Table of Contents
It’s super interesting to see that many of the sites that suffered greatly in Google’s helpful content update of September 2023 are able to rank really well in the new ChatGPT Search engine.
On one hand this is exciting. However, it’s not all good news. For every one of these queries impacted by Google’s HCU, I found that the answer a searcher was likely looking for was supplied by ChatGPT’s response.
Let’s look at some queries.
The sites shown here have given permission for me to share.
Keyword 1: cheese popcorn
This keyword has been in decline since the HCU. It showed some mild signs of life with the August core update but has not fully recovered.
If I put this keyword into ChatGPT, the site is chosen as the second source. Exciting, right? Except that if we put ourselves in the shoes of someone who has asked ChatGPT about cheese popcorn we can see that the tool has taken the sources and combined the best of each into an answer that makes it much less necessary to actually visit these websites.
What about Bing?
It’s interesting to note that this page only ranks in an image carousel for Bing. We know that Bing’s index is one of the sources ChatGPT draws upon. However, it’s important to know that Bing is just a part of a “set of services” that ChatGPT Search uses.
Keyword 2: how to make sour cream in 5 minutes
Once again we can see that a page that started its decline with HCU is featured quite highly on ChatGPT Search’s results.
But also, if I just quickly want to know how to make sour cream in 5 minutes, I don’t need to visit the website.
This page, although featured at #2 in the sources ranks #1 in Bing organically.
Keyword 3: plant island breeding chart
Same story here:
In this case we can see that the entire chart requested by the searcher is given by ChatGPT.
This page ranks #1 on Bing in a mutli-site featured snippet.
Keyword 4: dnd ability score improvement
And here is another similar example.
Although this past was surfaced 4th by ChatGPT Search, it ranks #10 organically on Bing for this keyword.
Why do Bing rankings not match ChatGPT Search’s rankings?
I asked ChatGPT about this. It told me that ChatGPT search uses a combination of pre-trained knowledge and real-time web access to generate responses. Its content selection isn’t solely dependent on traditional search engine rankings. Instead, it leverages a blend of sources, including those that may not rank highly on platforms like Google or Bing.
What can we learn from this?
This article brings up two major questions.
1) Is this fair?
If I owned these websites, I’d be quite upset that ChatGPT is able to profit from reading my content and paraphrasing it in a way that makes readers not have to visit websites.
OpenAI argues that training on your publicly available content is fair use with longstanding legal precedents. They provide an opt out for those who do not want to be included.
I do not think it’s fair at all. Even if the information on these websites is publicly available, it was made so under the current model where there is an understanding that search engines can use a part of our content in exchange for potentially sending us traffic which we can then make money from, or use in some way to help our businesses.
On the other hand, I believe that this is one of the main reasons why Google’s helpful content guidance was created. While the content listed above was technically “unique” they all contained information that is known by many, and not originating with these websites. For many years, we had an opportunity to create this type of content and profit from it. But now that AI can summarize information that is generally known in the world, there is less need to create content unless it is truly bringing something new to the online world.
The obvious argument here is that with little incentive to create content, what will AI draw upon? I think that if you are a business with an existing audience and you are creating people-first content to answer your audience's questions, there will continue to be incentive for you. For example, if this article you are reading now is used as a source by ChatGPT Search in the future, it does not cost me anything, but rather, it brings more attention to my brand and overall helps improve my business. I am quite happy to have LLMs train on my content. But then I don't make money from ad impressions and clicks.
2) Can and should we be optimizing for LLM search?
We are in a whole new era. At a Pubcon recently, I enjoyed listening to Wil Reynolds speak about the traffic his search agency Seer Interactive is receiving from LLMs. It seemed miniscule compared to his search traffic. However, he plotted the progression over several months and if the trend continues, within 20 months he will have more AI driven traffic than organic search traffic.
Optimizing for LLMs is much more than just optimizing for Bing. These tools are using the entirety of their training knowledge and augmenting it with Bing (and as stated above, possibly other services as well.)
I would encourage you to start tracking your client’s presence in ChatGPT, Gemini and Perplexity. I’ll be writing more as I develop processes for my own clients. We won’t be able to tell what queries (or more accurately conversations) drove visitors, but we can see which pages are getting attention.
We also can influence what is shown about our business by studying where it pulls information from. Most of the information below is taken from my website. That means I have some way to control the narrative.
Again, I’ll have much more to share both in articles and in my newsletter.
I hope this post has been helpful. You’ll find my AI related posts here on my new website, YouKnowAI.com. And as always you can contact me on mariehaynes.com, or in the Search Bar.