If you manage a website or work in digital marketing, surely you have come across AI (artificial intelligence) copywriters. The available tools are easy, cheap, and often pretty effective — but what does it do to your SEO? What impact, if any, will the use of AI copy and content have on your rankings?
Since Google is the gatekeeper of all things SEO, we should first look to them for guidance on whether using AI content is worthwhile, specifically for search ranking.
Google has been back and forth on its guidelines for AI-generated content and has made claims to have algorithms that can detect and penalize AI copy. Until recently, most SEOs recognized that using AI content explicitly for ranking would be a risky strategy, bordering on black hat SEO.
However, in response to ChatGPT and other LLMs (large language models), Google has softened its stance. Speaking on behalf of Google, John Mueller has said that so long as AI content is written for human benefit, it may not fall afoul of Google’s guidelines.
Subsequently, Bankrate and other prominent sites have ranked well with AI-generated content, although those rankings could be temporary while Google and other search engines catch up. Some SEO companies have reported that ranking gains were short-term, following the addition of AI-generated content. After an initial boost, much of the new value created by the AI content disappeared.
At Coalition, we’ve taken a negative stance towards AI content for SEO purposes. Why?
First, it’s helpful to understand how most of AI content generators work. By and large, they simply guess (using tons of sample data) what word should come next in a sentence, given a prompt. That means the answer is only as good as the underlying data. And around specific topics, data may be more sparse, resulting in less reliable or accurate outputs. Some experts have referred to AI content generators as exceptional “bullsh*t generators”.
Subsequently, we view the use of AI content as a liability for many of our clients. When an AI makes a claim about a service or product offered by businesses, and the business then publishes that content, they become liable. While fact-checking, editorial review, and other steps can help correct some errors, they’re unlikely to catch them all. And frankly, given how much time can be invested in a thorough review effort, why not begin with human-created content in the first place?
Second reason is the goal of creating something beneficial to humans that ranks well in search. Again, Google’s big goal is to ensure that search results are beneficial to a particular person in a particular moment with a particular question. When SEOs focus on creating a positive benefit to a search user, they algorithm proof their SEO.
Statistical odds as to what word comes next create fluff rather than real beneficial information for human beings. Relying on largely AI-produced content ensures that a website’s potential benefit to end users declines rather than improves.
Third, if LLMs rely on the internet to consume content to inform their next gen versions, and the internet becomes home for predominantly AI-generated content, the result will be an AI informing an AI what human content looks like. Over multiple generations of testing and release, this consistently produces extremist and inaccurate outputs from the AI.
With the popularity of ChatGPT and others, the volume of AI-generated content going online has likely spiked significantly, which poses an existential threat to the quality of AI-generated content. In turn, LLMs will become more focused on detecting content generated by their own tooling, or by others. The arms race towards AI content generation will have to be matched closely by AI content detection as a result.
In turn, human-generated content will increase in value, especially when it can more demonstrably show human involvement in the content. We, as an SEO company, are investing more time in showing the humanity behind our content in a way that can be measured by algorithms and AI as a result.
Finally, Google cannot manually review all AI content for whether it is beneficial for humans or not—that requires human input, and human labor in editorial review is expensive. That means Google will rely on automation. We’ll likely see rules issued by Google on how to identify AI-generated content from human-generated content. And even content that is claimed as being human-generated, will likely have to pass a series of sniff tests.
AI content not clearly identified as AI content is likely to inherit some kind of penalty since it will be easier to assign a malicious or problematic intent in its use. Google has modeled this type of penalization or devaluation in its handling of links. The big difference between links and AI content is that links don’t require unique, allowed access to publish on your web properties.
As a result of these four ideas, we explicitly ban the use of AI content by our copywriters for any content that could be used as part of a page we want to rank.
What AI Content is Good For
AI can help generate content concepts and ideas, retrieving data points, statistics, and even quotes that are relevant to the content being produced. That’s why it’s crucial to know how to prompt these responses. The better the prompts you enter, the better information and ideas the AI will be able to generate for you.
Humans Need to Verify
Never just take AI-generated response at face value. The tool acts as a shortcut and provides planning assistance, but is not the be-all-end-all solution. Even the best AI content generators can still deliver wrong information, which is why it’s crucial to fact-check. Tools like ChatGPT are impressive but in many cases wrong. Make sure everything the tool generates is accurate.
Test AI Detection Tools
To reiterate, we do not condone the use of AI-generated copy for your website, as Google’s stance on this type of content can fluctuate. However, whatever you publish on your site should pass an AI detection test, such as this OpenAI detector. You should only publish this on an indexed webpage if it passes to minimize the risk of future penalties if (or when) Google eventually cracks down on AI-produced content.
To be safe, start with pages of lesser importance and lower hierarchy in the site structure if you’re going to experiment with AI-generated content. Try it out on a short blog post. Now is the wrong time to rewrite your homepage using AI.
Need humans to work on your website? Get in touch with Coalition for a free review.