these have consistently been in the 15B to 18B range since Oct 2020 including occasional rises and declines in traffic)," the spokesperson said.
In other words, Wikipedia's traffic is holding steady, the organization says — but data from GWI and others suggests that ChatGPT usage has rapidly outpaced it.
Web traffic estimator Similarweb appears to corroborate GWI's data. Wikipedia is currently ranked number eight globally in terms of traffic compared to all other sites, while ChatGPT is currently ranked number six with just over 4.5 billion monthly visits.
And if Wikipedia's traffic is experiencing any type of decline, that could also have to do with the rise of AI.
"I think the long-term downward trend for Wikipedia has largely come from search engines adding more 'zero-click' answers," Beer told Futurism. "Whereas before they might return a Wikipedia page as the top result, now you’re more likely to see a weather forecast, or a famous person’s height, or whatever it may be, within Google itself."
"Google’s AI features, along with ChatGPT, are likely compounding that pre-existing trend," he added.
The discussion highlights just how ubiquitous ChatGPT and related tech have become in just a few years. It's helping students write essays, teachers grade papers, summarize data for office workers, and sending clueless hikers astray.
But at what cost are we ditching carefully reviewed Wikipedia articles in favor of often misleading or sycophantic chatbots? Considering the latest data, the problem of hallucinations is only getting worse as AI chatbots become more advanced. Companies' efforts to nudge their chatbots' outputs in the right direction have also resulted in mayhem and confusion.
The subject of carefully reviewed information and its reliability is more pertinent than ever in a world filled with disinformation and AI slop. Just last month, the Trump administration threatened Wikipedia's tax-exempt status, accusing the Foundation of violating the law via baseless allegations about the encyclopedia spreading "propaganda."
To critics, it was pure grandstanding and an assault on the freedom of expression.
"Wikipedia is one of the last places online that shows the promise of the internet, housing more than 65 million articles written to inform, not persuade," the Wikimedia Foundation said in a statement at the time. "Our vision is a world in which every single human can freely share in the sum of all knowledge."
