More Cases. Better Clients. Real Results.
Start a Conversation with GAVLOne of the questions we’ve received the most often over the past year from law firm clients and potential clients is regarding the usefulness of AI-generated written content, specifically for blogs and practice area pages. There’s no denying the fact that AI-generated content looks good on the surface, and no one can beat the speed it can crank out article after article. It reads well to outsiders. But how well does it rank?
One of the most dangerous things about AI-generated content is potentially the fact that if you ask the LLM about the content it generated, it’ll say it’s going to rank incredibly well. It’ll say it’ll bring in traffic. It’ll say it has all the right keywords and hits up areas that are often searched for. The LLM’s tone is confident. The tone of the piece it generates is confident. LLMs like ChatGPT are incredibly confident they have just the right answer on how to rank and how to gain the most viewers and traffic over time.
This confidence LLMs have is part of the reason many people are quick to think that AI-generated content may perform well ranking-wise, and why the general public is often quick to believe everything LLMs have to say. But we know better now. We know that hallucinations are a thing, and we know that LLMs do get things wrong. The only way to know whether or not AI-generated content can rank well, especially over time, is to put it to the test.
The Numbers Test
Search Engine Land, along with SE Ranking, ran a 16-month experiment. They launched 20 brand new domains with zero authority and published 100 AI-generated pieces of content on each site, every site covering a different area of interest. None of the content was edited by humans at all. No backlinks or internal links were added in. No extra SEO strategies were used. Within the first 36 days of being live, roughly 71% of the pages of content were indexed on Google. This means that Google saw them initially as valuable. 80% of the sites ranked for at least 100 keywords each.
Then, something happened between three and six months after the experiment was launched. Rankings began to tank. By the three-month mark roughly, only 3% of the pages remained in the top 100. Google still had the pages indexed, but users rarely saw them. By the 16-month mark, impressions and clicks remained low and showed no signs of recovery. The rankings had completely stagnated.
At GAVL, we have seen numbers and ranking drops like these ourselves when examining AI-generated content that new law firm clients had previously published on their websites. There may be an initial burst of impressions and traffic when the content is first launched because LLMs are pretty decent at determining keyword gaps and what information is most commonly searched for online, but that burst is just a singular, short-lived burst. It burns bright then burns itself to ashes.
Why? There are a handful of reasons.
Why Doesn’t AI-Generated Content Have Lasting Power?
AI-generated content by itself lacks several vital signals that Google uses to determine whether a piece of content is valuable to users and helps build trust. These vital signals include, but are not limited to:
- Content Uniqueness. One of the largest issues with AI-generated content is that LLMs are programmed to use what information already exists online. This means that any content it produces will mirror what exists already, and will not contain any unique insights, experiences, data, or information. Google looks to rank content that provides value as well as new, unique factors. Let’s use an example of medical malpractice. While an LLM may be capable of summarizing what other law firms are writing about medical malpractice cases, they are unable to provide unique insights about those cases that may help readers understand why medical malpractice cases are more complex than other types of personal injury cases, for example.
- Accuracy. While LLMs come across as incredibly factual, they can and do make mistakes. Google does expect information to be accurate in order to designate it as credible.
- Authority. Without building elements like backlinks, new domains will struggle to compete with other sites that are more established.
- Structure. Without internal links forming connections between various pages and any related content, Google fails to parse any easily-recognizable site structure or hierarchy. Topical organization is a massive part of SEO, and while LLMs understand these concepts, they can’t do the publication tasks for website teams. SEO is still very much a task better left to humans.
- Expertise. AI-generated content doesn’t include real-world expertise or experiences. When content is written by a human, you can include author information and credentials when publishing. These help build trust. Trust is a vital signal for Google, especially in the areas of health, finance, and law.
- Content Written by AI. Google understands how to identify AI-generated language by looking for established patterns, just as any AI-detection tool does. Since these patterns are so easy to spot, Google has started to devalue and even penalize websites that have content written by LLMs.
What Is AI Good For?
AI-generated content, similar to all tasks LLMs are capable of doing, is only as good as how it’s utilized. AI is a tool. Nothing more, nothing less. When AI is utilized as a tool to help writers organize, build better outlines, better implement keywords, suggest internal links, and build stronger headlines and subheaders, this is one way that law firms and content creation teams can utilize LLMs to help SEO efforts instead of get in the way of SEO completely.
For more information or to learn how agencies like GAVL help use AI to make our lives easier and not get in the way of Google rankings, reach out to us to learn more. We have your backs.
