内容来源：https://www.wired.com/story/real-estate-is-entering-its-ai-slop-era/

内容总结：

【房产市场观察】AI虚拟装修席卷美国房产中介，消费者担忧被“照骗”误导

当你在房产平台上浏览美国富兰克林的待售房屋时，可能会被一段展示豪华装修的视频吸引：四柱床、藏酒丰富的酒窖、按摩浴缸一应俱全，笑容得体的中介正用舒缓语调讲解。但这一切美好景象实则全是人工智能生成的虚拟幻影——房间实际空无一物，中介的语音表情由文本生成，连镜头运动都由AI操控。

前脸书产品经理古普塔创立的AutoReel应用正掀起行业变革。该工具可将房屋图片自动转为视频，目前每日生成500-1000条新房源视频，已覆盖美国、新西兰和印度市场。全美房地产经纪人协会创新战略总监韦斯曼透露，行业AI使用率已飙升至80%-90%。

这种技术革新正在制造新的消费风险。密歇根州房主伊丽莎白近期发现，当地房源图片出现违反物理结构的“悬空楼梯”，厨房橱柜离奇消失，草地替代了后院地砖。她在Reddit平台发布的对比图引发1200多条评论，网友直指AI修饰已超出合理范围。

尽管从业者强调技术能节省数千美元成本和一周制作时间，但行业伦理问题日益凸显。美国房地产协会联合创始人哈伯指出，AI会机械堆砌“坐落于黄金地段”等套路化描述，而专业经纪人应当善用技术而非放弃思考。

更令人担忧的是，目前美国对AI生成房源内容尚未建立明确监管规范。全国房地产经纪人协会已警示，使用AI图像的法律边界仍属“灰色地带”，违规者可能面临罚款和诉讼。有摄影师实测发现，即便采用防造假技术，AI仍会在空房间自动生成虚拟沙发。

随着购房者对社会媒体中AI视频的反感情绪蔓延，房产行业正面临信任考验。正如业内人士所言：“人们将做出人生最大投资，他们不愿在踏进房门之前就遭受欺骗。”

中文翻译：

当你在富兰克林市的房产 listings 中寻觅新居时，一段竖屏视频映入眼帘：宽敞的房间里陈列着四柱床，藏酒丰富的酒窖，还有精致的浸泡式浴缸。视频角落里，笑容可掬的房产经纪人正用令人放松的语调讲解这处梦想之家。一切看起来完美无瑕——或许完美得有些失真。

蹊跷何在？视频里所有画面皆由人工智能生成。真实房源其实是空置状态，奢华家具全凭虚拟舞台布置技术呈现。经纪人的画外音与表情源自文字指令生成，就连镜头在房间内的缓慢平移也是AI编排的结果——因为根本不存在实拍的摄像机。

"任何房产经纪人都能在家里几分钟内轻松制作出这样的视频，"Facebook前产品经理、Snapchat前软件工程师阿洛克·古普塔表示。他联合创立的AutoReel应用程序，可帮助经纪人将房源图片转化为视频。据他透露，目前每天有500至1000个新房源视频通过该应用生成，全美乃至新西兰和印度的经纪人都在运用这项技术营销数千处房产。

这只是众多AI工具的冰山一角。随着OpenAI的ChatGPT和谷歌的Gemini等知名工具的普及，房地产行业正被迅速重塑成某种难辨真伪的形态。

"最近几周我参加了多场行业会议，现场调研显示，100位观众中约有80%到90%举手表示正在使用AI，"全美最大房地产行业协会"全国房地产经纪人协会"的创新策略总监丹·韦斯曼透露，"使用量呈现爆发式增长。"

与多数行业相同，房地产巨头们正争相拥抱这波生成式AI浪潮。这些工具承诺提升效率、降低成本、颠覆客户体验的每个环节。但对于租房或购房这类人生重大消费决策，AI生成的图片、视频和房源描述反而可能加剧交易过程中的不确定性。

密歇根州乡村地区的业主伊丽莎白（因隐私考虑未公开姓氏）长期关注当地房源信息以掌握房产估值。她回忆道："大约两三周前，我首次发现某处房源的真实照片被AI处理过。"最引人注目的是图片泛着的黄调——这已成为AI生成图像的民间识别标志，甚至催生了号称能"去黄化"的AI修正工具。

"继续翻看照片时，我注意到诸多不合逻辑之处：楼梯通向虚无，整体画面充满卡通感。"她的怀疑在查看到同一物业的第二组照片时得到证实：原始图像中厨房橱柜不翼而飞，后院地砖被草坪替代，窗户尺寸也被大幅修改。将这些对比图发布到Reddit的"轻度恼人"版块后，她收到了1200多条评论。

"这完全是在误导消费者，扭曲房屋真实特征。"她强调，虽然房产中介惯用鱼眼镜头营造空间感，但AI技术"让我们进入了全新的造假维度"。

伊丽莎白的遭遇并非个例。近月来社交媒体上频现消费者对AI误导性房源展示的投诉：纽约市StreetEasy平台上将狭小阁楼包装成主卧的公寓，底特律某住宅被AI添加全新屋顶的外立面改造。《连线》杂志联系纽约涉事经纪人未获回应，底特律房源联系人则称图片由其他经纪人提前发布，不确定是否使用AI。

尽管如此，行业领袖们并未表现出担忧，部分人士直言可观的利润空间让使用AI工具成为必然选择。"既然能用ChatGPT在45秒内免费完成，何必花500美元把空房间照片交给虚拟布置师处理四天？"持牌经纪人、美国房地产协会联合创始人贾森·哈伯反问道，"虚拟渲染已存在20年，如今AI能轻松实现，整个衍生行业正在洗牌，相关从业者不得不转型。"

作为拥有2.2万名会员的协会负责人，哈伯强调房地产从业者必须披露AI使用情况，如同过去要求明示虚拟布置技术。欺骗性操作可能引发罚款与诉讼，全国房地产经纪人协会已提醒会员，AI生成图像的法律界定仍处于"灰色地带"。该组织职业道德准则明确禁止使用误导性图像。

哈伯指出行业已出现"AI滥用流行病"的典型症状：ChatGPT"几乎必然"在房源描述中插入"坐落于"（nestled）一词。他在Instagram发文揭露，诸如"坐落于黄金地段""坐落于城市核心区""坐落于两栋住宅之间"等套话，都是经纪人直接复制AI文本的铁证。

这位科技拥抱者同时警告："技术不能成为放弃独立思考的借口。如果你只会机械操作，就称不上优秀经纪人——缺乏创造力，无法形成差异化优势。"

AI视频生成器联合创始人古普塔则表示，随着社交媒体成为主要获客渠道，保持用户专注的高质量视频至关重要。使用AutoReel可节省"500至1000美元"专业摄像费用，并将制作周期缩短近一周。"两年前客户普遍拒绝这类技术，2024年他们开始主动询价，如今最常问的是'如何入门'。"

但并非所有从业者都全盘接受。拥有近十万订阅者的房地产摄影教学频道主内森·库尔试用过AutoReel等AI工具后认为，竖屏视频拍摄本身仍是性价比高的附加服务，尽管复杂制作仍需要更高成本。

此外，AI幻觉问题持续存在。古普塔称AutoReel经过数百万真实房源视频训练，能避免无中生有——通过设计缩放而非旋转镜头来防止虚构细节。在用真实房源照片的测试中该技术表现良好，但处理伊丽莎白发现的修饰照片时，系统仍生成了虚假沙发。

库尔指出，即便AI产品能以假乱真，许多用户早已厌倦社交媒体信息流中的AI生成视频，而购房决策的特殊性更放大这种反感："购房者即将做出人生最大投资，他们不愿在踏进房门之前就受到欺骗。"

英文来源：

As you’re hunting through real estate listings for a new home in Franklin, Tennessee, you come across a vertical video showing off expansive rooms featuring a four-poster bed, a fully stocked wine cellar, and a soaking tub. In the corner of the video, a smiling real estate agent narrates the walk-through of your dream home in a soothing tone. It looks perfect—maybe a little too perfect.
The catch? Everything in the video is AI-generated. The real property is completely empty, and the luxury furniture is a product of virtual staging. The realtor’s voice-over and expressions were born from text prompts. Even the camera’s slow pan over each room is orchestrated by AI, because there was no actual video camera involved.
Any real estate agent can create “exactly that, at home, in minutes,” says Alok Gupta, a former product manager at Facebook and software engineer at Snapchat who cofounded AutoReel, an app that allows realtors to turn images from their property listings into videos. He said that between 500 and 1,000 new listing videos are being created with AutoReel every day, with realtors across the US and even in New Zealand and India using the technology to market thousands of properties.
This is one of many AI tools, including more familiar ones like OpenAI’s ChatGPT and Google’s Gemini, that are quickly reshaping the real estate industry into something that isn’t necessarily, well, real.
“I’ve been at a few conferences over the past few weeks, and just anecdotally speaking, we’ll ask out of 100 people in the audience how many are using AI, and I’d say 80 to 90 percent of people raise their hand,” says Dan Weisman, the director of innovation strategy at the National Association of Realtors, the largest real estate trade association in the US. “We are seeing this huge uptick in people using it.”
Like most industries, the biggest names in this one are rushing to embrace a wave of generative AI products making big promises about increasing productivity, cutting costs, and revolutionizing every aspect of the consumer experience. But when it comes to renting or buying a home, which are typically the costliest parts of adult life, the use of AI-generated photos, videos, and listing descriptions can make the process feel even riskier.
Elizabeth, a homeowner in rural Michigan who did not want her last name used due to privacy concerns, keeps an eye on local real estate listings to stay abreast of her own home’s value.
“About two or three weeks ago was the first time that I’ve truly seen actual pictures of a house fed through AI,” she says. The first thing that stuck out to her was that the images in the listing had a yellowish hue, which has become a colloquially recognized sign of AI to the point that there are now other AI tools that purport to “UnYellow” AI-generated images.
“And then, as I was scrolling through the photos, I noticed that some things just weren’t making sense. There were stairways leading to nowhere,” Elizabeth says. “In general, it just looked cartoonified.”
Her suspicions were confirmed when she came across a second listing for the same property and saw the original images that had been transformed. In the edited versions, kitchen cabinets were missing, backyard pavement was replaced by grass, and windows were dramatically resized. Elizabeth posted the two sets of images on Reddit in the popular “mildly infuriating” subreddit, and more than 1,200 people commented.
“This is misleading. It’s distorting the features of the house,” she continued. She says real estate listings often use a fish-eye lens to make rooms look bigger, but with AI “we’ve entered a whole new realm.”
Elizabeth isn’t alone. Examples of consumer outrage over potentially AI-driven misleading listings have cropped up all over social media in recent months, from a New York City apartment listing on StreetEasy where a tiny loft became a master bedroom to an edited house facade, complete with a new roof, in Detroit. WIRED reached out to the real estate agent behind the New York City listing and has not received a response. A real estate agent whose number was included on the Detroit listing said the images were made by a broker and posted prematurely and that she wasn’t sure if AI was used.
Still, industry leaders aren’t necessarily concerned, and some say the profit margin makes the tools a no-brainer.
“Why would I send my photos of an empty room to a virtual stager, have them spend four days and send it back to me at a charge of 500 bucks when I can just do it in ChatGPT for free in 45 seconds?” asks Jason Haber, a licensed realtor and cofounder of the American Real Estate Association. “We’ve done virtual renderings for 20 years, so the fact that you can just do it now on AI, there was a whole cottage industry of virtual renderings and those people are now looking for a new job.”
In his position leading an association with over 22,000 members, Haber stresses that real estate professionals must disclose the use of AI, just like virtual staging has been disclosed in the past. Deceptive real estate practices can lead to fines and lawsuits, and the National Association of Realtors has advised realtors that the legal territory around using AI-generated images is still “murky.” The organization’s code of ethics prohibits using misleading images.
Haber said there are telltale signs of an “epidemic” of lazy AI use in the industry that outsiders may not be able to spot. Specifically, Haber says that ChatGPT “almost always” inserts the word “nestled” into copy it generates for real estate listings. The phrases “nestled in a prime location,” “nestled in the heart of the city,” and “nestled between two other homes” can be evidence of agents copying and pasting directly from the AI chatbot, he wrote on Instagram.
Haber says while he embraces technology, it’s not an excuse to “check your brain at the door.”
“If you become just a toll taker, you’re not a really good agent. You have no differentiation, you’re not creative, you’re just another agent,” Haber says.
Gupta, the cofounder of the AI real estate video generator, says as social media has become a primary avenue to reach consumers, high-quality videos that can hold their attention have become crucial. Using AutoReel could save “$500 to $1,000” and up to a week’s turnaround time from professional videographers, he claims.
“When we started this two years ago, we kind of got a no from customers,” Gupta says. “In 2024, they started saying ‘tell us more.’ And then this year, they’ve been asking How do I get started?”
But not everyone in the industry is entirely convinced. Nathan Cool, a real estate photographer who runs an educational YouTube channel with almost 100,000 subscribers, has experimented with AI tools, including AutoReel. He still feels like shooting vertical video is an easy and cheap add-on to his other services, although some productions can be more involved and costlier.
Plus, there’s the continuous issue of AI hallucinations. Gupta says that AutoReel is trained on millions of real property videos and has been fine-tuned to avoid inserting things that aren’t there—generating videos that zoom in versus rotate to try to stop the AI from creating features that aren’t really there. In a single test run with real photos from a real estate listing, this worked, although uploading the edited images from the listing Elizabeth found resulted in AutoReel adding in a fake couch.
But even when the AI products are realistic enough to fool consumers, Cool says that many everyday people are already sick of AI-generated videos in their social media feeds, and looking for a home raises the stakes.
“People that want to buy a house, they’re going to make the largest investment of their lifetime,” he said. “They don’t want to be fooled before they ever arrive.”