Tiyu Xilu Station, colloquially known as “Hell West,” is a stark manifestation of Guangzhou Metro’s overcrowding.
Every day, hundreds of thousands of people chat and walk through the station concourse with their phones in hand, their silhouettes cast as dark outlines by the advertising light boxes.
But now, in addition to bosses, clients, and deadlines, commuters who rise early and work late, still half-asleep, must also face a new kind of fear—
That’s the “AI fake-person ad” ubiquitous in subway stations, constantly triggering the uncanny valley effect.
AI is too heavy-handed; the ads are really irritating to watch.
The proliferation of AI has always been quietly gradual; or rather, you know it’s bound to take off, but you never expect it to happen so quickly—like suddenly realizing that everyone around you is watching Hongguo short dramas overnight.
The same goes for AI-powered advertisements in subway stations.
AI advertising is actually nothing new. As early as two years ago, Shenzhen Metro’s own “Your Face” ad series, along with somewhat eerie creatives from Fliggy and JD.com, had already trended on social media.
But having AI-generated ads is one thing; having only AI-generated ads is another. Today, standing in a subway station and looking around, you’ll notice that nearly all the illuminated billboard ads are labeled “Image generated by AI,” delivering a completely different kind of psychological impact:
This day arrived much sooner than expected.
Of course, if it weren’t for the overly “AI‑flavored” advertising, we wouldn’t have even noticed the AI‑generated label tucked away in the corner of the poster.
In this regard, popular models such as GPT Images cannot be absolved of responsibility.
Image | A poster generated in one minute using ChatGPT
With every model update, readers in the comment sections of iFanr and APPSO invariably ask whether the “grainy” quality of the generated images has been addressed—only to end up disappointed. At least for now, AI-generated outputs that haven’t undergone post-processing still tend to reveal telltale flaws.
But what’s even more frustrating is that brands are racing to slap these blatantly fake, stock‑photo‑ready visuals straight onto their posters—thus kicking off a veritable feast of “fake people.”
The protein powder ad that has recently dominated the screens on Guangzhou’s subway could be described as having a “fake human scent” so overpowering it practically assaults your nose:
An utterly typical AI “average face,” with skin lacking any textural detail and a greasy, oily visual texture—enough to make anyone who’s scrolled through Douyin or used AI image‑generation tools utter the words, “I’ve seen this girl somewhere before.”
Even when I traveled to Changsha, I couldn’t escape the relentless push of this advertisement—
However, as the flagship product switched from the blue can to the white-and-gold can, the rider’s cycling jersey also changed from blue and black to a gray-and-white color scheme, inadvertently reviving the “Is the dress blue and black or white and gold?” debate that went viral 11 years ago.
Cruise tourism, which caters to high-end leisure needs, has likewise not been immune to this trend.
In a cruise ship advertisement video displayed at a subway station, a group of friends—three or five of them—hold wine glasses and gather on the deck, chatting happily; everything looks perfectly normal… until they reveal teeth that are almost eerily identical, as if copied and pasted.
Chimelong Paradise, which is preparing for its Halloween event at the end of the month, has devised an entire series of themed “alternate-space” promotional ads.
But whether it’s the “oily‑faced” characters or the out‑of‑place font on the interdimensional gate, none of it reveals even a hint of the park’s true setting; it doesn’t look like a theme park at all, but rather resembles a low‑quality escape‑room poster tucked away in a rundown building on Beijing Road.
From outdoor mall roll-up banners to interior train car decorations, similar examples could go on indefinitely—after all, these days, “traditional‑style advertising” created without AI is the real niche.
But when you look at these ads, you’re likely to share the same question I have: What exactly are AI-powered ads telling us?
Literally speaking, the purpose of “advertising” is to widely disseminate important information; however, it’s clear that with AI-driven advertising, we don’t actually receive any meaningful information—
AI personas can endorse health products, but they are inherently perfect, making the resulting health‑related imagery of little real relevance to ordinary consumers; in train‑carriage posters, an AI “dopamine girl” holds a cup of milk tea, yet everyone knows full well she could never have actually drunk it; and fictional cruise‑ship and theme‑park settings offer no additional insights for those of us who might actually visit these places.
If traditional advertising seeks to convey a product, service, brand, or idea, then in AI-driven advertising, the form has already diverged from the function, with nearly all information lost—leaving us with nothing but the anxiety of falling behind the AI wave.
Of course, we shouldn’t overemphasize traditional advertising either; there are quite a few poorly executed campaigns in that space as well.
The classic car commercials that fans love to see inserted mid‑TV show often just feature a car driving through deserts, over bridges, and across icy plains, set to soaring music, only to wrap up hastily before most viewers have even remembered the model’s name.
Image | Automotive advertising practice by Bilibili UP主 @Hong Hongs
But the difference is that the faces, lighting, and scenes mass-produced by AI happen to bear the most readily visible hallmarks of industrialization. This sense of incongruity, coupled with the uncanny valley effect that hits you head-on, almost instinctively triggers rejection and discomfort.
In this sense, AI advertising has unwittingly become a “developer” of low-quality ads.
The worst aspect of AI advertising is also its greatest contribution. That inescapable “AI flavor” acts like a fluorescent marker in academia, reminding us that formulaic, assembly-line‑produced ads have long since infiltrated every corner.
In other words, what we may truly dislike is not just AI itself, but the increasingly brazenly careless attitude that new technologies have enabled.
Creativity and sincerity are the secret ingredients of great content.
Ironically, before AI completely took over, subway stations weren’t exactly short on good advertisements.
Last March, the children’s clothing brand balabala launched a special photo exhibition titled “Chinese Children” at subway stations in multiple cities. In each child captured by the camera lens, there is their own unique blend of happiness and worries:
Some, having been passed over for class monitor, held back tears, their eyes filled with a touch of grievance and resentment; others hid in the courtyard, spying on their parents’ movements while secretly stashing away their old clothes, hoping their parents would buy them new ones; and still others, spotting soldiers riding by on their grandfather’s tricycle, instinctively mimicked the salute they’d seen on TV…
Swipe left to see more content
These subtle expressions on the children’s faces are something that today’s AI still cannot comprehend or replicate; yet it is precisely these details that give advertisements their emotional impact.
Meanwhile, also at Tiyu Xilu Station, during last year’s National Games, Heaven&Hell created a ketchup ad titled “The Tomato That Wants to Win Is at Heinz,” featuring dozens of tomato stems dressed up as athletes. The playful campaign cleverly highlighted both the “healthiness” of tomatoes and its support for the spirit of sportsmanship.
Unsurprisingly, the Xiaohongshu posts praising this ad quickly surpassed 100,000 likes. During that period, whenever I entered a subway station, I could always see passersby stopping in front of the ad and pulling out their phones to take pictures—
This is precisely the value of creativity.
The above discussion might easily lead one to believe that the widespread adoption of AI is the direct cause of a “downgrading of advertising aesthetics.” But quite the opposite is true: over the past period, there have been numerous cases of AI-assisted creative work as well:
For example, in June this year, the nearly ninety-year-old You Benchang authorized an AI digital avatar to star in a short commercial titled “At 93, I’ve Got My First Robot.” At the time, the “Uncle” was already too frail to handle high-intensity filming, but in the world created by AI, he appeared full of vigor, sharing the simple joy of a toilet‑assisting robot that lets elderly people get up at night with peace of mind.
At the end of the commercial, You Benchang himself appeared on screen to record a congratulatory video; and this ad also became the artist’s final work.
Although the on-screen disclaimer—“Both the visuals and the ad’s voiceover were generated with AI assistance”—appeared from the very first second, even Bilibili’s notoriously discerning audience didn’t fault the AI this time. That’s because the technology didn’t steal the spotlight; instead, it helped the ad convey what it truly intended to communicate—making it an ad with “soul.”
Ultimately, advertising is a service industry that serves brand philosophies and the audience’s sense of inner satisfaction; however, judging from the current state of subway stations, many practitioners seem to be producing ads through a manufacturing‑oriented lens, mistakenly believing that the sheer volume and speed of visual‑content production can substitute for creativity and sincerity.
How can we save public visual spaces?
Since AI is capable of making advertisements vibrant and engaging, why do the ones that ultimately appear in subway stations so often turn out to be the most bland?
The most straightforward answer, of course, is cost.
Traditional advertising production is a lengthy workflow: finding a photographer, renting a studio, hiring models, and even sourcing an appropriate background image all require constant coordination among budget constraints, timelines, copyright permissions, and design requirements.
Image | Pixabay@Pexels
Take Adobe Stock / Tuchong Creative as an example: commercial stock images come with different licensing tiers. If you’re just posting them on a WeChat official account, it’s relatively straightforward, with readily available, low-cost licensing packages. However, when used in “outdoor advertising” contexts—such as subway light boxes or storefront signage—the licensing process becomes extremely complex. You must call a dedicated hotline to negotiate outdoor commercial licenses, and the price for a single asset can easily soar to over a thousand yuan.
With AI, a single prompt can replace the myriad complex communications and friction that once plagued the process. For advertising agencies that need to deliver quickly, this is, of course, an irresistible advantage.
However, production and creativity are never the same thing. AI can only ensure the speed at which an idea is turned into a finished product, but it cannot guarantee that the idea itself is worth bringing to life.
Moreover, decision-making authority in commercial advertising has always been highly decentralized.
Designers may want to make things more interesting, but the account manager has to ensure the project timeline, while the brand’s marketing department—as the client—must consider leadership and market acceptance. When opinions diverge, the solution that ultimately survives multiple rounds of revisions may well be the most mediocre one.
More subtly, AI can now provide what appears to be a sophisticated explanation for this mediocre solution. Once AI is deployed, embracing new technology becomes politically correct, while whether the ad is creative or not becomes a secondary concern.
As a result, it’s easy to understand why AI-powered advertising is rapidly taking over subway stations—it represents the biggest common denominator of our time.
But we all know that individual rationality does not always lead to collective rationality.
While advertisements in subway stations may indeed be a compromise that all parties can reluctantly accept, the ubiquitous AI-generated avatars are subjecting millions of daily commuters to relentless visual overload and undermining the visual quality of urban public spaces.
Image | Xiaohongshu @Xunmeng Daben Gou
As for visual pollution in public spaces, we might borrow an approach from law and economics: whoever can prevent harm at a lower cost should be the one to take action.
For ordinary citizens who don’t want their eyes to be bombarded by AI-driven ads, the only options are to avoid entering subway stations altogether or to stare straight ahead at all times without looking at the advertisements—both of which are hardly practical. By contrast, tackling the issue at the source—by holding advertisers accountable for “who pollutes, who cleans up”—is clearly more cost-effective and feasible.
In fact, many cities renowned for their public visual spaces have imposed strict regulations on outdoor advertising: Paris’s Local Advertising Ordinance treats outdoor signs and advertisements as part of the city’s public landscape, requiring shopfronts and signage to harmonize with building facades in terms of color, shape, and lighting, thereby creating the familiar Parisian ambiance; meanwhile, New York has adopted its Zoning Code to ensure the cohesive and coordinated appearance of Manhattan’s streetfronts.
Perhaps one day, Chinese cities will also need to establish corresponding regulations for outdoor advertising management, preventing unregulated growth of such advertisements from becoming part of the “urban aesthetic ugliness.”
And until then, at the very least, we can use social media to make our voices heard, striving to ensure that brands listen to their audiences—until those heartfelt, beautifully crafted ads once again take center stage.
#Welcome to follow the iFanr official WeChat account: iFanr (WeChat ID: ifanr), where we bring you the latest and most exciting content第一时间.