Many experienced SEOs love their work because it's always evolving, but the accelerating rate of Google updates and SERP changes might be making their heads spin.
It’s no exaggeration to say that the most fascinating (and dizzying) changes in SEO in recent years have involved AI and machine learning, both as they’re used by Google and by SEOs (and their tools/platforms).
But few announcements in the field of AI have caused as much of a stir as the recent unveiling of ChatGPT.
In this post, I’ll cover:
- The impact and implications of ChatGPT and its underlying technology
- The many ways in which AI/machine learning is already being used related to SEO
What Is ChatGPT?
ChatGPT is the most accessible and stunning implementation of the technology of AI research firm OpenAI. Its purpose is to make the experience of the most advanced iteration of interactive AI chat accessible to the masses.
If you want to know more about ChatGPT and its history, Roger Montti has an excellent writeup at Search Engine Journal, where he defined it as “a large language model chatbot developed by OpenAI based on GPT-3.5. It has a remarkable ability to interact in conversational dialogue form and provide responses that can appear surprisingly human.”
ChatGPT and SEO
There are two reasons SEOs in particular have been so worked up about ChatGPT: speculation that it could replace traditional search engines like Google, and create usable content without involving human writers.
Is ChatGPT a Google Killer?
Some SEOs have speculated that given its ability to provide quick, detailed, coherent answers to a great many questions, ChatGPT or future implementations of its underlying engine could become preferred by searchers over traditional search engines such as Google.
The thinking goes that, parallel to Google’s own Answer Boxes, ChatGPT answers eliminate the increased user effort involved in scanning a list of results and making judgments on which are best, then having to click through to an external web site.
However, there are several problems with this scenario:
- ChatGPT can’t (yet) answer all questions or user needs. Though its training sources are very large, it still draws from a much smaller pool of information than does Google. It also has little knowledge of products and ecommerce, one of the major reasons users go to Google. Of course, if the technology is eventually applied to a data set comparable to Google’s, then this objection would fall away.
- ChatGPT isn’t always accurate. Users have discovered many instances in which ChatGPT provides incorrect or misleading information. To be fair, this is a problem which has plagued Google’s Answer Boxes also, but at least with Google, users can look beyond the Answer Box to peruse other results. (Also, Google claims to be taking big steps to improve the accuracy of its Featured Snippets.)
OpenAI released a new version of ChatGPT on January 10, 2023, but Search Engine Journal notes that it still is limited to information before 2021, and found that it still failed some test queries it had failed previously.
ChatGPT’s summary of its capabilities and limitations (Source)
- ChatGPT is still too query dependent. The quality of answers ChatGPT gives is highly dependent on how clear and well-structured the user’s queries are. This makes it similar to the early days of Google, when users found they had to learn to structure and refine their queries just so in order to get the responses they wanted. In recent years Google has largely overcome that limitation with technologies such as RankBrain and BERT that help it better recognize the actual intent of varied human queries.
- Google is already firing back. In fact, it appears Google’s response to any perceived threat from AI chatbot implementations may already be in formation. For example, Google and its DeepMind project recently announced MultiMedQA, an open-source large language model for health information. It includes facilities to attempt to vet health information for factors such as factuality, precision, potential harm, and bias.Its intent is to allow users to interact with a chatbot in a way they might in an initial consultation with a health professional. While not yet incorporated into Google Search, it certainly could be in the near future.
Of course, ChatGPT and related technologies will probably overcome many of these limitations in the future, but at this point it still lags behind the utility of Google and other traditional search engines.
Google and its parent Alphabet are perhaps the entities most heavily invested in AI research in the world, so it seems likely that they are working to meet and exceed any challenges thrown up by third-party AI developers.
In fact, Google has already incorporated many functions similar to what ChatGPT does. For example, prompting users to enter follow up queries and offering aids to explore a topic more deeply after its own instant answers.
However, we should not downplay the significant achievement that ChatGPT represents. Its current limitations will lessen over time, and the profound impression it has made on both the public and the tech communities indicates conversational AI will be with us from here on out.
Can ChatGPT Replace the Need for Human Writers?
Over the past several years SEOs have become more and more focused on content as an essential part of their SEO strategy.
And for very good reasons. Google has been lessening its dependence on links and other easily-gamable authority signals in favor of increasing its ability to evaluate the relevancy, authority, and reliability of pages and sites that might answer a given query.
That means that it is paramount to have the best possible authoritative, complete, and accurate content on pages you want to rank well.
Recommended Reading: Is Full Content Automation Possible?
But creating such content is time consuming and expensive. It’s understandable that SEOs would be on the lookout for anything that might help reduce the cost of creating content while increasing the amount of content that can be added to pages.
So ChatGPT’s ability to generate what appears to be original content that, at least at first glance, seems human written, seemed like the equivalent of discovering the goose laying golden eggs. Add to that its ability to produce varied forms of content, such as FAQs, summaries, data tables, etc., and it seemed like a slam dunk solution to our content woes.
However, as it turns out, it is not the panacea to our content-production woes some might have hoped. Here are some of the reasons ChatGPT is not ready for prime time as a content generator for sites:
- ChatGPT usually fails the most reliable AI detectors. As the quest to create credible machine-generated content has amped up, so also has a whole micro industry dedicated to its detection. Glenn Gabe compiled a comprehensive list of the best AI content detectors, and shows that ChatGPT content unaltered by humans consistently gets red-flagged by the best detectors. As of this writing we seem to be entering a sort of arms race escalation, as users experiment with various hacks (such as having another AI rewrite the first AI’s content, with new or improved detection software then emerging able to sniff out many of those attempts.
Example of output from one AI text detector from Glenn Gabe’s post (used by permission)
- ChatGPT is sometimes inaccurate… or worse. As good as ChatGPT is at providing a basic answer to a question and making it sound credible, more and more testers report seeing clearly incorrect or misleading information in its replies. Even worse, it has been known to generate racist or sexist replies. Some of which can be faulted to how users frame their questions, so AI chatbot creators will need to improve their ability to better understand the real intent of queries.
- ChatGPT is bad at Google’s E-E-A-T standard. In its most recent release of its Quality Raters Guidelines, Google continued to emphasize that for topic where the answers matter to users, there should be evidence of either Experience (first “E”) or Expertise (second “E”) for the author or the content relative to the topic. ChatGPT content displays neither. Typically for more complex topics its responses sound like those you’d get from a generic writer assigned to do a few Google searches and assemble some information gleaned into paragraphs.
- ChatGPT probably doesn’t get past Google’s Helpful Content System. Google introduced its Helpful Content System into its ranking algorithms in August 2022 and updated it in December. This algorithm is specifically designed to sniff out and devalue pages (and possibly entire sites) that don’t have content “written by humans for humans.” If the third party AI detectors noted above are good at doing this, you can bet Google is even better.
- ChatGPT content will be watermarked. OpenAI and many other AI chatbot developers are voluntarily participating in development of watermarking systems so that AI content is self-identifying. These systems create telltale patterns of punctuation and word choice that is undetectable to humans but can be easily noticed by machines that know the key.
My conclusion is that, at least in its present form, ChatGPT and its underlying GPT-3 engine are neither a threat to traditional search engines nor a viable solution for creating content that is likely to rank in Google.
However, Google’s John Mueller recently clarified that Google isn’t against AI content per se. The issue is AI content not being able to come up to Google’s standards for content that should rank in search. But Mueller went on to say that if at some point AI is able to generate content that met those standards, then Google wouldn’t care about the source of the content.
It should be noted that OpenAI plans to release GPT-4 in 2023, which will bring its information database more up to date and undoubtedly provide many improvements to the output of ChatGPT.
So Does ChatGPT Have Any Use for SEO?
Even with the above-stated limitations and caveats, there may be some ways in which ChatGPT could be useful for SEO, particularly as part of the content production workflow. While human involvement is still irreplaceable for quality content, tools like ChatGPT could take some of the “grunt work” out of the process.
Here are some ways ChatGPT could be useful for content production:
- Basic topical research: When exploring a new topic, some queries on ChatGPT could give you some ideas for further research, or suggest a basic structure for your inquiries.
- Summarize large texts: With a browser plugin like Summarize you can get summaries of large online articles from ChatGPT.
- Create an FAQ: You can input text in ChatGPT and ask it to turn it into an FAQ. Here’s an example where I asked it to create an FAQ from its own response to my query: “Explain quantum mechanics”:
- Generate RegEx and Excel Formulas: ChatGPT can actually build regular expressions and complex Excel formulas from your plain English instructions. (See this Search Engine Journal post for examples.)
- Create starter text: Use ChatGPT to generate starter text that you then assign to a qualified writer in a content brief. Pro tip: seoClarity users can place ChatGPT output into Content Fusion with a main topic specified and generate a complete content brief including topics and keywords that should be included.
ChatGPT text inserted into seoClarity Content Fusion
For an even larger list of possible SEO uses for ChatGPT, I recommend this resource from Aleyda Solis.
Since, as we’ve established above, ChatGPT’s raw output does not (yet) provide content that meets Google’s guidelines for quality, its best use for now may be as an assistant for qualified human writers.
Sites relying on unedited AI content will likely not be able to rank well, especially for competitive terms.
Enterprise SEOs should approach anything being promoted as a “magic bullet” for SEO with a healthy amount of skepticism. Their organic positioning and traffic are far too valuable to be risked on unproven strategies, especially when, as in the case of AI-generated content at present, it seems clear the result does not meet Google standards and might be easily detected by them.
That being said, the current iteration of ChatGPT deserves much of the attention it has received. Beyond question it is a major advance in chatbot capabilities. Enterprise SEOs should keep an eye on it as it develops.
Here’s a chart by Mark Williams-Cook posted on Twitter that maps out a likely scenario for the present and future of ChatGPT. As the technology advances, some judicious uses of it may be warranted.
Beyond AI Content: Other AI SEO Applications
While ChatGPT has shifted our attention for now to content-related applications of AI technology, we shouldn’t forget that AI has many other uses related to SEO.
AI and Google Itself
Certainly the broadest and deepest-reaching applications of AI/machine learning in the SEO world are being created by Google itself. I’ll only touch on this here as it has been well covered elsewhere, and is, of course, the aspect of SEO AI we have the least control over.
The overall intent of Google’s use of AI seems to center around being able to understand both user intent and the relevance and authority of content, and match the two up better and better.
Here’s a quick survey of some of Google’s publicly-known applications of AI/machine learning:
- Rank Brain helps Google understand how the words searchers use relate to entities and concepts.
- Neural matching takes a broader, more contextual look at both queries and content.
- BERT is a natural language processing system that helps Google better understand the meaning of words used in queries.
- MUM helps Google better understand the way languages work.
AI and Keyword Research
Identifying the keywords your site should focus on has become an increasingly complex task over the years.
Long gone are the days when simply putting a keyword on a page enough times made any difference for ranking in Google. With the AI tools listed above Google is able to look beyond keywords on a page to evaluate the total relevance and thoroughness of the content in relation to potential queries.
On top of that, the number of potential keywords for any given topic is always growing.
So it follows that our keyword research needs to become more sophisticated, while also able to account for the massive scale of available content for a topic.
seoClarity has risen to this challenge with industry-leading applications of AI/machine learning to more than meet the needs of enterprise keyword research in 2023.
Our Research Grid is the largest and freshest dataset of industry-driven keywords on the planet. We use AI both in determining which keywords to add and track as well as in determining things like search intent, true search competitors, and true search volume (the latter from a complex proprietary algorithm using clickstream data mixed with other sources).
seoClarity’s Research Grid Hub
seoClarity’s Topic Explorer is by far the largest topical research engine available, with over 30 billion topics and over a billion added each month. We use AI in Topic Explorer to build super-useful capabilities like Intelligent Match, which finds relevant keywords no other tool can identify, using our proprietary Intent Similarity™️ to show how closely overlapping their audience is to your topic.
AI and Content Optimization
The factors that make a piece of content more likely to rank well are complex and variable. How are they variable? Google has told us that its ranking algorithm these days is actually a mix of many separate algorithms, and which come into play during any particular search varies according to the intent, context, and many other factors.
This would leave organic content creators simply guessing in the dark if it weren’t for artificial intelligence.
seoClarity uses AI in Content Fusion to reverse engineer the characteristics of top-ranked content to guide writers in what they need to do to create content more likely to rank well. We use that information to produce a Content Score that updates in real time as your writer incorporates our suggestions to improve the content.
The Content Score in Content Fusion
AI and the Future of SEO
The one certainty about AI and its involvement in SEO is that it is here to stay. Whether we’re looking at the scale of what Google does or the scale of an enterprise web site, the amount of data, insight analysis, and execution to be done long ago moved beyond the capacities of even the largest SEO teams.
It’s undoubtable that the capabilities of AI-based tools and services will continue to increase in the years ahead. As with most technological innovations, we should neither fear it nor ignore it. Instead, wise SEOs will concentrate on discerning where the technology is truly helpful in enabling them to do more at scale while staying aware of the elements that will always need human judgment and intervention.
At present, the smartest move might be to invest in SEO tools and platforms that already incorporate AI in ways that have been well tested and proven. seoClarity stands out in that aspect, incorporating AI and machine learning to move you from data to insights to actions at enterprise speed and scale.