Google tests AI-generated headlines in Discover, replacing publisher titles with misleading clickbait

Google is experimentally replacing news headlines with AI clickbait nonsense

Did you know that BG3 players exploit children? Are you aware that Qi2 slows older Pixels? If we wrote those misleading headlines, readers would rip us a new one -- but Google is experimentally beginning to replace the original headlines on stories it serves with AI nonsense like that. I read a lot of my bedtime news via Google Discover, aka "swipe right on your Samsung Galaxy or Google Pixel homescreen until you see a news feed appear," and that's where these new AI headlines are beginning to show up. They're not all bad. For example, "Origami model wins prize" and "Hyundai, Kia gain share" seem fine, even if not remotely as interesting as the original headlines. ("Hyundai and Kia are lapping the competition as US market share reaches a new record" and "14-year-old wins prize for origami that can hold 10,000 times its own weight" sound like they're actually worth a click!) But in the seeming attempt to boil down every story to four words or less, Google's new headline experiment is attaching plenty of misleading and inane headlines to journalists' work, and with little disclosure that Google's AI is rewriting them. The very first one I saw was "Steam Machine price revealed," which it most certainly was not! Valve won't reveal that till next year. Ars Technica's original headline was the far more reasonable "Valve's Steam Machine looks like a console, but don't expect it to be priced like one." "Microsoft developers using AI"? No shit, Sherlock. (That one was tacked on my colleague Tom Warren's story about "How Microsoft's developers are using AI" -- Google removed the six letters that make a silly headline into a real one!) I also saw Google try to claim that "AMD GPU tops Nvidia," as if AMD had announced a new groundbreaking graphics card, when the actual Wccftech story is about how a single German retailer managed to sell more AMD units than Nvidia units within a single week's span. Wccftech's headline was relatively responsible, but Google turned it into clickbait. Then there are the headlines that simply don't make sense out of context, something real human editors avoid like plague. What does "Schedule 1 farming backup" mean? How about "AI tag debate heats"? Make no mistake, the problem isn't just that these AI headlines are bad. It's that Google is taking away our agency to market our own work, like if we'd written a book and the bookstore decided to replace its cover. We try hard to craft headlines that invite readers in, ones that responsibly encapsulate the news, ones that help you understand why a story matters right away, and get you excited when it's justified. (Does my headline for this story seem the right amount of excited?) And yet Google seems to think it can just replace these headlines, in a way that might confuse our readers and think we're the ones generating clickbait, since our publication's names appear right next to them. Google does disclose that something about these news items is "Generated with AI, which can make mistakes," but not what, and readers only see that message if they tap the "See more" button: It's too easy for readers to think we intentionally send our stories to Google Discover with these headlines. The good news is, this is a Google experiment. If there's enough backlash, the company probably won't proceed. "These screenshots show a small UI experiment for a subset of Discover users," Google spokesperson Mallory Deleon tells The Verge. "We are testing a new design that changes the placement of existing headlines to make topic details easier to digest before they explore links from across the web." But the overall trend at Google has been to prioritize its own products at the expense of sending clicks to news websites. While the company swears it isn't destroying the web with AI search, you'd be hard pressed to find a news outlet that agrees, and even Google has admitted in court that "the open web is already in rapid decline." It's the reason The Verge now has a subscription: we can't survive Google Zero without your help.

Google Discover Trials AI Rewrites of News Story Headlines

Discover is a major source of news for millions of people around the world, and Google is continuing its experiments with AI on the platform. A new test uses AI to rewrite the headlines provided by publications, and sometimes it's doing so inaccurately. Spotted by The Verge, the trial is showing select users AI-generated headlines without the original post's title included until you click through. The AI-generated headlines shorten the description to four words with at least nine different instances appearing in the website's research. Some AI-generated rewrites have misunderstood the article and show false information in the replacement headline. An article from Ars Technica, titled "Valve's Steam Machine looks like a console, but don't expect it to be priced like one," was rewritten by the AI as "Steam Machine price revealed." Valve has yet to publicly comment on the price of its upcoming gadget. Another example saw a PC Gamer article with an original headline detailing how some Baldur's Gate 3 players were building an in-game army of non-player characters who are designed to look like children. It was retitled by Google's AI to "BG3 players exploit children," without a reference to those children being NPCs in the game. Some other examples took away the unique angle of a story they were recommending. An article written by The Verge on how Microsoft's team is using AI was retitled to "Microsoft developers using AI," losing the story's original context. PCMag was unable to activate AI-generated headlines. A Google spokesperson told The Verge that only a select "subset of Discover users" would see the "small UI experiment." The spokesperson for Google said, "We are testing a new design that changes the placement of existing headlines to make topic details easier to digest before they explore links from across the web." Google previously tested its own summaries of stories appearing in Discover, taking the article and providing its own AI-generated synopsis. Google said at the time it wanted to test ways to make it easier for readers to decide which websites to visit.

Google Discover is testing AI-generated headlines and they aren't good

Artificial intelligence is showing up everywhere in Google's services these days, whether or not people want them and sometimes in places where they really don't make a lick of sense. The latest trial from Google appears to be giving articles the AI treatment in Google Discover. The Verge noticed that some articles were being displayed in Google Discover with AI-generated headlines different from the ones in the original posts. And to the surprise of absolutely no one, some of these headlines are misleading or flat-out wrong. For instance, one rewritten headline claimed "Steam Machine price revealed," but the Ars Technica article's actual headline was "Valve's Steam Machine looks like a console, but don't expect it to be priced like one." No costs have been shared yet for the hardware, either in that post or elsewhere from Valve. In our own explorations, Engadget staff also found that Discover was providing original headlines accompanied by AI-generated summaries. In both cases, the content is tagged as "Generated with AI, which can make mistakes." But it sure would be nice if the company just didn't use AI at all in this situation and thus avoided the mistakes entirely. The instances The Verge found were apparently "a small UI experiment for a subset of Discover users," Google rep Mallory Deleon told the publication. "We are testing a new design that changes the placement of existing headlines to make topic details easier to digest before they explore links from across the web." That sounds innocuous enough, but Google has a history of hostility towards online media its frequent role as middleman between publishers and readers. Web publishers have made multiple attempts over the years to get compensation from Google for displaying portions of their content, and in at least two instances, Google has responded by cutting out those sources from search results and later claiming that showing news doesn't do much for the bottom line of its ad business. For those of you who do in fact want more AI in your Google Search experience, you're in luck. AI Mode, the chatbot that's already been called outright "theft" by the News Media Alliance, is getting an even more symbiotic integration into the mobile search platform. Google Search's Vice President of Product Robby Stein posted yesterday on X that the company is testing having AI Mode accessible on the same screen as an AI Overview rather than the two services existing in separate tabs.

Google is replacing Discover news headlines with laughably awful AI-generated titles

Google's AI also generated the title "Steam Machine price revealed" for an Ars Technica story that doesn't reveal the price at all. Either way, the search giant is churning out four-word headlines of mostly crappy quality. A Google representative told The Verge that this was just a test for now rather than a full-scale feature release: This statement suggests that these AI-generated titles won't see a broad release, at least not in their current state. Nevertheless, it's concerning that Google saw fit to push this experiment out to any users in the first place when the results are so obviously awful. There's also no visible label or disclosure that this is an AI-generated title, and no disclosure that Google is behind these headlines rather than a publisher. So we can see more than a few readers getting angry at publications after being duped by a low-effort clickbait title.

Google's toying with nonsense AI-made headlines on articles like ours in the Discover feed, so please don't blame me for clickbait like 'BG3 players exploit children'

Pop quiz, hotshot: what was the headline on this article when you clicked on it? Was it classic PC Gamer style -- witty, insightful, with the rare power of capturing the essence of a story with neither artifice nor evasion, and unintentionally but unmistakably implying the incredible mental powers and physical beauty of the writer? Or did it say something like "The headlines are all screwed"? If the latter, bad luck sport: you may have fallen prey to Google's latest experiment with its Discover newsfeed, replacing human-created headlines with sometimes-meaningless AI slop. As spotted by The Verge, one of the corporation's latest AI adventures is cramming it into your news feed, taking headlines like ours and condensing them into something at best shorter and clickbait-ier, and at worst actively nonsensical. So, for instance, Lincoln's headline, "'Child labor is unbeatable': Baldur's Gate 3 players discover how to build an army of unkillable kids through the power of polymorph and German media laws" became, ah, "BG3 players exploit children." Harvey's "Schedule 1 creator had a backup plan if Steam rejected it -- pack up the product, don a farmer's hat, and 'pivot it to be a farming game' like Stardew Valley" became the completely incomprehensible "Schedule 1 farming backup". It's not just us it's happening to, of course (though you may, and perhaps should, get angry about the bastardisation of our precious words most of all). Poor Ars Technica, for instance, had an article with the original headline "Valve's Steam Machine looks like a console, but don't expect it to be priced like one." All very reasonable. Google's AI turned that into "Steam Machine price revealed," which is actively misleading. That last one really sticks in my craw. Those of you who studied something useful at university and therefore don't work in journalism might not be aware of this, but Google has certain rules it likes sites like ours to follow, and woe betide any who violate them, because they could find themselves demoted in the algorithm. The rules mostly make sense! For instance, ol' Goog is very insistent that sites don't write things like 'release date revealed' in the headlines for articles that, you know, aren't about a release date being revealed, or which are actually about a wide release window. Entirely fair. Then, of course, Google's own hallucination engine turns around and slaps precisely that kind of misleading headline on Ars Technica's story, which did not have a misleading headline originally. It feels like a bad joke for the corporation to throw out its own rules like this in pursuit of slapping a shareholder-powered "AI-driven" badge on yet another enshittified product. Even worse, Google tucks away its "AI-generated" notice behind a See More button, leaving readers likely to assume the terrible headlines belong to the sites in question themselves. "These screenshots show a small UI experiment for a subset of Discover users," a Google rep told The Verge. "We are testing a new design that changes the placement of existing headlines to make topic details easier to digest before they explore links from across the web." Well, I'd say mission failed on that front. With any luck, this is an experiment that's soon to end.