#29: AI + Consumer Digital Behavior
Perhaps SEO is here to stay?
“Pay attention to what you pay attention to.”
Lately, that’s meant noticing trends that bubble up. All sorts of trends. From pickleball to cold plunges, and even which international destinations are trending with Americans, like Colombia and Portugal (purely anecdotal, based on my own feed and conversations).
I’m curious to see which trends will stay for good, and which trends will be old news soon enough. Which trends may be worth building a business around, and which ones will be gone by the time you’ve spun up an LLC for it.
Another trend I’ve been paying close attention to is how often I search for information on ChatGPT over Google. I’ve begun to rely more on ChatGPT for complex questions and Google when I’m in browse mode. Personally, Google works better when I’m unsure what I’m looking for. But that might just be me.
Search engine optimization (SEO) has been upended by the rise of large language models (LLMs). Today, you can ask ChatGPT a specific question and get a direct, tailored answer. It’s both efficient and effective, as you don’t have to sort through pages of blue links to maybe find what you’re looking for and receive a curated answer to your original question.
An entire new industry has popped up as a result, named answer engine optimization (AEO), attempting to decode the inner workings of an LLM. With AEO, you focus on how LLMs scrape your website, particularly which pieces of content are more frequently shown than others. Armed with this information, you can create content that shows up more frequently in LLM outputs. In doing so, you can better control how your brand is represented. As a consumer brand operator, I’m starting to shift my attention from SEO to AEO as I see an increasing number of website sessions referred by LLMs (mainly OpenAI and Perplexity).
But maybe I should pump the brakes. Check out this study by The Salt Agency.
The Salt Agency analyzed website traffic across many different industries and compared conversion rates between organic (think finding a website on Google that isn’t a sponsored link) and LLM-driven traffic (where a prompt output led you to the website). The study’s main finding is that although LLM traffic is increasing each month, LLM traffic’s quality (measured by key event conversion rate) tends to be worse than organic traffic.
Which is a big deal. This challenges the idea that LLMs’ specific, curated answers lead to better outcomes (however you define “better”) than a traditional Google search.
Before I go on, I must call out that The Salt Agency is a SEO agency, so they are certainly talking about their own book when concluding that organic traffic converts at a higher rate than LLM traffic. This point justifies that SEO on traditional search engines is still worth investing in, which is The Salt Agency’s business model.
Let’s look at a few of the data points.
The “Consumer E-Commerce” sector showed a near 700bps (24.12% vs. 17.56%) beat in key event conversion rate by organic traffic over LLM traffic. The study doesn’t specify what “key event conversion rate” means for each sector, but for e-commerce, I’d guess it refers to actions like adding an item to cart or completing a purchase.
I would have thought that providing context via an LLM prompt and receiving a curated answer would lead to a higher conversion rate than typing on a search engine and having to sort through various links. But perhaps not.
A few notes here. Perhaps LLM traffic is more pragmatic. Because users are expecting to receive the “correct” answer to their question, the moment they realize the website that the LLM provided them doesn’t have what they’re looking for, they leave. I’d love to look at bounce rates (when a customer spends under ten seconds on a website as per Google Analytics) for LLM traffic to help prove this hypothesis out.
Also, maybe the LLM traffic’s lower conversion rate is tied to the current state of the LLM shopping experience. Sure, ChatGPT and Perplexity have introduced shopping features, however the user experience is not nearly as robust as Google, particularly in the way of reviews, ratings, and shipping timelines. Although, I am impressed by Perplexity’s shopping recommendation engine that will provide context on a product’s best use case. Google does a better job at showcasing products in its shopping carousels, however its LLM foes are right behind.
On the flip side, the “Health” industry saw LLMs win on conversion rate versus organic (13.24% vs. 12.88%), albeit a small victory. I can personally attest to the efficiency of LLMs when it comes to ailments. Type your symptoms into Google and you’ll find yourself going down the rabbit hole that is WebMD or the Mayo Clinic. You start with a simple headache, and ten minutes later, you’re spiraling through worst-case scenarios.
LLMs allow you to type in the full context of your ailment, asking clarifying questions before they provide a few potential diagnoses. I recently had a knee injury and was blown away by how detailed ChatGPT was at asking for additional information on my symptoms and the “why” behind their guesses on what could be the root cause of my pain. It was a significantly better experience over Google. With that said, I am assuming that “Health” refers to health-related queries and questions. Unfortunately, sector definitions are not included in the study.
I find that LLMs are excellent at providing information about a topic. If you want to learn about all the causes of knee pain, ChatGPT will break them all down for you in a clear, concise way. But where LLMs seem to lag is for dynamic inquiries, like commerce. As mentioned above, the shopping experience is still early but improving quickly.
I’m looking forward to more studies coming out about this topic. LLMs are great, but for now, Google search still has a seat at the table.