You might recognize Gemini as the feature present in every Google service you utilize — whether you desire it or not.

While it has been a consistent, sometimes unwelcome aspect of Gmail for at least the previous year, it’s a comparatively recent addition to Maps. And guess what? It’s actually quite useful.

To put it through its paces, I had Gemini organize a full-day itinerary for me around the city. After about an hour or so of having Gemini locate various places for me — playgrounds near the new light rail extension, restaurants themed on vehicles that are family-friendly, you catch the drift — I was impressed. Some suggestions were predictable, but I also saved a few locations that I hadn’t considered before.

Gemini had a big task ahead of it: my own expectations. I’m an avid user of Google Maps. I rely on it for routine activities like navigating, but I also enjoy aimlessly browsing to discover anything new that grabs my attention. This method can lead to discovering real gems. I’ve stumbled upon bike trails, playgrounds, secret parks, and new coffee shops to test out. If I could, I would spend all day every day exploring Seattle via public transport, visiting bookstores and upscale stationery stores. It’s how I typically unwind on a day off, yet I often feel overwhelmed by the endless options, which leads me to revisit familiar neighborhoods. Thus, I had Gemini plan a route into unfamiliar areas.

Gemini appears as “Ask Maps” and provides a text box once you tap on it. It responds to inquiries using data in Google Maps with user feedback but can also access information from additional sources. If you inquire about whether to carry an umbrella for your trip across town, it can check the weather for you — that sort of functionality.

I provided it with my parameters: I would be using public transit and wanted a lunch stop, a pleasant walk somewhere, and a café where I could work on my laptop, in that order. I wanted to explore two different neighborhoods and needed to return home by 4:30. Its initial suggestions were very aligned with my preferences — a café adjacent to a bookstore and a dependable coffee shop downtown — but I had visited both places recently. After some back-and-forth, it was settled: tacos, greenery, and a Scandinavian-inspired café.

Tacos Chukis was known to me, yet I had never dined there. I nearly passed by the venue since it is located at the back of a building along with half a dozen other shops, lacking a sidewalk sign. However, Gemini guided me to the right place, and just in time: it had only opened 15 minutes before I arrived. My AI itinerary hinted that the house special with grilled pineapple was a popular choice, and I soon understood why. After three fantastic tacos, it was time to proceed to my next location.

However, I was ahead of schedule, so I asked Gemini to locate a unique shop nearby that I could visit before heading north to the park. It confidently recommended Elliott Bay Books — a fantastic spot, but it was certainly not “one block east,” as it claimed. This was the sole significant error I encountered during this experience, but it could have been quite inconvenient if I had followed its directions. Did I mention it was pouring rain outside?

Once I politely informed Gemini that it was suggesting I walk 10 minutes in the wrong direction, it recalibrated and directed me to Kobo: a lovely little shop with Japanese products. I had visited their other location a few times but didn’t realize there was one in the vicinity.

By the time I arrived at Volunteer Park, the front of my jacket was drenched. My umbrella was doing most of the work, but I needed it to shield my backpack that contained my laptop. Therefore, I was quite wet. Gemini had suggested a scenic route around the park or a tour of the conservatory — effectively a large greenhouse — if I wished to dry off. Simple choice.

Plants are truly amazing, you know. Did you realize there’s a tree that hollows itself to attract ants to inhabit it? And the ants fend off potential threats to protect the tree? That’s fascinating. That tree exists at the Volunteer Park Conservatory, which I had seen but never visited. Entry cost $6, and I felt a bit annoyed with Gemini for not mentioning that, but it proved to be a minor price to pay for time spent in a warm, serene sanctuary on a rainy day.

An employee at the conservatory noticed me admiring one of the towering palms and taking pictures with my phone. She guided me to the cacti room and insisted on capturing a photo of me with the giant cacti. “It’s so gorgeous in here I could cry!” she exclaimed as she left me among the cacti. I had to concur. There’s something ethereal about a cactus, and they come in so many diverse forms and sizes, down to the iconic saguaro that I mainly know from cartoons. There are fluffy ones, large round ones that resemble the world’s ugliest footstools, and another that appears to be coated in peeling wax paper. Up close, they possess a haunting and magical quality, akin to seeing an owl in its natural habitat. Not the type of scenery I expected on a very wet day in the Pacific Northwest.

I’ve recently been pondering how tech companies all seem to want us to use AI to make more purchases — especially after a recent discussion with my friend Will Sattelberg at 9to5Google. The way every AI demonstration culminates in booking a flight or purchasing a new pair of sneakers is becoming quite tiresome. However, it’s not solely a tech company issue — I’ve also been reflecting on my own habit of seeking a transaction whenever I leave home.

I tend to be drawn to places where I can depart with a new book, or a coffee, or a small treat, partially as a way to mitigate the anxiety of simply existing in the world. How do you select a place on a map when faced with thousands of options? What if I make the wrong choice and have a terrible experience? Acquiring a little keepsake reassures me that I visited somewhere worthwhile, or so I think. Yet the feeling never seems to endure, and soon enough, I find myself back on Google Maps plotting another adventure to track down that one perfectly curated home goods shop that will solve all my problems.

Regardless, I did leave the conservatory with some souvenirs — a soggy admission ticket and a pair of child-sized gardening tools from the gift shop. My kid adoring digging in the yard while I remove weeds, and it’s something I hope to encourage more. Sometimes, the perfect outing can just be playing in the dirt right outside, you know? Anyway, a warm, dry route 10 bus awaited me to take me to my final stop of the day: coffee.

I hadn’t previously heard of Day Made Kaffe, which struck me as strange, given that I frequent the area where it’s situated. Based on Gemini’s description of a minimalist yet inviting and laptop-friendly café, it also sounded like my type of venue. I had actually been there before, I realized upon entering, before it was a coffee shop. The establishment used to be a — you guessed it! — fancy home goods store where I purchased a couple of Christmas gifts in 2024. Time does indeed feel like a flat circle, etc.

Gemini didn’t steer me wrong; Day Made is incredibly my scene. The coffee was excellent, and the atmosphere was fantastic. The cardamom bun that Gemini recommended was not in stock, so I opted for a pastry with guava jelly as a substitute for the dreary weather. I watched the Artemis II launch on mute, left the café at 3:40 as Gemini advised, and boarded my final bus of the day. The time when I stepped through my front door at home? 4:26. Perfectly timed.

If my big adventure in the city was a success — and I believe it was! — it was made possible by people, not Gemini. People composed the reviews and suggestions that led me to Tacos Chukis. Gemini is merely the intermediary. However, when managing a vast and often intimidating dataset like that in Google Maps, a tool like Gemini seems incredibly valuable.

I frequently depend on user reviews to determine how genuinely kid-friendly a place actually is, and Gemini allows me to survey a multitude of those reviews across a broad area simultaneously to identify a location that serves dinosaur-shaped chicken nuggets alongside craft cocktails. Yet crucially, when it comes time to navigate from point A to point B, the LLM doesn’t just improvise and guide you through transit directions — you simply access the transit directions in Google Maps, which incorporates accurate real-time updates.

Gemini also does a commendable job of displaying its reasoning as it offers suggestions, allowing you to see the basis for its assertions. It isn’t free of hallucinations, and that’s a significant concern when relying on it to navigate the real world. Nevertheless, being aware of this limitation, I still consider it an impressive tool — whether you’re seeking a nearby restaurant equipped with high chairs right now because everyone is famished and cranky, or on a more relaxed journey of exploration.

Photography by Allison Johnson / The Verge

In January, folk musician Murphy Campbell discovered multiple tracks on her Spotify account that were not hers. These songs, which she had recorded, had never been uploaded by her to Spotify, and the vocals seemed peculiar.

She quickly concluded that someone had taken her performances from YouTube, crafted AI renditions, and posted them on streaming services under her name. I analyzed one of the tracks, “Four Marys,” using two separate AI detection tools, both indicating it was likely AI-generated.

Campbell was astonished, stating, “I was under the impression that we should have more safeguards before someone could do that. But, you know, it’s a lesson learned,” she told The Verge. It took a while for her to have the fraudulent songs taken down, “I became a nuisance,” she remarked. Even after that, it wasn’t a complete success. While the plagiarized tracks seem to have been removed from YouTube Music and Apple Music, at least one remains on Spotify, albeit under a different artist name but with the same title. Now there are several Murphy Campbells — “Naturally, I was excited about that,” remarked the genuine Murphy Campbell.

Spotify is trialing a new feature that would enable artists to manually verify tracks before they show up on their profiles, but Campbell is doubtful after her experience. “I feel like, each time a massive entity makes such commitments to musicians, it tends not to live up to expectations, but I’m curious to see how it works out in the future,” she expressed.

However, this was merely the onset of Campbell’s troubles.

On the day a Rolling Stone article was released detailing Campbell’s encounter with AI impersonators, a collection of videos were uploaded to YouTube via the distributor Vydia. These videos are not publicly viewable, and it remains unclear if anyone aside from the uploader, known as Murphy Rider, has viewed them. YouTube opted not to provide comments for this report.

These were utilized to assert ownership of content from several of Murphy Campbell’s videos. She received a notification from YouTube stating: “You are now sharing revenues with the copyright holders of the music identified in your video, Darling Corey.” The most puzzling aspect is that the tracks at the center of these claims are all in the public domain, including the renowned “In the Pines,” which dates back to at least the 1870s and has been reinterpreted by numerous artists from Lead Belly to Nirvana (as “Where Did You Sleep Last Night”).

Vydia has since resolved those claims, and spokesperson Roy LaManna stated that the individual who uploaded the videos has been banned from the platform. Of the more than 6,000,000 claims submitted by Vydia through YouTube’s Content ID system, only 0.02 percent were deemed invalid, which LaManna claims is “remarkable by industry standards.” He added, “We take pride in executing this correctly.”

LaManna also asserted that Vydia has no ties to Timeless IR or the AI videos that were released on streaming services labeled under Campbell’s name. Although the coincidence is dubious, LaManna insisted that the two occurrences are distinct.

Vydia has faced substantial backlash, including, as LaManna noted, “literal death threats” which resulted in office evacuations. Campbell is not letting Vydia off the hook, though she acknowledges that it’s not solely responsible. The intersection of generative AI, music distribution, and copyright is intricate, with numerous points of vulnerability and chances for exploitation. “I believe it runs deeper than we can comprehend,” Campbell stated.