Hallucinating Headlines: The AI-Powered Rise of Fake News

The number of AI-powered fake news sites has now surpassed the number of real local newspaper sites in the U.S.

How? AI tools have made creating entire fake news sites quicker and easier than before — taking one person minutes to create what once took days for dozens and dozens of people.

Researchers say we crossed this threshold in June 2024, a “sad milestone” by their reckoning.[i] As traditional, trusted sources of local news shut down, they’re getting replaced with sensationalistic and often divisive fake news sites. What’s more, many of these fake news sites pose as hometown newspapers.

They’re anything but.

These sites produce disinformation in bulk and give it a home. In turn, the articles on these fake news sites fuel social media posts by the thousands and thousands. Unsuspecting social media users fall for the clickbait-y headlines, click the links, read the articles, and get exposed to yet more “news” on those sites – which they then share on their social feeds thinking the stories are legit. And the cycle continues.

As a result, social media feeds find themselves flooded with falsehoods, misrepresentations, and flat-out lies. Researchers spotted the first of them in mid-2023, and they number of them are growing rapidly today.

In all, the rise of AI-powered fake news sites now plays a major role in the spread of disinformation.

What is disinformation — and misinformation?

When we talk about so-called “fake news,” we’re really talking about disinformation and misinformation. You might see and hear those two terms used interchangeably. They’re different, yet they’re closely related.

Disinformation is intentionally spreading misleading info.

Misinformation is unintentionally spreading misleading info (the person sharing the info thinks it’s true).

This way, you can see how disinformation spreads. A bad actor posts a deepfake with deliberately misleading info — a form of disinformation. From there, others take the misleading info at face value and pass it along as truth via social media — a form of misinformation.

The bad actors behind disinformation campaigns know this relationship well. Indeed, they feed it. In many ways, they rely on others to amplify their message for them.

The explosion of fake news sites

With that, we’re seeing an explosion of fake news sites with content nearly, if not entirely, created by AI — with bad actors pushing the buttons.

Funded by partisan operations in the U.S. and by disinformation operations abroad, these sites pose as legitimate news sources yet push fake news that suits their agenda — whether to undermine elections, tarnish the reputation of candidates, create rifts in public opinion, or simply foster a sense of unease.

One media watchdog organization put some striking figures to the recent onrush of fake news sites. In May 2023, the organization found 49 sites that it defined as “Unreliable AI-Generated News Websites,” or UAINS. In February 2024, that number grew to more than 700 UAINS.[ii]

Per the watchdog group, these sites run with little to no human oversight. Additionally, they try to pass themselves off as legitimate by presenting their AI “authors” as people.[iii] Brazenly, at least one publisher had to say this when confronted with the fact that his “reporter” bylines were really AI bots:

The goal was to create “AI personas” that can eventually “grow into having their own following,” maybe even one day becoming a TV anchor. “Each AI persona has a unique style … Some sort of — this is probably not the right word — personality style to it.” [iv]

Beyond spreading disinformation, these sites are profitable. Recent research found that among the top 100 digital advertisers, 55% of them had their ads placed on disinformation sites. Across all industries and brands, 67% of those with digital ads wound up on disinformation sites.[v]

To clarify, these advertisers support these disinformation sites unwittingly. The researchers cite the way that online advertising platforms algorithmically place ads on various sites as the culprit. Not the advertisers themselves.

So as we talk about disinformation sites cropping up at alarming rates, we also see bad actors profiting as they prop them up.

Many AI-powered fake news sites try to pass themselves off as “hometown” papers

Follow-up research pushes the estimated number of AI-powered fake news sites yet higher. In June, analysts discovered 1,265 sites targeting U.S. internet users with fake news – many posing as “local” news outlets. Shockingly, that figure surpasses the number of local newspapers still running in the U.S., at 1,213 outlets.[vi] (Side note: between 2005 and 2022, some 2,500 local newspapers shuttered in the U.S.[vii])

The actors and interests behind these sites follow a straightforward formula. In word salad fashion, they’ll mix the name of a town with classic publication names like Times, Post, or Chronicle to try to give themselves an air of credibility. Yet the content they post is anything but credible. AI generates the content from tip-to-tail, all to suit the disinformation the site wants to pump out.

The U.S. isn’t alone here. Similar sites have cropped up in the European Union as well. The European Union’s Disinformation Lab (EU DisinfoLab) found that outside actors mimicked several legitimate European sites and used them to spread disinformation.[viii] Legitimate sites that outside actors mimicked included Bild, The Guardian, and the NATO website.

How can you spot an AI-generated fake news website?

The answer is that it’s getting tougher and tougher.

Fake news sites once gave off several cues that they were indeed fake, whether because they were created by earlier, cruder versions of AI tools or by human content creators. They simply didn’t look, feel, or read right. That’s because it took a lot of manual work to create a fake news site and make it look legitimate.

For starters, the site needed a sharp visual design and an easy way of surfacing articles to readers. It also meant cooking up a virtual staff, including bios of owners, publishers, editors, and bylines for the writers on the site. It also called for creating credible “About” pages and other deeper site content that legitimate news sites feature. Oh, and it needed a nice logo too. Then, and only then, could the actors behind these sites start writing fake news articles.

Now, AI does all this in minutes.

The Poynter Institute for Media Studies, a non-profit journalism school and research organization, showed how it indeed took minutes using several different AI tools.[ix] One tool created fake journalists, along with backgrounds, bylines, and photos. Another tool provided the framework of web code to design and build the site. As for the articles themselves, a few prompts into ChatGPT wrote serviceable, if not bland, articles in minutes as well.

As a result, these sites can look “real enough” to casual viewers. Taken at face value, all the trappings of a legitimate news site are there, with one exception — the articles. They’re fake. And they go on to do the damage that the bad actors behind them want them to do.

So, what news can you trust online?

The people who create these fake news sites rely on others to take the lies they push at face value — and then immediately react to the feelings they stir up. Outrage. Anger. Dark joy. Without pause. Without consideration. If an article or post you come across online acts taps into those emotions, it’s a sure-fire sign you should follow up and see if what you’ve stumbled across is really real.

Here are a few things you can do:

Seek out objective reporting.

Outside of a newspaper’s Op-Ed pages where editorial opinions get aired, legitimate editorial staff strive for objectivity—reporting multiple dimensions of a story and letting the facts speak for themselves. If you find articles that are blatantly one-sided or articles that blast one party while going excessively easy on another, consider that type of reporting a red flag.

Watch out for clickbait.

Sensationalism, raw plays to emotion, headlines that conjure outrage — they’re all profitable because they stir people up and get them to click. Content like this is the hallmark of fake news, and it’s certainly the hallmark of AI-powered fake news as well. Consider stories like these as red flags as well.

Use fact-checking resources.

Come across something questionable? Still uncertain of what you’re seeing? You can turn to one of the several fact-checking organizations and media outlets that make it their business to separate fact from fiction. Each day, they assess the latest claims making their way across the internet — and then figure out if they’re true, false, or somewhere in between.

Check other known and long-standing news sources.

Search for other reputable sources and see what they’re saying on the topic. If anything at all. If the accounts differ, or you can’t find other accounts at all, that might be a sign you’re looking at fake news.

Additionally, for a list of reputable information sources, along with the reasons they’re reputable, check out “10 Journalism Brands Where You Find Real Facts Rather Than Alternative Facts.” It’s published by Forbes and authored by an associate professor at The King’s College in New York City.[x] It certainly isn’t the end-all, be-all of lists, yet it provides you with a good starting point. Both left-leaning and right-leaning editorial boards are included in the list for balance.

Stick with trusted voter resources.

With Election Day coming around here in the U.S., expect many bad actors to push false voting info, polling results, and other fake news that tries to undermine your vote. Go straight to the source for voting info, like how to register, when, where, and how to vote — along with how to confirm your voting registration status. You can find all this info and far more with a visit to https://www.usa.gov/voting-and-elections.

You can find another excellent resource for voters at https://www.vote411.org, which is made possible by the League of Women Voters. Particularly helpful is the personalized voting info it offers. By entering your address, you can:

See what’s on your ballot.
Check your voter registration.
Find your polling place.
Discover upcoming debates in your area.

If you have further questions, contact your state, territory, or local election office. Once again, usa.gov offers a quick way to get that info at https://www.usa.gov/state-election-office.

[i] https://www.newsguardtech.com/press/sad-milestone-fake-local-news-sites-now-outnumber-real-local-newspaper-sites-in-u-s/

[ii] https://www.newsguardtech.com/press/newsguard-launches-2024-election-misinformation-tracking-center-rolls-out-new-election-safety-assurance-package-for-brand-advertising/

[iii] https://www.bloomberg.com/news/newsletters/2024-05-17/ai-fake-bylines-on-news-site-raise-questions-of-credibility-for-journalists

[iv] Ibid.

[v] https://www.nature.com/articles/s41586-024-07404-1

[vi] https://www.newsguardtech.com/press/sad-milestone-fake-local-news-sites-now-outnumber-real-local-newspaper-sites-in-u-s/

[vii] https://localnewsinitiative.northwestern.edu/research/state-of-local-news/2022/report/

[viii] https://www.cybercom.mil/Media/News/Article/3895345/russian-disinformation-campaign-doppelgnger-unmasked-a-web-of-deception/

[ix] https://www.poynter.org/fact-checking/2023/chatgpt-build-fake-news-organization-website/

[x] https://www.forbes.com/sites/berlinschoolofcreativeleadership/2017/02/01/10-journalism-brands-where-you-will-find-real-facts-rather-than-alternative-facts