Insights
What I learned about AIO
Reports on AI Overviews (AIO) are popping up everywhere.
In an effort to make sense of the data and trends, I’ve downloaded and read at least 5 studies in the past 2 weeks.
What did I learn?
1. AIO adoption is still limited.
AIOs appear for a relatively low number of search queries.
In a new whitepaper from Dave Cousin and Authoritas, only 29.9% of the 10,004 keywords analyzed triggered an AIO.
When considering search volume, the impact is even smaller.
AIOs showed for only 11.5% of total searches analyzed, or 2,601,780 monthly searches out of 22,708,250.
It’s interesting that their smaller batch of keywords yielded similar results to those Tom Capper reported on.
His analysis of 46,000 keywords showed AI Overviews appear approximately 10.5% of the time. (I discussed this in my last newsletter).
2. AIO and Featured Snippets
Do you have a greater chance of appearing in AIO if you have a Featured Snippet?
The short answer is ‘yes’ but the statistics I saw varied wildly.
The Cousin/Authoritas results state that 44% of the search terms with AIOs also have a Featured Snippet, while Capper found 27%.
According to the whitepaper, sites with Featured Snippets have a 63.3% likelihood of appearing in AI-generated results.
This tells us 2 things:
- AIOs do not seem to decrease the chance of a Featured Snippet appearing.
- Both AIOs and Featured Snippets are more likely to show for search terms that can be answered or fulfilled in the SERPs in a useful way for users.
This reinforces the ongoing advice to maintain best practices for Featured Snippets by providing clear, expert and well-structured content where these snippets could appear.
3. Search intent is a major factor in AOI.
As expected, search intent seems to be the primary factor in the appearance of AIOs.
AI Overviews will show for low-volume keywords and appear mostly for ‘how to’ query types.
For example, 'problem solving' queries had AIOs 74% of the time, according to Cousin.
He also found that ‘specific question’ intents (non-commercial) were the most likely to trigger AIOs (69%).
It’s important to note that the numbers vary greatly by topic, sector, and industry.
The results you see for ‘what is titanium dioxide used for’ will differ greatly from asking a question like, ‘when is the best time to visit Paris.’
The user intent with the latter could be part of broader research to make decisions on getting to Paris, avoiding the crowds, and where to stay.
Research-related queries still only account for 3.3% of AIOs.
This indicates that Google prefers to use AIOs for search queries where it can provide a direct and concise answer.
If the topics you write about are not prevalent in AIOs, then keep your focus on longer-form content, image optimization, and other elements that cater to broader user information needs.
As stated in one report:
“Focus on E-E-A-T factors for your content and site overall, focus on areas where you have topical-authority and expertise and don’t go after visibility in AIOs that aren’t your site’s area of expertise.”
Interacting with ChatGPT
This chart jumped out at me from Semrush's analysis of 80 million lines of clickstream data.
What does it tell you?
It tells me that search intent on ChatGPT is still not well understood.
People search differently on ChatGPT Search, so we need to think differently about their intent.
The whitepaper I cited above is relevant here, because it attempted to better classify user intent.
Cousins took the traditional search intent categories (navigational, informational, commercial, and transactional) and added 9 sub-intents to describe the full spectrum of user needs and behaviors.
If only 30% of searches on the chatbot are traditional queries, this suggests users are asking questions in new ways (conversational) and using ChatGPT for exploring ideas and solving problems.
The question is, how do we adapt as SEOs and marketers to gain traffic?
You’ve seen some of my opinions on this in recent newsletters, so I decided to take a different approach today.
I asked ChatGPT what to do!
This is not the full response and I’ve edited the answers for length:
- Optimize for “Retrievability”
Traditional SEO metrics now need an extra dimension. Ensure that your key brand information is available in formats that LLMs can easily parse.
- Produce High-Quality, In-Depth Content
- Invest in long-form content: Guides, case studies, white papers, and technical documentation that answer complex questions.
- Focus on expertise and authority [E-E-A-T]: This is especially important for sectors like education and technology.
3. Tailor Content for Conversational Queries
Consider:
- FAQ sections that answer common questions in a natural, conversational style.
- Long-tail keyword integration…mirror how users interact with conversational AI.
If you’d like to see the full list of ideas, reply and let me know. I’ll drop ChatGPT’s entire reply into a PDF and share it with you.
Remember this guy?
I mentioned Mike Hardaker's website woes 3 months ago—right after he attended the Google web creator event.
Mark founded the Mountain Weekly News, but after losing 97% of his search traffic, I thought he'd give up.
Nope. He's still posting new content on his website and on X.
You have to give him credit for keeping his sense of humor intact.
Do you ski?
Related: Lily Ray takes a reflective look at the fallout from the 2023 Helpful Content Update.
I hope these insights help. For more tips, follow me here.
