You might go to an urgent care facility’s website to make an appointment or read an explanation of what causes fevers. But most people are probably not looking to urgent care websites for an explanation of what happens when unicorns consume ketamine.
But that—and millions of other pages about things that don’t make much sense—have suddenly been popping up on the website of a New York City-based urgent care clinic called Nao Medical. The company, which says it has 16 locations in New York City and Long Island, appears to be using AI to generate a vast flood of well-written and sensibly structured—but not particularly accurate—posts about popular topics in an effort to rank higher on Google.
One post on Nao Medical’s website explains a medical condition it calls “Derek Jeter Herpes Tree,” which is not actually a medical condition but that Nao says nevertheless is “a rare viral infection” that affects trees. There are posts promising that Ivermectin can help COVID-19 patients regain their sense of smell (not true), about getting iodine poisoning from eating too much shrimp (extremely unlikely), about a “cutting-edge technology” called the Cloud 5 Zinc Shell (not a real thing) and a “revolutionary healthcare solution” called Chicken Par Y (not a real thing either.) Google just about anything you can think of and “Nao Medical” and you will find a long list of posts, some of which appear to make some sense, others which don’t.
“Are you tired of the traditional colonoscopy procedure? Looking for a more enjoyable and engaging way to protect your health? Look no further than color guard!” one post reads, going on to explain that the color guard (yes, the people who wave flags in marching bands) can be a “fun and effective” alternative to colonoscopy procedures. (It appears to have gotten the color guard confused with Cologuard, an at-home colon cancer screening test.)
We were promised that AI would change the world for the better, freeing humans from meaningless tasks and putting a wealth of information at our fingertips. But now that AI is here, the first-hand negative consequences are also piling up. Sites like Nao Medical appear to be using AI to generate millions of pages of content that are not only useless, but also spreading or amplifying misinformation that exists in other parts of the Internet.
Read More: The AI Arms Race Is Changing Everything
The eerie thing about the Nao Medical posts is that they are not total nonsense. The article “Why Are Unicorns Healthier Than Dragons” makes a decent case, with bullet points and headers, as to why the former imaginary creature has superior health to the latter imaginary creature. (One explanation: “They often retreat to serene meadows and engage in meditation, allowing them to maintain a calm and balanced state of mind.”) The article extolling the “Surprising Health Benefits of Ginger Bread Shrek” does not describe why the post links the character Shrek to ginger bread, but does provide a recipe and the promise that “by enjoying ginger bread Shrek, you can not only satisfy your sweet tooth but also give your immune system a natural boost.”
According to Google algorithm and search engine optimization (SEO) experts, Nao Medical appears to be using AI to write millions and millions of posts, structuring them to have headings and keywords that people are using to search in Google.
“It’s clear to see that AI is being used in some way to falsely inflate this site’s organic visibility through mass content production,” says Nick Boyle, SEO director at the Audit Lab, a British digital marketing agency.
Google looks for what it calls E-E-A-T when it decides which websites show up first in organic search: Experience, Expertise, Authoritativeness, and Trust. Since Nao Medical is a legitimate group of medical centers with thousands of reviews on Google maps, Google perceives them as having experience and trust. By creating “copious amounts” of content on certain topics, Nao is appearing to try to convince Google that it also has authority on certain topics, says Marie Haynes, a Google algorithm expert with her own consulting firm who looked into Nao Medical’s web strategy for TIME.
Google could be ranking Nao’s pages higher because it can easily find plausible-sounding information in Nao’s many AI-generated posts, which link externally and which have bolded subheadings with useful titles like “Fish Oil: Pros and Cons” (in a post about which is a better health choice: fish oil or Nao Medical.)
Nao Medical had about 11,000 daily impressions until early June, Haynes says. Then the volume of pages on the site naomedical.com started to spike—between the first time she checked, on July 27, and the second time, July 31, the site went from about 15 million pages to 25 million. Now, the site has around 1.4 million daily impressions. “That doesn’t happen organically. It has to be black hat,” she says, referring to people who violate laws or ethical standards on the Internet.
Nao Medical’s president, Arihant Jain, responded to initial questions about Nao Medical and its business model, saying the clinic boomed during the pandemic and then has struggled as patient volume dropped and as insurers refuse to pay for services they had once covered. Jain did not respond to subsequent questions about Nao's internet posts.
After this article was first published, Nao Medical confirmed in a blog post it used AI to generate content and that it "slipped up on a few of these articles." It also advised people not to take medical advice from the internet.
After this article was first published, Nao Medical confirmed in a blog post it used AI to generate content and that it "slipped up on a few of these articles." It also advised people not to take medical advice from the Internet.
The company’s strategy is actually very clever, Haynes says. Nao appears to be addressing topics people are searching for but that don’t turn up a lot of organic results, making it easier for Nao to rise to the top.
Nao Medical ranks on the first page if you are Googling “caffeine and autism,” for example. This is probably because Nao has created a “topical hub” with around 22 published pages about caffeine and autism, making the site seem authoritative on the subject.. The AI appears to have structured the posts well, with headings addressing topics and subtopics that Google knows people look for, such as “can caffeine worsen anxiety in individuals with autism?”
Nao’s success in getting to the top of Google results also illuminates a problem with the search engine that has been circulating on the Internet: that Google search “is dying” because it is coming up with too many useless results. One reason: that there are too many people whose job it is to “game their way to the top of Google,” wrote engineer Dmitri Brereton in a viral post in Feb. 2022.
Google search has adapted to address new spam techniques and low-quality content for 25 years, and its goal is to show “helpful, relevant content that’s created for people, rather than to attain search engine rankings,” Danny Sullivan, public liaison for Google’s search business said in a statement provided to TIME. Google regularly launches updates to search to make sure it’s providing relevant information; in August 2022 it released an update that it said would lead to “more content by people, for people.” Using AI to generate content for the sole purpose of manipulating ranking results violates Google’s spam policy, the company says. But despite rising interest in generative AI, Google says, 99% of visits to websites from search have remained spam-free.
Read More: In the Rush to AI, We Can't Afford to Trust Big Tech
Nao’s strategy seems like it would have limited benefits, given that most of the people running across this content online probably do not live within the clinic’s service area. But Jain told me that after Nao’s in-person patient volumes dropped after the pandemic, the company decided to offer more telemedicine, psychiatry, and nutrition services. “Our plans are to grow outside of New York,” he told me; the company already offers telemedicine in a few other states.
Experts warned about AI flooding the internet with useless content but much of the AI-generated content popping up so far appears to be misleading or worse. And there’s something particularly disturbing about a health care company using AI to improve its results. Patients could assume from its post volumes that Nao is more of a medical authority than it actually is.
Haynes, the SEO consultant, says this type of SEO tactic is unlikely to work for long. For one thing, Google is currently implementing an AI content system that is built to detect “unhelpful” content, and, according to Haynes, many of Nao’s nonsense posts would likely fall into this category. So Google will catch up with Nao sooner or later.
That’s what happened to a product review website called Conch-House.com, which put up around 6,000 posts a day in 2022, gaining 6 million users by scraping content from Amazon reviews. The site brought in $800,000 in revenue in two months—until Google shut it down.
Besides, says Haynes, Google is constantly changing how search works. Generative AI could completely upend how Google processes and shows results; one experiment uses generative AI to improve results. That may mean it won’t make sense for websites to use AI to create content to answer questions because Google’s search will already be answering them—thanks, of course, to AI.
Update, Aug. 10: This story has been updated to include a link to a new blog post from Nao Medical confirming its use of AI to generate articles.
ncG1vNJzZmismaKyb6%2FOpmZva2BnhHJ8jqeYqGWdmrGqr8ClZKCnn5y5pnnAomY%3D