How Horoscope.com Uses 10,000 AI Articles to Rank in Google Search Results
As artificial intelligence technology rapidly advances, many publishers are beginning to experiment with using generative AI models to assist in content creation. One such company leading the charge is Ingenio, the parent company of popular astrology and horoscope websites like Horoscope.com, Astrology.com, and SunSigns.com.
According to a recent article in Digiday, Ingenio has published over 10,000 articles that were created using GPT-3, one of the most powerful language models developed by OpenAI. This has allowed them to drastically scale up content production while keeping costs down.
Let‘s take a closer look at exactly how Horoscope.com and its sister sites are leveraging AI to fuel their content strategy, how the AI-generated articles are performing in search engines, and what we can learn from their approach.
Integrating GPT-3 to Generate 90% of New Article Content
In an interview with Digiday, Josh Jaffe, president of media at Ingenio, revealed that the company began experimenting with AI-assisted content creation back in 2021. They built a custom integration between GPT-3 and their content management system.
This allows their writers and editors to automatically generate a draft of an article that is around 90% complete just by entering a topic or title. The remaining 10% involves light editing, fact-checking, and optimization by a human before publishing.
Using this AI-powered process, a single piece of content now costs a tiny fraction of what a fully human-written article used to cost to produce. The efficiency gains have been massive.
"We can publish faster than ever before," said Jaffe. "We can publish [for] less expensive than ever before. We can publish 1,000 articles for the cost of what one article used to cost to produce."
Launching New Sites and Scaling Content with AI
With this new AI content generation capability, Ingenio has been able to quickly expand into new topic areas. For example, they launched a brand new website called DreamsCloud.com focused entirely on dream analysis and interpretation.
This site now has over 700 articles about different dreams and their meanings, nearly all of which were generated by AI. Without this technology, creating a comprehensive dream interpretation resource would have taken years and cost hundreds of thousands of dollars in writer fees. Instead, they built it in a matter of months.
In addition to new verticals, Ingenio also used GPT-3 to breathe new life into some of their existing properties. They published over 10,000 AI-generated astrology and horoscope articles on SunSigns.com as well as 1,000+ articles on their other core sites like Horoscope.com and Astrology.com.
Celebrity profiles have been an especially fruitful area. "We wouldn‘t have been able to write up 10,000 celebrity profiles without generative AI," noted Jaffe. For any given celeb, they can instantly generate an article about that person‘s astrological sign, natal chart, personality analysis, predictions for the year ahead based on planetary transits, and so on.
Of course, astrology is a very niche subject matter that lends itself well to this type of scaled content generation. The "logic" of astrological analysis based on star signs, planetary positions, aspects between celestial bodies, etc. can be codified into an AI system much more readily than other more open-ended topics.
But the key takeaway is that AI allows publishers to produce huge amounts of niche content quickly and affordably in a way that would have been unthinkable just a few years ago. The implications for SEO and organic traffic acquisition are significant.
AI-Generated Articles Indexed and Ranking in Google
Now for the million dollar question – how are all of these AI-generated articles actually performing in search engines? Can synthetic content really drive organic traffic as effectively as human-written content?
The short answer is yes, absolutely. We looked at dozens of examples of AI-generated articles on Horoscope.com ranking on the first page of Google for their target keywords. Google is clearly indexing this content and rewarding it with prime SERP real estate.
For instance, let‘s look at an article on Horoscope.com covering Alicia Keys‘ 42nd birthday. It‘s titled "Alicia Keys Turns 42: What Her Astrological Profile Reveals About Her Future." This article has a byline of simply "by Horoscope.com" and when analyzed by an AI content detection tool, it is flagged as nearly 100% likely to be AI-generated.
Yet when we plug the URL into an SEO tool like Ahrefs, we can see that Google has indexed the article and it ranks for several relevant keywords, including:
- "alicia keys zodiac sign" – Ranks #10
- "alicia keys sign" – Ranks #1
- "what sign is alicia keys" – Ranks #6
Keep in mind that this is just one article out of thousands that Horoscope.com has published. While not every AI-generated post is a home run, clearly many of them are performing quite well in terms of driving search traffic.
This proves that Google does NOT penalize content simply for being AI-generated. As long as the content is relevant, provides value, and comes from an authoritative domain, it has just as much of a chance of ranking as human-written content.
Limitations of AI-Generated Content
With all of that said, we did notice some limitations and areas for improvement when it comes to Horoscope.com‘s AI-generated content. Namely around content quality and depth.
While the articles are certainly readable and relevant to the topic, they have a certain "mechanical" quality to them. The language tends to be somewhat repetitive and formulaic. They cover the basics but lack the deeper insights, personal anecdotes, and original research that the best human-written content has.
In fairness, this is simply a limitation of current language models. GPT-3 is incredible at mimicking patterns of human language but struggles with truly original ideation or analysis. The generated text is rarely factually wrong but often lacks substance.
However, this will almost certainly improve over time as the AI models get more sophisticated. And there are already ways to compensate for these shortcomings.
AI-Generated Content Works Better With Human Editing
Perhaps the most effective approach is to view AI-generated content as a starting point rather than a finished product. Have the AI do 80-90% of the writing, then have a human editor review, fact-check, optimize and enhance it before hitting publish.
For example, an editor could punch up the article with a more engaging intro, add in some interesting stats and examples to support the key points, include relevant imagery or multimedia embeds, and make sure it‘s well-optimized for target keywords and user intent.
There are even AI-powered SEO tools like SEO.ai to assist with this optimization process, giving editors specific suggestions on keywords, headings, links, and other ranking factors to include.
This human + AI hybrid model seems to be the best of both worlds. The AI can handle a lot of the "grunt work" of researching a topic and generating a comprehensive first draft. Then an experienced editor can efficiently turn that draft into a truly high-quality, publishable piece of content.
In the case of Horoscope.com, investing just 10-15 minutes optimizing each article would almost certainly pay off in terms of significantly better rankings, more organic traffic, and improved engagement metrics.
The Future of AI-Generated Content
While the quality of purely AI-generated content may be hit-or-miss today, it‘s only going to get better over time as the language models advance. Already, GPT-4 can generate much more coherent and insightful text compared to GPT-3. And who knows what GPT-5 and beyond will be capable of.
At a certain point, the content produced by AI will be virtually indistinguishable from the best human-written content. At that stage, the role of human writers and editors will shift towards prompt engineering, fact-checking, and optimization rather than original drafting.
AI will enable individual bloggers and small editorial teams to compete with much larger publishers in terms of content quantity and quality. For example, an astrology blogger working alone could conceivably publish hundreds of in-depth, well-researched articles per month with the help of AI tools.
In the meantime, the most effective publishers will be those who combine the speed and cost efficiency of AI-generated content with the editorial oversight and optimization savvy of human experts. The likes of Horoscope.com are giving us a sneak peek of this hybrid future.
It will be an adjustment for many content creators and there will undoubtedly be disruption in the industry as the technology matures. But in the end, AI will be a net positive for publishers by enabling them to produce better content faster and cheaper than ever before. The trailblazers who embrace this shift will reap the benefits.
