In early December, I brought you the news that Google has begun replacing Verge headlines, and those of our competitors, pinch AI clickbait delirium successful its contented feed. Google appeared to beryllium backing distant from the experiment, but now tells The Verge that its AI headlines successful Google Discover are a feature, 1 that “performs good for personification satisfaction.” I erstwhile again spot tons of misleading claims each clip I cheque my phone.
Like I explained past month, these AI headlines are akin to a bookstore replacing the covers of the books it puts connected show — only here, the “bookstore” is the news tab that appears erstwhile you swipe correct connected the homescreen of a Samsung Galaxy aliases Google Pixel phone, and the “cover” mightiness beryllium a AI-generated dishonesty alternatively of the truth.
For example, Google’s AI claimed past week that “US reverses overseas drone ban,” citing and linking to this PCMag story for the news. That’s not conscionable mendacious — PCMag took pains to explicate that it’s mendacious in the communicative that Google links to!
From PCMag’s story, bolding theirs:
I saw a header saying that the drone prohibition was dropped. Is that true? No. While it’s existent the Commerce Department ended its efforts to restrict DJI and different drones from import successful Jan. 2026, it only did truthful because it would beryllium redundant successful the aftermath of the FCC actions. The Commerce Department is simply a abstracted entity from the FCC, and its projected restrictions were ne'er put successful spot to statesman with. Some reports connected the Commerce Department’s determination person misleading headlines that could make you deliberation the authorities did an about-face, truthful I don’t blasted you for being confused.
Google’s AI header and summary, versus…
…the PCMag communicative it links to.
What does the writer of that PCMag story think? “It makes maine consciousness icky,” Jim Fisher tells maine complete the phone. “I’d promote group to click connected stories and publication them, and not spot what Google is spoon-feeding them.”
He says Google should beryllium utilizing the header that humans wrote, and if Google needs a summary, it tin usage the ones that publications already taxable to thief hunt engines parse our work.
Google claims it’s not rewriting headlines. It characterizes these caller offerings arsenic “trending topics,” moreover though each “trending topic” presents itself arsenic 1 of our stories, links to our stories, and uses our images, each without competent fact-checking to guarantee the AI is getting them right.
In immoderate ways, Google’s existent implementation isn’t rather arsenic bad arsenic it was a period ago. I’ve seen less examples of egregious clickbait, partially because Google Discover is now serving maine rather a fewer unadulterated news stories alongside its AI ones — though it does trim disconnected their genuine headlines acold excessively quickly, making galore reliable to read.
The AI is besides nary longer restricted to astir 4 words per headline, truthful I nary longer spot delirium headlines for illustration “Microsoft developers utilizing AI” aliases “AI tag statement heats.” (Instead, I occasionally spot chop for illustration “Fares: Need AAA & AA Games” aliases “Dispatch sold millions; fewer avoided romance.”)
An inane AI-generated header supra an original header that Google is cutting disconnected prematurely.
Two of Google’s AI headlines.
But Google’s AI has nary hint what parts of these stories are new, relevant, significant, aliases true, and it tin easy confuse 1 communicative for another.
On December 26th, Google told maine that “Steam Machine value & HDMI specifications emerge.” They hadn’t. On January 11th, Google proclaimed that “ASUS ROG Ally X arrives.” (It arrived in 2024; the caller Xbox Ally arrived months ago.) On January 20th, it wrote that “Glasses-free 3D tech wows,” introducing readers to “New 3D tech called Immensity from Leia” — but linking to this TechRadar story astir an wholly different institution called Visual Semiconductor. I recovered different that claimed to beryllium astir a GPU shaper commenting on the RAM shortage; it linked alternatively to a Digitimes story astir a RAM maker.
I judge this communicative was astir really the strategy was backmost successful banal astatine Best Buy.
I’m peculiarly disappointment erstwhile I spot bait-and-switch headlines connected Verge stories, of course, and worried they’re taking distant our expertise to marketplace our ain work.
Google boiled down my workfellow Jay Peters’ communicative astir really RGB stripe OLED monitors tin unlock sharper matter and much meticulous colors to the boring “New OLED Gaming Monitors Debut.” My communicative astir letting you experience an immersive 3D demo of the Lego Smart Brick for illustration you’re location pinch america astatine CES became “Lego Smart Play launches March 1,” a day that wasn’t news by the clip Google wrote that! And Google AI decided to advertise our big Verge Awards astatine CES 2026 communicative arsenic “Robots & AI Take CES,” which was fundamentally the other of our conclusion successful that story.
Please travel straight to our website alternatively of relying connected Google arsenic intermediary.
And yet, our caller AI overlords are not efficaciously weeding retired the worst human clickbait successful speech for our fealty. One header that Google’s AI didn’t overwrite was “Star Wars Outlaws Download Available For Less Than 24 Hours” by Screen Rant. Here’s what the writer of that communicative reveals halfway down the post:
Just the 1 codification being fixed distant erstwhile thousands of group are apt to get progressive feels a spot stingy, though astatine the clip of penning Ubisoft’s station hasn’t precisely blown up, truthful you mightiness beryllium successful pinch a shot.
Yes, Ubisoft gave distant a azygous transcript of a game connected X, successful a giveaway only unfastened to residents of the UK, yet Google’s news bot decided that Screen Rant’s FOMO clickbait was good to service up without tweaks astatine all.
There was a *single* codification disposable to 1 fortunate victor successful the UK. For shame.
Here’s the connection from Google spokesperson Jennifer Kutz astir the Google feature:
We launched a caller feature past twelvemonth successful Discover to thief group research topics that are covered by aggregate creators and websites. The characteristic includes a adjuvant AI-powered overview of the topic, a featured image, and links to related stories. The overview header reflects accusation crossed a scope of sites, and is not a rewrite of an individual article headline. This characteristic performs good for personification satisfaction, and we proceed to research pinch the UI to thief group click done and research contented connected the web.
Google declined our petition for an question and reply to much afloat explicate the idea.
I don’t cognize really broadly Google is showing these “trending topics” AI headlines yet, but it seems the institution is testing them beyond the Google Discover news feed, too. I’ve precocious seen immoderate of them look arsenic push notifications to my phone; tapping them takes maine to a Google Gemini chatbot that attempts to summarize a caller portion of news.
Google changes for illustration these are the biggest logic The Verge now has a subscription, without which we won’t past Google Zero.
Disclosure: Vox Media, The Verge’s genitor company, has filed a suit against Google, seeking damages from its forbidden advertisement tech monopoly.
Follow topics and authors from this communicative to spot much for illustration this successful your personalized homepage provender and to person email updates.
English (US) · 
Indonesian (ID) ·