Both Google and Bing have dove head first into the world of generative AI by fully embracing these technologies to provide a richer search experience within their results. Both of which are transforming the result pages and search engines themselves into being more of what Marissa Mayer once called an “answer engine”. By providing generative text above the search results, that attempts to answer a user’s question or deliver relevant information to a user’s query instead of providing a list of links to websites.
This of course has alarmed many SEOs and publishers that are concerned that traffic that they once got from search engine result pages will be limited or gone completely as users will no longer find the need to visit websites when they can get all the information they requested straight from the search engine themselves.
While I understand how many are concerned about these developments, and after testing Google’s search and generative experience myself, it is a bit concerning. However, if we examine similar instances in the past where search engines have included the same types of results and also consider the realities of the search economy, we will see that the impact of these developments may not be as bad as many are assuming, and in some instances there maybe no change at all.
Now that is not to say that we will not see changes in some regard. Of course some websites will lose out on traffic and be forced to develop new content strategies to fill the gap, however others will not see a significant change at all. No matter what, there will still be a significant need for SEO as a working marketing channel to engage with the search engines in various ways.
SEO Should Already Be Dead, But Isn’t
One of the biggest reasons that I am confident that SEO will still remain a viable marketing channel in the future is that over the course of the last 14 years of my career, SEO has supposedly been on its deathbed dozens of times. But despite that it prevailed on as a viable marketing channel. That does not mean that it is completely invincible, but it does mean that we should be wary of any over speculation of it’s demise. Here are three examples of moments in the past when SEO should have died but continued on to be a viable marketing tactic.
Search Ads Should Have Killed SEO, But Hasn’t
Over the years Google has aggressively taken over more of the screen real estate above the fold with advertising opportunities. So much so that many result pages include nothing but advertising above the fold. Take a look at this screenshot as an example of every square inch of the result page utilized as an advertising opportunity, before scrolling down to the organic results. Of course this phenomenon is mostly seen for commercial queries where advertising rates are highest, however they are increasingly being seen more and more for long tail queries and even branded searches. This has been a steady increase over the years however has not stippled the demand for SEO. If anything, this increased reliance on advertising above the fold, has pushed more companies to invest even more into creative content strategies and technical SEO as a result of ranking within the top five organic positions even more valuable. So despite its appearances an increase in advertising has not killed SEO and in many cases only made the demand for SEO stronger.
Featured Snippets/Knowledge Graph Should Have Killed SEO, But Hasn’t
During my recent review of Google’s new Search Generative Experience (SGE) I mentioned that their generative AI results section is very similar to the Featured Snippets. In both the SGE section and Featured Snippets they typically take up the same amount of screen real estate, include around the same level of content depth, and provide not enough information for most users to be satisfied. The one striking difference is that SGE does not include citations for the content found. However, it does include a carousel of links that are somehow related.
If the similarities between Google’s SGE and Featured Snippets are accurate then, it is likely we will not see much of an impact from SGE as we did not see much of an impact from Featured Snippets. Of course there were definitely examples where websites that previously ranked number one for a popular search term did receive less traffic after a Featured Snippet was included on the page however pages that were cited within Feature Snippets received overwhelmingly increased amounts of traffic. So much so that many businesses sought to be mentioned as a Featured Snippet as part of their SEO strategy. This has spurred a whole new subtactic within SEO for optimizing for Featured Snippets.
I believe that Google’s SGE will result in the same behavior. Businesses will want to be included within SGE results and will seek out opportunities to optimize their content to do so. I’ll discuss that further down the page but for now I’m not concerned that SGE will impact SEO as much in part because Featured Snippets didn’t as well.
Zero Click Should Have Killed SEO, But Hasn’t
About 3 years ago there was reporting of an increase in zero click searches. A zero click search is when a search query is run without clicking any of the results or advertisements on the page. This likely is the result of an increase in things like Feature Snippets and Knowledge Graph integration that provides more information in the search results without having to click onto a website or advertisement.
While this is obviously concerning for publishers and SEOs that rely on traffic from search engines, in the end it appears that this increase in zero clicks has not resulted in a significant decrease in organic traffic. Many believe this is because at the same time that zero clicks increased, so did query volume as a whole. Meaning that more people are using search engines at the same time that more people are not clicking on search results. This is likely because elements such as Feature Snippets and Knowledge Graph have changed user behavior and query intent. Users have become trained to use search engines for more than browsing the internet and instead for using them to get answers or seek information in the same way that Marissa Mayer’s “answer engine” is supposed to work.
I have seen this play out in real time as friends or family members will pull out their phone and say “let’s see what Google says” and ask it a simple question like “how many ounces in a cup?” or “who is Martin Sheen’s wife?” In the past users would likely not have used a search engine in this way at all but now have been trained to do so because the information is available.
The increase in search volume for new queries like this has not impacted organic traffic significantly because both have increased at the same time.
Many (Most?) Queries Aren’t SGE Worthy
One of the lessons learned from Bing’s first roll out of their generative AI feature is to not put too much trust in the initial set of results. This is because of the overwhelming prevalence of hallucinations within Large Language Models. In other words many of the results found within Bing’s generative results were completely fabricated and not accurate at all. This resulted in not only incorrect results for users but also embarrassing moments with the media when they were trying to promote a new cutting edge technology.
Google seems to be more careful in not repeating this mistake by limiting the types of queries that it’s SGE’s feature will be used on. For example queries that are classified as Your Money Your Life will not include a SGE result. Other queries that may touch on ethnic or racial issues are also not included and it is likely that many queries that Google has never surfaced results for yet will not be included as well because they have not had enough time/data to process a reliable result for that query yet. This type of query alone could end up being 15% to 20% of all queries in itself.
Considering there will still be many types of queries that SGE will not be applied to means that there will still be plenty of opportunities for SEOs and content marketers to build robust content strategies to secure targeted organic traffic. This might mean engaging the customer at a different point in the marketing funnel or creating content that is adjacent to target keywords. Either way it is still possible to gain organic traffic utilizing queries that do not serve SGE or AI results.
Search Engines Will Calibrate
At the time of writing this article it’s been a little over a week that Google has rolled out its test of SGE to select users. Even in that small time frame, they’ve made changes to both the layout of the result page and also the frequency at which SGE results are displayed. It is likely they’ll continue to experiment with how SGE works based on user feedback during this initial beta test. But they will also make modifications and adjustments after launching to the public as well.
Suffice to say that what we know now about Google’s generative results is likely not what is eventually launched or set long term. It is likely that many of the features that appear to take traffic away from publishers will actually not exist at all or will be less prevalent than what is currently expected. One reason that this is likely is because of the likely impact of SGE on Google’s advertising revenue. In some cases it looks like ads are placed above the results and in other cases below. However any marketing professional that has experience with UX or CRO will tell you these types of changes will have significant impact on ad CTR. Therefore if Google’s primary concern is ad revenue it is likely they will not allow SGE to disrupt the status quo too much.
For SEOs it’s important in the short term to think of creative strategies to offset any traffic changes initially until the search results are less volatile and the search engines have found a comfortable place for SGE.
LLMs are Easy to Manipulate
When I first started playing with ChatGPT I gave some examples of something that I called Artificial Intelligence Optimization. Which is simply training a Large Language Model to accept information based on a series of prompts and responses in a logical format. This seems to work to various degrees. However my test then was limited to ChatGPT and not search engines. Now it seems a similar approach can be utilized on Google.
In this example Britney convinces Google’s SGE that Jennifer Lopez loves grilled cheese sandwiches. As you can see her process for implementing this level of manipulation is pretty straightforward and quite frankly a lot easier to execute than an entire content strategy or outreach link building campaign.
Others have theorized the opportunity to manipulate the systems from a more traditional SEO approach through content optimization and maybe even some old school SEO tactics. I think there are many untapped opportunities to optimize for SGE that are very similar to feature snippet optimization. Essentially when we approach how to do SEO for this type of results we need to focus on optimizing not web pages or URLs but specific pieces of information. Most likely this approach to optimization, will be similar to entity optimization or semantic SEO that considers entities, query intent, and language to promote an idea or entities of information within the context of other entities.
No matter what our strategies and tactics will look like in the future, it is obvious SEO as an industry, and the practice, can and will evolve to encompass manipulating Large Language Models in some way. It is what we have always done, evolved in an effort to help promote our clients within the search results and I don’t see that stopping anytime soon simply because of a new piece of technology.
SEO will continue to live and thrive, and I am going to thrive with it. Will you join me?
