Wikipedia says traffic is falling due to AI search summaries and social video
64 comments
·October 21, 2025irjustin
mmooss
Many said (some still say) that Wikipedia is not a replacement for traditional encyclopedias with articles written by domain experts, such as Britannica.
Many more scoffed at that, saying those people were just stuck in their old ways and unable to adjust to the obviously superior new thing.
Is that you? AI applications are different than Wikipedia and are better in some ways: Coverage is much greater - you can get a detailed article on almost any topic. And if you have questions after reading a Wikipedia article, Wikipedia can't help you; the AI software can answer them. Also, it's a bit easier to find the information you want.
Personally, I'm with the first group, at the top if this comment. And now truth, accuracy, and epistemology, and public interest in those things, take another major hit in the post-truth era.
scuff3d
Right, and where are all those LLMs without the billions upon billions of lines of text written by humans? A not insignificant number coming from Wikipedia?
Also, LLMs don't produce truth. They don't have a concept of it. Or lies for that matter. If you are using LLMs do study something you know nothing about the information provided by them is as good as useless if you don't verify it with external sources written by a person. Wikipedia isn't perfect, nothing is, but I trust their model a shitload more then an LLM.
consumer451
> post-truth era.
I know it’s completely normalized and the official name, but this has to be the most dangerous euphemism of our time.
It’s the era of lies.
zzo38computer
I agree, we will need Wikipedia and it is OK if the traffic falls, and that AI is not a replacement (and videos are not a replacement either).
Printed texts are still useful but so is Wikipedia (I continue to use both).
al_borland
> Become self sustaining through investments so it can live for 1000 years.
I always wondered why more companies or organizations didn’t do this. Pile up money during the good years to allow themselves to not need continued outside income to keep going, so they can do what is right instead of compromising their vision for the sake of hitting quarterly earnings. That isn’t to say they can’t keep making money, but do it for the right reasons that will keep the core business around for the long run.
nickff
If a company retains earnings, it has to pay taxes on them (as profits), but the money is still at risk (from a shareholder’s point of view) if something bad happens to the company (lawsuit or market problem). Shareholders usually want to receive whatever money the company has saved up, to safeguard it from being lost for no reason, and so that they (the shareholder) can put it to use elsewhere. This would change if the government stopped taxing retained earnings.
Okx
Unfortunately, Wikipedia doesn't seem interested in this. Their revenue is more than enough be able to invest and sustain the site forever, but they just increase expenses on non-core outgoings https://en.wikipedia.org/wiki/WP:CANCER
skeeter2020
It takes a pretty altruistic leadership team to plan for 1000 years; even the multi-billionaire tech bros only plan to live for ~200 or so.
al_borland
“A society grows great when old men plant trees whose shade they know they shall never sit in.”
I recently visited Scotland and on a visit to a distillery they mentioned they bought land in the US to grow trees that will make their barrels one day. The trees take over 100 years to grow (if I remember correctly). How is it we can invest ~200 years into a glass of scotch, yet we aren’t willing to take the same care and long term thinking in most other areas.
Even without being around for 1,000 years, I’d think doing this would de-stress and de-risk. Somewhere along way it became a bad thing to have a good, stable, long-lasting business. The only thing that seems to matter now is growth, even if they means instability, stress, excessive risk, and a short stay.
aeternum
I think we need Wikipedia competitors.
Wikipedia is a victim of it's own success, it was excellent at avoiding bias for quite awhile and the vast majority of articles are extremely well written.
However it's massive popularity and dominance have also led to, well this guy put it best: https://en.wikipedia.org/wiki/John_Dalberg-Acton,_1st_Baron_...
arunharidas
Elon Musk is launching Grokipedia this week as a competition to Wikipedia.
johnnyanmac
>I've always scoffed at the Wikimedia Foundation's warchest and continuously increasing annual spending.
what are they increasing spending on? Are they still trying to branch out to other initiatives?
I understand, even with static pages, that hosting one of the largest websites in the world won't be cheap, but it can't be rising that much, right?
intended
The production of factually accurate content is a pretty expensive job.
We who were born before this era really took off, are spoiled by the journalism standards and information purity levels of the past, especially post the fall of the USSR.
Wikipedia is impressive on what it manages to coordinate on a daily basis, especially given only 644 FT staff.
adventured
For 2023-2024, their budget was ~$177 million. Travel & events was 7.4% of their expenses. Processing fees on donations was 6.4%.
Grants & movement support was 25%.
Hosting was 3.4%. Facilities was 1.4%.
The Wikimedia Foundation is another Komen Foundation.
ryanwhitney
They reported a headcount of 644 for 2024–2025.
It's all very open if anyone wants to track down details themselves: https://meta.wikimedia.org/wiki/Category:Wikimedia_Foundatio...
2025–2026 is in-progress: https://meta.wikimedia.org/wiki/Wikimedia_Foundation_Annual_...
>Similar to last year, technology-related work represents nearly half of the Foundation's budget at 47% alongside priorities to protect volunteers and defend the projects of an additional 29% – a total of 76% of the Foundation's annual budget. Expenses for finance, risk management, fundraising, and operations account for the remaining 24%.
johnnyanmac
So, as a non profit the other 55% or so (approximately 95m) goes to salaries? That's interesting considering how much work is performed by volunteers.
mrcwinn
Why are they paying 6.4% on processing fees? What is "movement support" and where is the travel to? Do they have to publicly disclose these disbursements anywhere? This seems sketchy at best.
SanjayMehta
Maybe they can use the immense volunteer talent at their disposal to build their own AI/LLM.
I'm sure all those editors with decades of experience can do quickly outdo OpenAI and Grok and what have you.
wodenokoto
I think Wikimedia needs expensive outreach and modernisation to stay relevant. And most of those attempts will be in vain.
A poor comparison is how much money coca cola spends on advertisement, even though it is one of the best known brands in the entire world. And most of their advertisement is simply "This is our name, we exists", not even a value proposition or call to action.
If Wikimedia sets themselves up to pay for servers and maintenance for perpetuity, they will fall into obscurity.
With that being said, I also don't think they are spending their money in a good way.
Razengan
Wikipedia is the best use of the Internet, the best reason & outcome of its existence.
Right up there with anime torrenting sites.
But seriously, AI trained on Wikipedia should donate to Wikipedia. Why are the AI companies not doing this, or are they?
RickJWagner
Not for me.
Wikipedia had its day, in between print encyclopedias and quick query AI. Its place in history is now set.
Something else will come along soon enough.
kibwen
Until LLMs gain the ability to cite their sources, they will be, at best, a search engine on top of Wikipedia, and not a replacement for it.
somenameforme
Most/all LLMs have already been able to cite sources for quite some time now.
Angostura
…. And it will be worse
hackyhacky
Without Wikipedia, where will AIs get their (factual) training data? Reddit?
01HNNWZ0MV43FF
Disagree. Say I'm looking for a list of countries and their populations.
Wikipedia almost certainly has this in a nice table, which I can sort by any column, and all the countries are hyperlinked to their own articles, and it probably links to the concept of population estimation too.
There will be a primary source - But would a primary source also have articles on every country? That are ad-free, that follow a consistent format? That are editable? Then it's just Wikipedia again. If not, then you have to rely on the LLM to knit together these sources.
I don't see wikis dying yet.
At work, I had rigged one of my internal tools so that when you were looking at a system's health report, it also linked to an internal wiki page where we could track human-edited notes about that system over time. I don't think an AI can do this, because you can't fine-tune it, you can't be sure it's lossless round-tripping, and if it has to do a web search, then it has to search for the wiki you said is obsolete.
OpenStreetMap does the same thing. Their UIs automatically deep-link every key into their wiki. So if you click on a drinking fountain, it will say something like "amenity:drinking_water" and the UI doesn't know what that is, but it links you to the wiki page where someone's certainly put example pictures and explained the most useful ways to tag it.
There has to be a ground truth. Wikipedia and alike are a very strong middle point on the Pareto frontier between primary sources (or oral tradition, for OSM) and LLM summary
al_borland
LLMs are useless without source material.
AI companies should be donating large sums of money to Wikipedia and other such sites to keep them healthy. Without good sources, we’re going to have AI training off AI slop.
cindyllm
[dead]
crazygringo
LLM's have definitely replaced 90% of what I used to look up on a Wikipedia, simply because they integrate from so many more additional sources.
But at the same time I continue to contribute edits to Wikipedia. Because it's the source of so much data. To me, it doesn't matter if the information I contribute gets consumed on Wikipedia or consumed via LLM. Either way, it's helping people.
Wikipedia isn't going away, even if its website stops being the primary way most people get information from it.
crmd
As a 501(c)(3) nonprofit that is not dependent on web traffic for revenue, is a decline in traffic necessarily bad?
I always assumed the need for metastatic growth was limited to VC-backed and ad-revenue dependent companies.
intended
The warning sign is not traffic for ads, although this will result in a drop in donations eventually.
It means that now, people are paying for their AI subscriptions, while they don’t see Wikipedia at all.
The primary source is being intermediated - which is the opposite of what the net was supposed to achieve.
This is the piracy argument, except this time its not little old ladies doing it, but massive for profit firms.
qingcharles
They are highly dependent on web traffic for revenue.
And their costs are even increasing because while human viewers are decreasing they are getting hugged to death by AI scrapes.
khamidou
If you look up their latest annual report (https://wikimediafoundation.org/wp-content/uploads/2025/04/W...) you can see that they're allocating ~1.7% of their expenses towards hosting.
I doubt that they're getting "hugged to death" by AI scrapers.
cm2012
Isn't it true that only around 10% of Wikipedia massive budget is used to actually run the core website? The rest goes to bloated initiatives in the Wikimedia foundations orbit.
crmd
How is their revenue traffic-dependent?
AstroBen
Their traffic is potential donations
Something tells me a person is way less likely to donate if they're consuming the content through an LLM middleman
johnnyanmac
scraping Wikipedia feels like the stupidest possible move. You can in fact download the entire encyclopedia at any time and take all the time in the world parsing offline.
For such purposes, I'd naively just setup some weekly job to download Wikipedia and then run a "scrape" on that. Even weekly may be overkill; a monthly snapshot may do more than enough.
yorwba
You can download twice-monthly database dumps, but they consist of the raw wikitext, so you need to do a bunch of extra work to render templates and stuff. Meanwhile, if you write a generic scraper, it can connect to Wikipedia like it connects to any other website and get the correctly-rendered HTML. People who aren't interested in Wikipedia specifically but want to download pretty much the entire internet unsurprisingly choose the latter option.
sublinear
> As Miller puts it, “With fewer visits to Wikipedia, fewer volunteers may grow and enrich the content, and fewer individual donors may support this work.”
lwansbrough
Contributors are a tiny % of users. I'm sure they've got some room for improvement on incentivizing new contributors. But Wikipedia is a gift to humanity and I hope we find new ways for them to be paid for their contributions to AI.
undeveloper
> Contributors are a tiny % of users most of them were wikipedia users in some form before they were contributors I imagine
KPGv2
1/3 of all donations are from the banner. I just went and looked at their annual report, which disclosed this.
arjie
It's all right. Wikipedia was a magical device for its time, and it's still a great aggregator of information. It will probably last forever as such a link aggregator. Read-time curation is obviously far better than write-time curation, but the former used to be very hard. Now we have the former for cheap so it makes sense for Wikipedia to be Yet Another Source into the read-time curator. And the existence of a source database like Wikipedia makes many of these tools work a lot better.
People rightfully get upset about individual editors having specific agendas on Wikipedia and I get it. Often that is the case. But the chat interface for LLMs allows for a back and forth where you can force them to look past some text to get closer to a truth.
For my part, I think it's nice to be part of making that base substrate of human knowledge in an open way, and some kinds of fixes to Wikipedia articles are very easy. So what little I do, I'll keep doing. Makes me happy to help.
Some of the fruit is really low-hanging, take a look at this garbage someone added to an article:
https://en.wikipedia.org/w/index.php?title=Salvadoran_gang_c...
01HNNWZ0MV43FF
> Read-time curation is obviously far better than write-time curation, but the former used to be very hard. Now we have the former for cheap
It's _kinda_ cheap. Wikipedia is so cheap you can fit it all on a phone and search it instantly.
I agree overall but LLMs are just so heavy. I don't know if most people can afford to run one locally, and they're lossy. Both on a phone would be great. I fret a lot about data ownership, you know
codinhood
AI seems obvious, but social video? Are they saying people watch TikToks instead of reading Wikipedia, or people who used to look things up don’t bother anymore because of TikTok?
byzantinegene
tiktok seems to be the primary medium by which Gen Alpha obtain their news and knowledge
undeveloper
Gen Z as well. Many well into their 20s heavily consume tiktok.
bad_username
My personal traffic to Wikipedia fell after around 2019, when activist editors took over, and the site ceased to be trustworthy for a lot of important topics.
null
intended
I’m a pro-market solution person, but markets are a tool.
This is the kind of capitalistic behavior that is repugnant to our idea of how things should work.
This is not what the commons is for - taking the work of creators, repackaging it, and using platform capability to re-sell it.
At this point, I am coming around to the argument that governments should make their own local/national AI.
Venn1
This made me curious enough to check the stats for my little site. According to Cloudflare’s AI Overview, over the last 24 hours the breakdown is:
665 ChatGPT-User
396 Bingbot
296 Googlebot
037 PerplexityBot
Fascinating.
loloquwowndueo
Out of how many visits total?
About 80% of traffic to my sites (a few personal blogs and a community site) is from ai bots, search engine spiders or seo scrapers.
Razengan
How is it not conflict of interest when Google's AI summary (which is sometimes hilariously wrong) takes a click away from websites that pay for ads? Specially if it was trained on those websites
d--b
traffic falling means wikipedia will be cheaper to run. since they don’t rely on ads, it’d likely not affecting their revenues either (assuming those who don’t use it anymore weren’5 those givîng to it)
Gigachad
It’s probably not good long term for donations or attracting future editors. Possibly less people interested in editing when it becomes more unpaid work for AI companies than actually serving real viewers.
I'll go out on a limb and say we _need_ Wikipedia and it's okay that traffic falls.
Physical print encyclopedias got replaced by Wikipedia, but AI isn't a replacement (can't ever see how either). While AI is a method of easier access for the end user, the purpose of Wikipedia stands on its own.
I've always scoffed at the Wikimedia Foundation's warchest and continuously increasing annual spending. I say now is the time to save money. Become self sustaining through investments so it can live for 1000 years.
To me, it is an existence for the common good and should be governed as such.