One-Third of Web Pages are Now Written With AI, Says Pew Research (pewresearch.org) 38
A Pew Research senior data scientist analyzed 500,000 English-language webpages from the past five years with an AI detection tool to estimate how many were likely written or edited by AI:
Of all the pages in this sample, 10% show significant signs of AI authorship... If we filter old webpages out of our samples and look at only the pages published after the release of ChatGPT, the trend is even more pronounced. In the July 2026 snapshot, signs of AI authorship can be found in over one-third of pages published after ChatGPT was released. This is in line with other studies that have shown that large shares of recently published pages on the internet were likely written or substantially edited by AI.
AI-authored text is not evenly spread across the web. When ChatGPT was first released, the kinds of linguistic patterns that can signal AI authorship appeared at similar rates across the main top-level web domains (.com, .org, .edu and .gov). But in samples from 2026, around one-in-ten pages with a .com domain show signs of AI authorship — about double the share on .org domains (4.6%), and 10 times the rate on .edu or .gov domains (both around 1%).
They even identified specific ways online text has changed, "comparing the internet of today to a snapshot from 2023."
Of all the pages in this sample, 10% show significant signs of AI authorship... If we filter old webpages out of our samples and look at only the pages published after the release of ChatGPT, the trend is even more pronounced. In the July 2026 snapshot, signs of AI authorship can be found in over one-third of pages published after ChatGPT was released. This is in line with other studies that have shown that large shares of recently published pages on the internet were likely written or substantially edited by AI.
AI-authored text is not evenly spread across the web. When ChatGPT was first released, the kinds of linguistic patterns that can signal AI authorship appeared at similar rates across the main top-level web domains (.com, .org, .edu and .gov). But in samples from 2026, around one-in-ten pages with a .com domain show signs of AI authorship — about double the share on .org domains (4.6%), and 10 times the rate on .edu or .gov domains (both around 1%).
They even identified specific ways online text has changed, "comparing the internet of today to a snapshot from 2023."
- Em dashes appear twice as often
- Oxford commas increased 63%.
- Words used more frequently by AI (like "delve," "interplay" or "testament") have more than doubled in usage.
- "Negative parallelism" comparisons ("it's not just X, it's Y") have nearly tripled.
Re: no surprise (Score:4, Insightful)
And the last 15 years the corporate web pages have been dumbed down to irrelevancy so you can't find any useful info anymore.
Re: (Score:2)
You think these web pages are created just to tick a box and not to actually be used by people? Probably. In that case it does not matter that they are slop.
Re: (Score:2)
And the last 15 years the corporate web pages have been dumbed down to irrelevancy so you can't find any useful info anymore.
And to think my mod points expired yesterday. This is my number one issue with web pages. Lack of information. If you're going to advertise a product or service, how about explaining what it does. Don't wave your hand and market speak everything. Flat out state, "This will allow you to create graphics beyond your imagination." (possibly a bit hyperbolic, but at least you know it's a graphics program)
The same goes for local businesses and what hours they're open. Don't make people scroll to the bottom o
Re: (Score:1)
It's not that it's fake. It's that most work is predictable.
I think it's missing some nuance (Score:5, Funny)
It's not the interplay between the decrease in literacy [futurism.com] in high school and college and artificial intelligence's improvement in text generation, it's a testament to the complexity of the problem — being critical to delve into the cognitive, cultural, and other foundations of human language and thought — in addition to examining not just the production, but also the consumption of language by both the human brain and large language models.
Re: (Score:1)
—
I see what you did there.
Re: I think it's missing some nuance (Score:2)
Re: (Score:2)
It is probably too short for any reliable verdict.
Re: (Score:2)
What's surprising is, even though you hit all of the things that were signaling AI, this still reads like something a human wrote. I wonder why?
Re: (Score:2)
Maybe because it was?
I also find that I find more and more channels on Youtube where a human is reading what sounds like an AI assisted script, as it's full of these same AI-isms. It's annoying when everyone starts sounding the same even when they are reading from a script.
Re: (Score:2)
Re: (Score:2)
Wow. This statement is basically empty while using complexity to obscure that fact. I would applaud you, but I have seen similar quite when correcting exams...
Stil a work of art.
Breaking news (Score:2)
A new tool that's really good at written language is being used by folks to edit copy. And the unsavory are using it to externalize costs and monetize slop. All this seems to distract everyone from the big issues like ownership and mass surveillance.
Expect "make it look human" post-processing (Score:1)
Expect AI-written material to be put through a "cleanup to make it look less like AI and more like a person" post-processing step in the very near future.
It's probably already happening.
Re: (Score:1)
Mass spamming (Score:2)
Yeah there have been tons of spam site popping up that are now produced with AI text instead of less sophisticated templates, it would be interesting to see if there's a downtick in the previous spam site markers.
SEO bait pages that we didn't need anyway (Score:2)
Google Search results are a joke these days.
Maybe Firefox should build a filter deleting all such pages.
Easy fix (Score:2)
Re: (Score:2)
Unfettered human greed will prevent that from happening. Unless we have put a pretty low cap on individual wealth, say 10-50M EUR or so and make circumvention an immediate "you loose everything" event. I would be all for that.
Reminds me of some medieval cultures that needed about 2-3 days of work per week to sustain themselves comfortably. Seems we have regressed from that, mostly because some really big assholes cannot get enough.
Re: (Score:2)
Reminds me of some medieval cultures that needed about 2-3 days of work per week to sustain themselves comfortably. Seems we have regressed from that, mostly because some really big assholes cannot get enough.
Those were hunter-gatherer cultures, not medieval.
Hunter-gatherer cultures produce minimal amounts of science, math, writing, medicine, art, clothing, or anything else. They work enough to survive.
Confounding variables (Score:2)
I'm curious how they dealt with the fact the language changes over time. It makes sense that people writing after AI would write more like an AI than people from before AI.
Re: (Score:2)
They deal with it by adding 'likely' in the conclusions.
I believe their conclusions though. But they also say 'significant signs of AI authorship' which is very stretchable and more cautious compared to other studies.
ugh (Score:2)
That last one really chaps my hide. Those mountains of fake drama posts punctuated by the 'negative parallelism' structure (now I have a name for the thing I hate.) It's everywhere and it's infuriating.
All you can do is try to ignore it and focus on your own (AI-free) stuff. Of course anything you post online will then become free AI fodder. It's hard to get away from it.
Some people write their love letters (Score:2)
is this really news? (Score:2)
Who cares? (Score:3)
If the product is an easily read, accurate website, that's a good outcome. Is writing website content a job we ought to worry about protecting? Next you'll suggest that writing user help screens that said nothing of value was worthwhile (yes I'm still resentful about that 35 years on).
Worthless (Score:2)
What is the point of having an AI write your website? You may as well not do it, because nobody will visit it anyways. People will just as the LLM directly. Talk about self-defeating stupidity.
Re: (Score:2)
News pages easily detectable (Score:2)
it's coming from both directions (Score:2)
On the one hand, ai is indeed getting better at writing banal fluff pieces that fill the Internet.
On the other, human writing is growing so awful in both subject and quality (this morning's bbc contains an article where one of their interns sprightly explains she's been able to train her self to wake without an alarm clock, golly!) that little of value is being supplanted.
Hell, I expect an ai editor of slashdot would post fewer dupes and likely be better at summarizing the point rather than cnp'ing whole 10
Of course (Score:2)
NOBODY write HTML code by hand these days, 1991 was a long time ago.
AI detectors? (Score:2)
I find the AI detectors more suspect than the Dark Web..
Snake oil is what those detectors are. Their false positive rate is huge.
So they found whatnow, and think that's okay? (Score:2)
There are some very obvious errors in their methodology if they've claimed to find "significant signs of AI authorship" in webpages that were written before ChatGPT.