At 25, Wikipedia faces a double threat: the rise of AI and the decline of local media | CBC Radio

When Wikipedia first emerged in 2001, it was nonetheless a time when most needed to be affected person for info — ready for the high-pitched scree and its answering cry as the laptop linked, painstakingly, to the web through dial-up. 

And the concept of an open supply encyclopedia that may very well be up to date by anybody in actual time — or its equal in these pre-fibre-optic days — sparked questions and lots of criticism about how correct that info may very well be. 

Quick-forward 25 years and Wikipedia is now the ninth most visited website on the web, with almost 15 billion guests every month, looking and modifying its greater than 65 million articles.   

However regardless of its speedy ascent in the early years and regular progress thereafter, Wikipedia isn’t as seen because it was once. Now, if you Google a query, the prime search consequence will probably be a Wiki hyperlink, however its AI will even handily synthesize the reply for you above it. And ChatGPT? That cuts Wikipedia out altogether.

Now, human guests to the website are on the decline, dropping by roughly eight per cent in components of 2025, whereas massive language fashions (LLMs) — chatbots or different types of AI that may condense phrases and info — are hammering Wikipedia’s servers and utilizing it as a coaching floor. 

If these traits proceed, alongside the decline in local information shops which might be Wikipedia’s important sources, the future is “extra dire than you assume,” says Zachary McDowell, an affiliate professor of communication research at the College of Illinois in Chicago and the writer of Wikipedia and the Illustration of Actuality.

Take a look at it like a pyramid of info accessibility, with LLMs at the prime, Wikipedia in the center and conventional information media on the backside, he mentioned.

“As you erode all the secondary sources under and you then begin to erode Wikipedia, what you might have is one thing that can inevitably crash in upon itself,” he mentioned.

“It’s been proven over and over once more that if you feed these [AI] methods artificial knowledge, if you feed them issues which were created by different AI sources, they find yourself with what they check with as mannequin collapse.” 

In layman’s phrases, it’s thought of digital inbreeding — when AI-generated info will get fed again in on itself once more and once more, growing the quantity of errors and inaccuracies.  

Human failsafe

Wikipedia’s founder, Jimmy Wales, expresses extra concern about the monetary implications of the growing demand that LLMs are inserting on the on-line encyclopedia. He notes the want for extra databases and servers to assist that further visitors from “AI crawlers” was the cause behind the offers it introduced with a number of AI companions in January, together with with Amazon, Meta and Microsoft. 

“The typical donation to Wikipedia is about $10 [US],” he mentioned. “Individuals aren’t donating to subsidize OpenAI.”

However McDowell’s issues about these AI crawlers making it tougher to entry impartial, correct info? Wales mentioned he doesn’t share them in terms of Wikipedia. 

“We don’t take heed to AI; Wikipedia is written by people and one of our strongest coverage factors is that every part in Wikipedia must … have a high quality supply,” he mentioned. “That’s the pathway into Wikipedia … human-created, human-vetted data.” 

The primary draft of historical past

However McDowell and Wales agree that media focus — particularly in small local newspapers and information stations — impacts Wikipedia, however in a bigger sense, it additionally impacts the skill to precisely seize a file for historical past.   

Conglomeration erodes the “neutrality” for which Wikipedia and conventional media try, McDowell mentioned. 

“These conglomerates, many of which have very political leanings, at the moment are pushing a specific ideology and agenda.”

WATCH | How AI can be utilized to interrupt an election:

Can you notice the deepfake? How AI is threatening elections

AI-generated faux movies are getting used for scams and web gags, however what occurs once they’re created to intrude in elections? CBC’s Catharine Tunney breaks down how the know-how will be weaponized and appears at whether or not Canada is prepared for a deepfake election.

In Canada, greater than 250 local information publications or broadcasts have shuttered between 2008 and Oct. 1, 2025, in response to the advocacy group Information Media Canada. 

“You already know, it is simpler in some methods to jot down a historical past of a small city from 30 years in the past than it’s from three years in the past if, as in lots of locations, the local newspaper has died and gone,” Wales mentioned in an interview with CBC Radio.

“That first draft of historical past isn’t even being captured in the first place. So there isn’t any query that that is a drawback.” 

The ‘Wikipedia detour’

AI, nevertheless, is simply rushing up what McDowell calls “the Wikipedia detour” — one thing that started a decade in the past, as Google began summarizing solutions on the search outcomes web page itself.  

Chopping Wikipedia out of the equation doesn’t simply have an effect on its skill to recruit editors or donors, it undermines digital and info literacy, as a result of folks don’t see the citations that type the basis of these articles.  

Nor are they inspired to dig deeper, in the means that what would possibly begin as a search about black holes finally brings you to the dates for an upcoming lunar eclipse. Wikipedia will be a rabbit gap, however in a great way.

It’s how Jess Wade has helped increase the profile of feminine scientists. The British physicist and assistant professor at Imperial School London in the U.Okay. has written greater than 2,200 biographies on Wikipedia of ladies and different marginalized teams who work in the sciences over the previous eight years, saying that the majority of her articles get visited as individuals are investigating a scientific idea and then bump into the undeniable fact that it was invented by a girl. 

And that enhances their visibility in actual time. Throughout the COVID-19 pandemic, Wade and a colleague added biographies about the ladies and folks of color on the entrance traces of the public well being disaster. As time wore on, she mentioned, the “previous white males” who had been showing in newspapers or on TV started to get replaced by some of the specialists she had included.

“I used to be actually struck by what number of broadcasters or lecturers or attorneys use Wikipedia as a first level of name when in search of info.”

How AI may help

There are methods, although, that Wikipedia is exploring learn how to use AI to enhance, together with its search expertise, as its interface hasn’t modified a lot lately. That might embody utilizing a chatbot, Wales mentioned.

And whereas the website’s 250,000 volunteer editors would nonetheless be the ones curating it into the future, he mentioned he can see AI doing a little easy automation — fixing a useless hyperlink in an article, for instance, by discovering potential replacements that a human may validate and resolve whether or not to incorporate.

“Automating some of the drudgery of engaged on Wikipedia may very well be very useful and type of make it greater high quality.”