The future of SEO-led content as a result of the introduction of Generative Artificial Intelligence (GAI) in search results is a hot topic as of late, and for good reason. The impact for publishers and website owners — we’ll call them content creators from here on — could be devastating.
Google’s new Search Generative Experience (SGE), pictured below, integrates GAI directly into the search results, in the form of in-depth answers and the added ability to ask follow-up questions directly in search results. The threat to content creators is that the searcher may rarely need to leave search results.
What’s more is, SGE answers are masticated from unknown sources and regurgitated with mixed attribution to the original sources, those who rely on the referral traffic to meet business objectives and keep the lights on. Without those clicks, why should content creators even produce the content?
We’ll attempt to answer this question as well as address the impact on how we interact with search results and how content creators should react. We’ll conduct a test to help us better understand Google’s SGE results and the implications for SEO content strategy.
Is SEO dead?
I’ve worked in SEO for over 13 years and I remember “SEO is Dead” reaching marketing-wide industry headlines at least three times prior. It’s fair to say, this integration of GAI in search results will absolutely rock the foundation of SEO-led content marketing and forever change the way searchers interact with search results, however, old truths will remain true.
Well that’s a relief. Before we investigate how SEO will be impacted by GAI and SGE, let’s explore the history of AI and SEO to help guide us to this answer.
AI & SEO: How we got here
AI’s integration into organic search has generally led to improved search results. Google first introduced AI in search when they introduced RankBrain in 2015, their machine-learning AI system, to help process search results and power part of their ranking algorithm.
After that, they employed Neural Matching in 2018, a way to better associate words to concepts. That same year they launched SpamBrain, their AI-based spam-prevention system to detect and remove useless or harmful content from search results.
Next, there was BERT in 2019, whose aim was to better understand how using different combinations of words can convey various meanings and intents. Then came MUM in 2021, the 1000x more powerful technology add-on that uses AI and natural language processing to do more than all the others combined before.
And finally, just a few months ago, Google revealed the Search Generative Experience (SGE), a new search engine powered by their new AI technology.
Important to note, SGE is not publicly available and currently a Google Labs search experiment, unless you sign up for Search Labs here, which — at the time of writing — it’s only available in the USA.
Gaming the search results was harder than ever and we now must rely on the quality of our content to meet E-E-A-T standards in order to earn visibility and traffic. What’s E-E-A-T (as it’s now known) you ask?
Experience – the firsthand or life experience the content creator has with the topic.
Expertise – the content creator's proficiency, skill, and knowledge on the topic.
Authoritativeness – the reputation the content creator has as a reliable source with the topic.
Trustworthiness – the safety, reliability, and accuracy of the content creator’s page.
What Google has been telling us for years — produce great content and you will be rewarded — was finally ringing true. And then SGE arrived.
The real issue: Attribution & traffic
Predating AI in search results, the introduction of the knowledge panel in 2012 was the first instance where attribution raised flags for content creators. The familiar panel served to the right of search results, had content creators questioning the sources and attribution of this data, with many believing their introduction led to fewer clicks to content creators.
Fuel was added to the fire with the introduction of “rich results” in 2016. While attribution was less of an issue with rich results since directly below it was a link to the source of the extracted text, their introduction resulted in what’s known as “zero-click searches.”
Popularized by Rand Fishkin in 2019, zero-click searches meant that searchers got the information they were looking for in the search results thanks to rich results and didn’t need to click through to the result, thus harming content creator’s referring traffic and resulting revenue.
Let’s look at this search result for “who invented pizza.” Content gets pulled directly into the search results from the source, but they, of course, get the direct attribution with a link below the extracted text in the rich result:
According to research conducted by SimilarWeb and Fishkin’s Sparktoro in 2020, two thirds of all searches end without a click, meaning only 33% of searches result in searchers clicking through to content creators. The rest get what they need directly from Google’s enhanced search results. This was becoming a major issue for content creators, and fast.
Now we have a very similar problem — greater zero-click searches — but attribution is considerably worse.
Initial attribution with SGE was super vague and indirectly attributed by three thumbnails in a carousel of sources located in the upper right corner of the SGE result, if they appear at all. This attribution (or lack thereof direct attribution) caused uproar in the industry, and pressure since mounted on Google to begin testing better attribution directly in SGE.
Beyond attribution in search results, content creators are questioning their content’s inclusion in LLMs without consent. Pierre Far (former Googler) stated, “Why should a website allow its content to be used to train a language model if it doesn’t get anything in return?”
The biggest to opt out so far is The New York Times, that updated their terms of service to explicitly prohibit the use of their content to train AI systems. Many hope, myself included, that this will help pressure LLM chatbots and SGE to adopt a direct attribution model.
The introduction of the ability to block GPTBot — ChatGPT’s web crawler that consumes content to use in its GAI responses — is a welcome change, though the real challenge is with Google’s SGE to adopt a more direct attribution model.
It’s more than likely that searchers interacting with the SGE experience will click through less than they would with traditional search results, or even rich results. Only time (and data) will tell if the zero-click searches will go north of two-thirds, but it’s safe to assume it will.
What does SGE mean for SEO content?
An SEO-led content strategy begins with a search, meaning that searchers will land on this content via conducting an organic search. With SGE integration imminent, how do we best approach SEO content strategy?
Know when SGE is most likely to be triggered
Analyzing “search intent” — the primary goal of a user’s query in a search engine — can help us decipher what stage of the funnel the user’s query is at, and thus will help give us more insight into which keywords are triggering SGE.
The three types of funnel stages are top of funnel (ToFu), middle of funnel (MoFu), or bottom of funnel (BoFu):
To use some keyword examples, let’s pretend I have a website all about pizza that sells pizza ovens and accessories where I create content aimed at pizza aficionados and people interested in making pizza at home.
Let’s ask both ChatGPT and Claude 2 for a list of keywords clustered by funnel stage, providing the above information to each LLM chatbot. Then let’s see which keywords generate an SGE result.
Keep in mind that I don’t necessarily agree with the funnel clustering decision each chatbot made, however keyword funnel clustering is somewhat subjective and significantly nuanced based on your specific website’s offering and needs.
*No, but the option to generate a SGE result was given directly at the top of the search results, like below:
Reviewing the lists above, Claude 2 gets an immediate point for following the instructions better and giving me actual keywords, which I explicitly asked for, whereas ChatGPT gave me topics instead.
As for the clustering, I think they both did pretty well, but at a push I would give a slight edge to Claude 2. Full disclaimer, I also prefer Claude 2 for my day-to-day use. Have a chat with Claude 2 if you’ve never tried it.
Key Findings from the SGE Test
The test results first and foremost reveal just how often an SGE result is triggered, which at 70% of the time for all keywords in our test is eye-opening and concerning. I’m hopeful that since SGE is still in a testing phase, that perhaps they are testing triggering SGE more often than not, however, only time will tell.
1. The mixed top of the funnel
At the top of the funnel, we assume that searches are more likely to interact with SGE since they are searching more broadly about a topic, and thus Google might trigger SGE more often for ToFu keywords.
In our test results, there’s a big discrepancy between ChatGPT and Claude 2 results. ChatGPT triggered SGE results for ToFu keywords 78% of the time, whereas Claude 2 only triggered SGE results a third of the time. This is likely based more on ChatGPT-returned topics rather than keywords, despite my prompt for keywords. However, if you take most of ChatGPT’s ToFu keywords, it’s clear there are broad keywords behind these topics.
As for Claude 2, ToFu keywords are only triggering SGE results a third of the time, which is great news for SEO-led content pillar and hub & spoke approaches. ToFu keywords are typically broad informational queries and root keywords, which often are your target keyword for your “pillar page” in a content pillar strategy or the “hub” in your hub-and-spoke approach.
While many ToFu keywords trigger SGE results, there are great candidates for content pillars, such as “pizza sauce” or “pizza dough” for my pizza-making website. All ToFu and MoFu keywords for “pizza sauce” and “pizza dough” in both chatbot lists do not trigger SGE results. “Recipe” keywords also didn’t trigger SGE results.
Since search behavior for ToFu keywords is exploratory in nature, we know searchers may be drawn more to the SGE. Finding a pillar where multiple levels of the funnel do not trigger SGE results like these will be an SEO-content goldmine.
2. The more promising middle
While middle of funnel (MoFu) keywords on average were the next best option for SGE-less results, we might not need to worry about that. Searchers in this phase of the funnel will likely place more importance and value in the source of the information for these types of searches.
With ToFu keywords, the intent is broad and informational, so the searcher is likely less informed on the topic and they may also be less familiar with the sources in the search results. This is where cutting through the noise of an SGE-less result is important. With MoFu keywords, the user is likely more familiar with the topic and thus also familiar with content creators in the search results.
For example, learning what a “pizza peel” (ToFu keyword) is feels like an answer that you can get from just about any source and feel comfortable with the answer, particularly an SGE result. It’s common knowledge; GAI can’t get that wrong, right? Whereas with a search for “best pizza dough recipe” (MoFu keyword), the user is more likely to want their answer from a reputable source such as BBC Good Food or perhaps a familiar pizza oven company they’ve come across already.
Folks are less likely to trust a SGE result since the attribution is less clear. While this search result might include SGE, searchers may find themselves scrolling right past GAI to look for reputable or familiar domain names in the search results.
This is particularly true for long-form SEO content, e.g., in-depth tutorials or how-tos typical of MoFu content. These can’t be served directly in search results and — like most users — I prefer to go to sites I trust, or that appear trustworthy and authoritative when looking for an in-depth guide or analysis. With SGE, it’s currently challenging to discern the source, meaning we have to put a lot of faith in Google's judgment. And whether it's a tutorial on fixing your car or your website, you will likely want to ensure you're using a reputable source.
3. The busy bottom of the funnel
BoFu is an entirely different beast altogether. At this stage of the funnel, these searchers are ready to purchase. SGE triggers for nearly all of our test results, and where it doesn’t, you are offered the ability to trigger a GAI result.
I suspect that we’ll feel more of the pinch here in terms of SGE dominating search results since BoFu keywords are more financially valuable to Google than other stages of the funnel.
Knowing the nuances in the search results will be paramount for BoFu keywords. For instance, our tests revealed a “buy” keyword without SGE would trigger a shopping carousel right at top followed by local results. However with SGE, the local results take precedence:
Listing your products in Google Shopping will still be critical, however earning placements in buyer guides and product roundups (see upper right in the SGE result above) will be just as important. Additionally, many BoFu keywords served up links to products under a “Here are some products to consider” heading, which were no doubt sourced by LLMs from buyer guides and product roundups.
Being intimately familiar with these new SGE results will be critical for this immensely valuable phase in the buyer journey.
Keep building great content
Content creators shouldn’t fear SGE and should instead focus on building great content that exhibits expertise and authority at all phases of the buyer journey.
Don't feel threatened by SGE
Let’s say you have a highly valuable page for “pizza stones” (ToFu keyword) which drives a lot of traffic and you’ve built supporting content linking to it — a true content pillar — and now there’s an SGE result for it. Should we kill the page?
Absolutely not. Your page may already be attributed in the SGE result, which will result in some referral traffic, albeit less than before. No reason to lose that, especially as Google’s attribution may improve.
Additionally, while SGE will have significant impact on ToFu content and likely a heavy impact on MoFu content, many searchers will still prefer to get the content directly from a reputable source — the horse’s mouth — and not want the information to be masticated and regurgitated by GAI, which is widely known to be prone to errors and inaccuracies.
Wired recently described SGE as “more nuisance than aide. It’s slow, ineffective, verbose, and cluttered — more artificial interference than intelligence.”
Depending on their search intent or the phase in their journey, searchers may scroll right past the SGE result from Google. Searchers may also skim through the SGE result, but find themselves drawn to the cited sources for a more in-depth and reputable read on the particular topic.
Become a topical authority
In order to be authoritative about a particular topic that’s important to your business — or even the crux of your business — you need to build great interconnected content around this topic to demonstrate your expertise and authority.
According to Google’s own E-E-A-T guidelines, your content should be “helpful and relevant,” and thus you’ll be rewarded with the oomph you need to rank well for the topic, given you are optimized and link well.
It’s hard work, but pays dividends when it all comes together. Being an expert on a particular topic that’s critical to your business where you appear in search results, SGE included, for each stage of the funnel in your customer’s journey will get you remembered. And that value goes far beyond SEO.
Being a true topical authority can equate searchers scrolling past SGE and competitor search results, seeking out your result. Being a true topical authority means searches on the topic will begin to include your brand name to filter out the noise to more effectively get to the authority they trust.
LLMs need our content
The reality is, in order for GAI to give us a response, it needs someone to write about it first. GAI can’t be the original source of the information, and must rely on what humans have produced. This is particularly true for new topics. Someone needs to write about any new topic for the LLMs to consume it and regurgitate it in GAI.
The introduction of SGE will likely have an impact on content farms, as we may well see these dry up and close shop as a result of dwindling referring traffic. If the content pool shrinks too much, then Google will have a massive problem on their hands: They can't answer user queries without content.
And while there will be closure and consolidation, as soon as one site lets their foot off the content production pedal, there will likely be another site there to pick up the slack.
The GAI content boom
With the introduction of ChatGPT in late 2022, came widespread speculation that many content creators were going to revert to producing content using GAI and content creator jobs were in danger, especially with headlines like “Will searchbots put me out of a job?”
There’s reason to be concerned, as LLM chatbots do a decent job of producing content that even AI detection systems can’t pick up. If the U.S. Constitution was flagged as being AI by all three “leading” AI-detection systems, what hope do we have knowing what we are reading was produced by AI?
In reality, a study by Morgan Stanley in April 2023 revealed that only 19% of respondents had used ChatGPT at all, and the majority of respondents said they were unlikely to use GAI tools in the next six months. Speaking with colleagues, few actually use LLM chatbots on any regular basis and no one I work with has ever used Google’s SGE.
It’s more than likely that the rush to use GAI for content generation has been by those keen to churn out content at scale, not your competitors looking to produce quality content. Low quality “content farms” are the most likely culprits using and abusing chatbot LLMs to try to game the search results.
After years of prohibiting GAI content, Google revised their stance to say their “focus is on the quality of content, rather than how content is produced.” However, content farms beware, Google also states that those “using automation—including AI—to generate content with the primary purpose of manipulating ranking in search results is a violation of our spam policies.”
Google’s John Mueller warns, “If you're using AI to write your content, it's going to be rehashed from other sites." No matter how good your prompts for GAI are, the responses are always going to be average.
Google’s algorithms are highly sophisticated and designed to detect low-quality content. As a result, regurgitated GAI content — even average, run-of-the-mill content providing little added value to what’s already out there — will not surface to the top of search rankings. Thus, the advent of GAI content actually raises the bar on what qualifies as quality content.
Google’s own aim is to deliver the best possible search results to its searchers, and thus, high-quality content will continue to be rewarded with top-ranking positions and SGE citations.
Is GAI coming for your job if you’re a content creator?
Don’t take my word for it — seriously, please don’t — but in my professional opinion, I would say if all you do is write, it might be advisable to get more involved in the entire content creation process: ideation, strategy, outlining, project management, workflows, editing, and so on. LLM chatbots cannot compete with a multifaceted content creator. And someone has to manage the bots. Take it from Claude:
Where to use humans and where to use GAI for SEO-led content?
GAI is excellent for supplementing content writing, for instance, to help brainstorm topics and ideas, generate outlines, and rework content. Use our AI Content Generator to automate repetitive tasks such as writing meta descriptions or translating content. Augmenting SEO-led content with GAI allows for more of a focus on the human element of your writing: storytelling.
Anything with first-party data is essentially impossible for GAI to produce without all the human inputs. Long-form formats like data studies, research reports, and elaborate guides are highly difficult for GAI to produce without heavy human guidance and editing.
GAI will also struggle to understand your customers and target audience the way that you do. You understand the behaviors that lead to conversions. Knowing this can allow you to produce the content that captures your audience at the right time in their journey.
GAI may help you produce this content, but it can’t necessarily help you know what content to produce. The relationship between GAI and content creators is not one of replacement, but rather collaboration.
The intersection of AI and SEO continues to be complex and rapidly evolving, and it’s anyone’s guess what’s next.
And while Google’s SGE will forever change how searchers interact with search results, it doesn't signal the end of SEO-led content. Rather, it emphasizes the importance of producing high-quality, authoritative content that can be surfaced in GAI responses.
SEO-led content remains a powerful tool earning visibility in organic search results, so long as content creators produce high-quality content tailored to their searcher’s needs with their unique human point of view.