The biggest change to SEO since the launch of Google?
Do the same, only more
On May 10th Google announced “Search Labs”, a beta test of new search functionality. Initially there are three things to test:
Search Generative Experience
Code Tips
Add to Sheets
The latter two are interesting, but not of immediate importance to marketers. The first is Google’s first attempt to incorporate generative AI into it’s search results. And it could have significantly implications for the future of SEO.
You can get access to Search Labs by opening your Google App on your phone (who knew they even had a Google app?), and clicking on the test-tube in the upper left hand corner. I requested access and was given it within a day, so it is worth your time to open (or download) the app and request access in order to test with the new experience.
For those without access, what is the new experience?
It’s not one thing. Over the last few days I have seen a few different “experiences” and it seems to be changing.
Sometimes when I search I get no AI at all. Other times the normal search results appear immediately and the AI starts “generating…”. It takes a second or two (or up to ten one time I counted) to get the AI results, which, while fast, feels very very slow for everyone used to the milliseconds that it takes for traditional results to appear. Sometimes the “AI section” appears with text that reads “Use AI results for this query?”, followed by a button that reads “Generate”. For these searches Google is leaving it up to the user whether to get AI results on that particular search.
For the most commercial searches (“best hotels for families in Paris”) any AI generations are pushed below the fold by traditional search ads. If I want the AI response I need to scroll down, or click on a new tab “converse”, which takes me away from search results entirely and drops me in a traditional chatbot interface.
Apart from the most commercial searches I have not found any patterns on which type of result I get on any given type of search, but I am sure Google is crunching all of the beta users the data meticulously.
AI generations are both slow and expensive (in comparison to traditional search). And speed in these things traditionally matter a lot. From Google’s blog back in 2012:
Our research shows that if search results are slowed by even a fraction of a second, people search less (seriously: A 400ms delay leads to a 0.44 percent drop in search volume).
I expect, if anything, users expectations have gone up in the last decade.
The cost issue is also real. While the cost to generate AI responses is coming down, it is estimated that it is still 10x more expensive to generate an AI response vs a list of links. If ever Google result included an AI-response it would cost Google an additional ~$12B/year (last year Google made $60B in profits). This is a cost Google can handle, but it would be material and could definitely hurt the stock price if they were not expected to find a way to mitigate it.
As costly as this is it is no where near as impactful if Google loses even a few points of share to Bing due to a worse search experience. A one-percentage point drop in market share is worth ~$30B. So the quality of the search engine matters far far more than any costs, even if those costs are significant.
Hence the beta testing.
Google is learning when users want the AI-generations. They are learning the impact of the slower response on usage. They are learning which types of AI responses are valuable and which are not.
So what does this mean for SEO?
The first thing that happens when an AI response is generated is what doesn’t happen. As far as I can tell there is never a “snippet” that appears in the search results. Snippets were first introduced in 2014, and have increased in prominence over the years. Elliminating them entirely is a pretty drastic change, but AI is mostly solving the same function the snippet did so it does not make much sense to include both. For those of you who are getting traffic from appearing in snippets, expect that to go away as SEG rolls out.
The second thing that most AI responses produce is a carousel of what appear to be reference links. These reference links can usually be found in the below search results, but they are not in the same order. For example, consider this search on “new weight loss drugs”:
Three things to note:
Interesting that this does not seem to be a “commercial search”. No search ads appeared above the AI generation. Maybe due to advertising restrictions in pharma?
The generative results are taking up the entire screen, pushing all organic results below the fold
The AI response includes “reference links” in a carousel to the right of the generation
The first carousel result is from ABC news. The same link is found below in position #9. The second result on the carousel is from the Eli Lilly investor website. That link doesn’t appear below until position #57. When you click on the carousel links you are jumped to the relevant part of the page that the AI was referencing in the generated text (You can also click to expand the number carousel links. I expect, like clicking for a second page of search results, that will be rare)
Google seems to be optimizing the carousel links so that the user can double check the accuracy of the generative text — and however that is done, the result seems to be very DIFFERENT rankings to traditional search results.
So whatever companies were doing to rank in traditional results will have to be re-thought if the goal is to rank in the AI-carousel. And I expect more and more clicks are going to go to the carousel.
How does one rank in the AI-generated carousel? I expect even the designer of the search experience at Google doesn’t really know. It’s all black-box AI. It is going to take years of testing and observations before anyone figures out how to crack the new code. When someone does I am unsure what they should do to best monetize that insight (tell everyone and get to the top of Hacker news and use the credibility somehow? Keep it a secret within an SEO agency? Start a new business driven by SEO? Sell the secret back to Google in exchange for keeping quiet?).
Here are some initial hypothesis from my testing:
Stating Facts: If the sites are ways to double check the accuracy of the AI generations, then it makes sense that they will match the clearly articulated facts that the AI is generating. But why one site and not another? Maybe clearly written? Each fact stand-alone within a paragraph? Non-controversial?
Contracting Facts: Sometimes the AI likes to give you “both side of the story”. Where the first strategy is about stating (more?) clearly what everyone else is saying, the second strategy is to give the other side of the story. If there are less sites doing that, then you are presumably more likely to win due to the reduced competition.
Other ways of looking at it: Related to #2, for general searches, the AI needs to guess what your intent is and then given an answer. If the search is “hotels in Paris” you may get a spot if you cover things like “the average cost of hotels in Paris” or “What are the best neighborhoods to stay in Paris”, or “what are the best reviewed hotels in Paris”. Perhaps this just means that long tail SEO got even more important since these long tail pages may start appearing in head term results?
The answers don’t seem to be much different than traditional content long tail SEO, and maybe have the most in common with targeting snippets. Which is an argument for dramatically increasing content generation — since inbound links are likely LESS important than they have been in the past. Thankfully AI is very good at generating high quality content.
I will explore more tomorrow what this means for eCommerce, Local Search and what companies should be doing about it right now (Thursday posts are for premium subscribers)
Keep it simple,
Edward
When an AI response is generated it also includes
Create presentation out of Kevin’s stuff and take it to Warburg CMOs
Write post for Wednesday!
https://www.kevin-indig.com/x-realizations-from-testing-search-generative-experience/?ref=growth-memo-newsletter
Google’s new search:
Sorry Microsoft, but I think with this one you might be about to slip back into second place on the old AI search engine front… Google is opening up access to Search Labs, their home of all things experimentation for AI search. The big ticket in the Labs is the Search Generative Experience, which displays an AI-generative paragraph at the top of your search results.
The Verge argues that this could truly revolutionise SEO - what if companies start scrambling to appear in the AI section rather than the links? It’s still a waitlist, but time to jump on it!
Have to hit a button to generate. Faster. Cheaper. Tests engagement and which types of queries people interested in.
Also: Link Carousel - will it get the majority of organic clicks? Rank in that not ten blue links.
Different ranking!
Link Carousel is based on how the AI answers. “by explicitly writing about an angle AI Snapshots highlight, websites might increase the chance of ranking in the carousel.”
Example: Credit Cards, carousel talks about risks/benefits
eCommerce: Google is pulling from reviews to turn it into structured data based on query request - like Raveable
3rd party reviews
Google reviews
Price
Thumbnail
Description
Attributes
Title
