While honest websites experienced the ups and downs of Google’s March 2024 update, SEOs and recipe bloggers noticed AI spam sites surging to the top of search results. One example as of yesterday ranked for over 217K queries, 14.9K of which rank in the top 10 – and that number has increased a day later. This is what’s going on and how the spammers continue to beat Google.
Surge In March 2024
The site that’s ranking is a subdomain. The main domain has been around since the summer of 2020. The spammy subdomain was first spotted by the Internet Archive on November 30 2022, coincidentally the launch **** of ChatGPT. The subdomain was in a half-finished and essentially dormant until March 2024 when it rapidly expanded and immediately began to rank for thousands of search queries.
Wednesday March 20th the site was ranking 14.9k search queries in the top 10. Thursday March 21st the site had 15.6K in the top 10. Even though Google just concluded their spam update, this particular site (and others like it) continue to rank for thousands of search queries and Google appears powerless to stop them.
Food Writer Reacts To AI Spam
A food writer and recipe book author, Robin Donovan (Instagram), called my attention to the AI site, telling me that others in a private Facebook group were livid about AI sites surging for recipe search queries.
It’s super obvious that the content is AI generated, even the images accompanying the articles are 100% AI. So it’s especially hurtful to those with experience, expertise and authoritativeness to see obviously AI content outrank them.
Robin was understandably upset:
“How on earth is this the best content? And meanwhile, bloggers who are professionally trained chefs, recipe developers, cookbook authors, and others with decades of training and experience are watching their sites be decimated with these updates. Sites that they’ve spent years building with well-researched, well-written (human-written!) articles, recipes that have been professionally developed and carefully tested, photographs they have spent hours prepping for and shooting.
They’ve done all the things Google has told them to do for years—write your own content, take your own photos, develop unique, high-quality recipes, be an expert in your subject area and have credentials to show it, don’t try to game the system, be genuine, create HELPFUL content. For what?”
Details About The Spam Sites Look Like
1. Hosted On Squarespace
The spam site is on a subdomain and both the subdomain and the main site are hosted on Squarespace. Why Squarespace? Just a guess but maybe that infrastructure tends to generally appear legit to Google (or it might not play a role).
2. All Images Are Colorful And Simple
All of the images are AI generated, created in a flat colorful style similar to what might be seen in an infographic, indicating that a templated prompt was used to create the thousands of images accompanying the articles.
The images are an important component of the articles. Each article features about seven images that relate to the overall topic of the article. Every article is very colorful. I don’t know if the images are intentionally colorful but the use of strong colors help images stand out in the SERPs and Google Discover when Google displays them.
I uploaded one of the images to ChatGPT and asked it to generate a prompt based on the image to create a new one in the same style.
Here’s an example of an AI generated image in the same style as the AI spam sites:
Example Of An AI Generated Image Used By A Spam Site
3. All Articles Follow A Rigid Template
The articles follow a templated structure, which varies depending on the type of article. Article topics range from comparisons, local destination travel, lifestyle, recipes, health benefits of certain foods and so on.
The health related articles follow this exact article structure:
a. Introduction Introduces the concept topic being discussed and its relevance to a specific health related topic. The concept topic can be a particular food, a type of diet and so on. b. Health Benefits Following the introduction, every discusses the health benefits associated with the concept topic. c. Fundamentals This section discusses the basics of the health topic that is being focused on (oral health, diabetes...), the importance of nutrition, and common problems or diseases associated with the health topic. d. Nutritional Guidance And Key Nutrients e. Dietary Choices And Impact On Health f. Tips and Lifestyle Advice g. Conclusion - A summary of the benefits
4. Underlying Prompt For AI-Generated Articles
I was curious about what a prompt that generates that content would look like so I asked ChatGPT to create one.
This is a generalized prompt that could have been used to create the health and diet articles:
"Write an article exploring the connection between [concept of dietary choice] and [specific health focus], following the template below: Introduction: Begin by introducing the concept of [concept of dietary choice], its definition, and why it's relevant to [specific health focus]. Discuss its growing popularity and how it aligns with contemporary health and lifestyle trends. Health Benefits: Elaborate on the general health benefits associated with [concept of dietary choice], focusing on its potential to enhance [specific health focus]. Fundamentals of [Health Topic]: Provide a background on the [specific health focus], including essential anatomy, the significance of nutrition, and prevalent conditions affecting this aspect of health. Key Nutrients and Their Impact: Detail the crucial nutrients that play a significant role in [specific health focus], including their sources and the health benefits they offer. Highlight the importance of certain vitamins, minerals, and other compounds. Influence of Dietary Choices: Analyze how specific dietary choices influenced by [concept of dietary choice] can impact [specific health focus], positively or negatively. Recommend beneficial foods and advise against certain types that may harm [specific health focus]. Practical Advice for Diet and Lifestyle: Offer practical suggestions for integrating beneficial foods into one's diet and making lifestyle adjustments to support [specific health focus]. This might include tips on meal preparation, portion control, and balancing different types of foods. Conclusion: Conclude the article by summarizing how adopting [concept of dietary choice] can contribute to improving [specific health focus]. Emphasize the balance and variety of nutrients this approach provides and its potential benefits beyond [specific health focus]. Ensure the article provides a comprehensive overview that is both informative and engaging, catering to readers interested in understanding the relationship between [concept of dietary choice] and [specific health focus]."
5. Content Tested By AI Checker Tools
The weird thing about these articles is that every article I tested with GPTZero AI Content Detector scored 100% as AI-generated. The Originality.AI content checker offered similar scores.
Screenshot Of GPTZero Score
Screenshot Of Originality.AI Score
The Squarespace templates are professional and the articles themselves are in a dry style that is informative but lacks signals of human authorship such as expressions of insight or experience and a complete absence of colloquialisms. Every article tested failed the AI detection tests.
6. How Do AI Spam Sites Rank?
It’s my hypothesis that the reason these spam sites rank is that they’re taking advantage of a loophole in Google’s algorithms that allows new content to receive an initial boost, what Google’s John Mueller has described as Google testing the website or the webpages out. This happens all the time and excites people when they publish a new site and see it ranking almost right away.
What is happening with this one AI generated website is that it is publishing massive amounts of webpages every day and those pages receive a boost to the top of the search engine results pages (SERPs) for the first 24 to 48 hours. They then begin to slip down the top ten and eventually into the second pages of the SERPs.But by that time there are new pages begining that journey from the initial boost, every day. This is a classic old school strategy known as churn and burn.
John Mueller has commented in the past about why Google ranks new websites at the top of search results.
He explained:
“In particular, with completely new websites, one of the difficulties that we have is we might not have a lot of signals for those websites so we have to make estimates.
And depending on how we make estimates, it can sometimes mean that in the beginning we show this website a little bit more visibly than like it turns out that the signals tell us in the end.
…But that can go both ways. It can go in the direction of like you’re shown very visibly in the beginning.
And it can also be that maybe you’re shown less visibly in the beginning and as we understand your website and how it fits in with the rest of the web then we can kind of adjust that.
…Sometimes it’s also new websites that show up that we try to pick up really quickly.”
7. AI Spam Is A Longstanding Problem With Google
Google has had a long problem with AI generated sites dominating certain search results and this is not the first time that Google has been overwhelmed by spam, particularly for relatively longtail phrases. This one spam site is not alone and is not an outliers. There are many others just like it following the same methods for ranking.
What makes this example noteworthy is that it went live at the same time that Google launched a spam update and it continued ranking at the top of the SERPs for hundreds of thousands of search queries (with 15.6K queries currently in the top 10).
The AI spam site has now popped out on the other side of the spam update and is thumbing it’s nose at Google. It is a humiliating and demoralizing experience for the thousands of honest and experienced bloggers who are outranked by content that lacks credibility, experience and authoritativeness- these articles don’t even list authors.
Featured Image by Shutterstock/ViDI Studio
Source link : Searchenginejournal.com
