Author: Jack Treseler
Odds are this isn’t the first time you’ve heard of ChatGPT. In fact, 100 million people signed up to use ChatGPT within two months of its launch. And with the introduction of GPT-4, it seems like every SEO is wondering whether it can help them become more efficient, scale strategies, or if it’s even safe for brands to begin with.
In this article, we’ll examine all those factors so that you can determine whether generative AI is actually the blessing to SEO that some have touted it to.
Table of contents:
What is ChatGPT and how did we get here?
ChatGPT originated from OpenAI’s development of a technology called “Generative Pre-training Transformers” (GPT for short) back in 2018. Basically, the company wants to create an AI that can better interact with humans and help people become more efficient.
GPT is designed to generate human-like text by predicting the next word in a sentence based on what the previous words were.
Programmers have been using GPT in chatbots, virtual assistants, and content generation tools since OpenAI released its first iteration called “GPT-1” (super original name, right?).
OpenAI used human trainers to improve GPT-1’s performance, ultimately creating two more versions (GPT-2 and GPT-3, respectively) before launching ChatGPT (and as of March 2023, GPT-4).
We’ll dive into how to use GPT a little later, but first, you need to be aware of the potential tradeoffs.
Inherent bias in ChatGPT
ChatGPT sounds like an incredible technology that SEOs can leverage to programmatically scale their work. But, there’s also a big risk—inherent bias. And to its credit, OpenAI is aware there’s bias in ChatGPT’s outputs and even acknowledged as much:
“The model can have various biases in its outputs—we have made progress on these but there’s still more to do. Per our recent blog post, we aim to make AI systems we build have reasonable default behaviors that reflect a wide swathe of users’ values, allow those systems to be customized within broad bounds, and get public input on what those bounds should be.” — OpenAI
Be that as it may, there’s a very big risk that the ongoing lack of diversity (in both thought and life experience among the leadership and engineers) could cause ChatGPT to continue a culture of racism, sexism, and confirmation bias already present in today’s top search engines.
For SEOs, this goes beyond perpetuating negative stereotypes. Biases can also affect ChatGPT’s ability to provide accurate or compliant content for highly regulated industries like fintech or alternative health. Remember:
ChatGPT is about predicting the next phrase in a string of content, not about verifying claims and facts. SEOs will still need to perform their own research, analysis, and revisions.
Because of this, there’s a very real risk that ChatGPT can generate misinformation or simply false information (more on this later).
Who is integrating ChatGPT into their platforms?
The biggest names integrating ChatGPT include Microsoft and its search engine Bing. Microsoft purchased an exclusive license to the underlying technology behind GPT-3 in 2020 after investing $1 billion in OpenAI in 2019.
Some of the biggest brands across several noteworthy industries have already integrated ChatGPT into their systems. Most notorious among them is Bing.
The mass integration of chatbots into search engines and operating systems has to do with how users interact with content on the web. Users want to consume information in the most expeditious manner possible, and asking questions the same way they would in everyday speech may be the easiest learning curve.
In essence, ChatGPT has already become intertwined with:
Search engines
Web browsers
Dating sites
Mobile games
Data infrastructure
Marketing
And as marketers, it’s probably best we figure out how to most effectively (and safely) use this technology.
How does ChatGPT work?
At its core, ChatGPT uses massive amounts of data and machine learning algorithms to understand how we talk and what we mean when we communicate. Then it uses generative AI to provide high-quality responses that SEOs can use to save time and increase productivity, freeing you up to focus on more strategy-intensive work.
Generative content in a nutshell
Generative AI platforms (like ChatGPT) take commands from users and create human-like content as the output. That might be as simple as asking it to suggest pizza restaurants in your area:
But, it’s also capable of incorporating data that you feed it, opening up a whole universe of potential applications. Generally, that data includes text and images, but it could also include other media. At this point, a user would feed the generative AI model a prompt or question. The model then uses the data it’s been trained on to create brand new, original content to answer the prompt.
Below is an example of how ChatGPT can work for users:
The prompt I fed ChatGPT + the output | The data it used to create the content |
 |  |
You can use ChatGPT to create blog articles, guides, and even code.
What algorithm does ChatGPT use?
ChatGPT is built on top of OpenAI’s GPT-4 family of large language models and has been fine-tuned using both supervised and reinforcement learning techniques. Basically, ChatGPT is capable of processing image and text inputs and producing text outputs.
Supervised learning is what it sounds like: engineers train the machine learning model with data that has labels and examples. They’re essentially holding its hand as it figures out what content to generate.
Reinforcement learning is when the machine learning model learns to take actions to maximize a reward model: If you look at the ChatGPT results, you’ll see a little thumbs up icon and a thumbs down icon (shown in the screenshot below).
Reinforcement learning means ChatGPT wants as many thumbs up as possible, and as few thumbs down as possible.
To translate: ChatGPT got a lot of training and feedback to get to where it is today. And with each new prompt, it wants to get better for you, so if you like the output, give it a thumbs up.
Where does ChatGPT get its information?
ChatGPT gets its information from a huge set of data consisting of books, articles, academic papers, websites, Wikipedia (and Wikimedia), and other publicly available information.
The content in that data includes topics around science, technology, history, literature, and more.
While ChatGPT can access a ton of information, there is one important caveat: The output isn’t based on the veracity of that information—it only looks for statistical patterns and produces the next most likely output.
ChatGPT is a language model, it’s not looking at what’s true or not.
In fact, the creators of ChatGPT said the same thing:
“GPT-4 has similar limitations as earlier GPT models. Most importantly, it still is not fully reliable (it ‘hallucinates’ facts and makes reasoning errors). Great care should be taken when using language model outputs, particularly in high-stakes contexts [...] or avoid high-stakes uses altogether.” — OpenAI
Is ChatGPT safe to use?
Honestly, it’s a mixed bag. Below are some concerns people have expressed.
Concern | Legit? | Popular consensus | How SEOs can work around this |
ChatGPT lies to you. | Yes. | ChatGPT can be a fountain of misinformation if not carefully monitored. In fact, the latest source of information comes from 2021. That means ChatGPT isn’t privy to any more-recent developments. | Feed ChatGPT the background information it needs to accurately create content. By “training” the model on specific information, your output will be much more accurate, and take less time to edit. |
ChatGPT’s responses are too robotic and unnatural. | Yes and no. | ChatGPT is a language model, and profiling language constructs is very much within Google's wheelhouse (it’s been using n-grams and lexical parity for a while). Tools already exist to identify AI written content. | The quality of your output depends on your input. Try adding conditions around tone of voice, writing personas, and audience to get a more natural output. Then make sure to edit the results. |
There’s a mobile app. | Yes. | ChatGPT didn’t launch with a mobile app, but there’s now an iOS app available. |
The basics of using ChatGPT for SEO
When using ChatGPT (or any generative AI model) to create content, you’ll want to review the content to make sure it doesn’t miss the mark, or worse yet, potentially get you into legal trouble.
It’s especially important for SEOs to have a heavy editing hand when it comes to the output ChatGPT gives you. Even more so if you’re in a highly regulated industry like fintech marketing, where Google is more critical of content you publish (commonly called “Your Money or Your Life” (YMYL) content).
Below is a breakdown of potential generative AI use-cases and their associated issues, organized by business type.
Business type | Common Problem(s) | How ChatGPT can help |
Enterprises ($100m+) | Efficiency and getting things across the finish line | Quickly generate articles, FAQs, or other content that has search demand. Use this for lead generation. |
Mid-sized startups | Limited resources, budget, and bandwidth | Provide prompts and let ChatGPT create unique content at scale. Then review the content and map to your target keywords. |
Franchises | Creating unique, localized content for each location | Provide prompts and create flexible page templates and localized content. |
Local SMBs | No time to create content | Use ChatGPT to quickly answer common questions customers have in minutes. |
eCommerce Brands | Creating unique content for category and product pages | Quickly generate unique content and FAQs, and even implement chatbot functionality to help shoppers. |
Highly regulated markets | Create compliant content that drives sales | Quickly scale the creation of content, knowledge centers, and dynamic widgets for better customer experiences. Be sure to have an expert review all output. |
While ChatGPT is limited, it provides enough capability for SEOs to get creative. Some SEOs have used it for a range of tasks, including:
Localizing content by country, region, and local idioms
Rewrite .htaccess rules for domain migrations and 301 redirects
Generate dynamic XML sitemaps by category
Create link building emails and subject lines
Keep in mind: As mentioned before, if you’re using ChatGPT to create content, Google can tell. Does Google care? Not really, it cares more about the quality and accuracy of the content. If you’re using ChatGPT to produce or translate content, don’t trust its output. Have an editor or translator give it a read first before publishing.
What SEOs shouldn't use ChatGPT for
Ethical usage of ChatGPT is absolutely critical—not just for your SEO, but also to ensure that you’re not inadvertently spreading misinformation.
01. Retrieving sources for claims
Why someone might want to: You would think—based on all the info ChatGPT has at its disposal—it’d be able to just give you the sources of information and save you some time.
Why you really shouldn’t: The sources are absolute garbage (i.e., they don’t exist). Remember, ChatGPT is a language model. It’s not looking at whether something is true or not. It’s looking at what the next likely text is given your input. Need proof? Talk to Alan Wolk, co-founder of TVRev. Here’s his experience:
“ChatGPT can be a fountain of misinformation. To wit: I asked it to provide me with some articles about FASTs. It came back with this seemingly legit list. The reporters are real (I even know some of them IRL) and work/have worked for the publications mentioned and these are topics they and the publications might write about. But.. none of these are real. Repeat: none of the articles in this list exist. The links turn up dead pages and a Google search indicates they've never existed-- ChatGPT made them up from whole cloth.” — Alan Wolk, co-founder of TVRev
Here is the output that ChatGPT gave Alan:
This may not be the case indefinitely, though: Bing’s chatbot “Prometheus” can integrate citations into sentences in the chat answer so that users can click to access those sources and verify the information. Jasper AI also cites its sources in its chat product.
02. Plagiarism
Why someone might want to: Look, we’re all strapped for time. And if someone wrote a great piece that’s already ranking well on Google, why not just use ChatGPT to change the way it’s written, so “your new content” appears unique to search engines? This has been a very common SEO hack that has been around for some time, so why not leverage ChatGPT to do so at scale?
Why you really shouldn’t: You’ll get caught. There are tell-tale signs content came from generative AI, and as said before, Google can use natural language processing to determine low quality, machine-generated content from high quality, unique content. Without human oversight, you’ll get busted.
03. Scraping Google search results.
It’s tempting to take something more complex (that developers generally have to do for you), and just have ChatGPT take care of it. This is especially tempting if you want to pair the top search results you scrape with ChatGPT’s ability to write content and build high quality, optimized content. However, remember Alan’s experience—the results ChatGPT gives you are not real. They are just content created to look like the output you ask for.
I asked ChatGPT to create a list of topics family law attorneys could write content about, but when asked to find the top three ranked URL for those topics, each URL shown did not actually exist.
How SEO professionals can use ChatGPT at scale
ChatGPT can be a massive boon to SEO if used properly. Outside of content creation, generative AI can help analyze large swaths of data and find trends human eyes can miss.
01. Use ChatGPT to identify Google entities and subtopics
Why: Entity analysis and competitor analysis can take an SEO between one to four hours, depending on how much they need to learn about the topic before they can begin analyzing the content. With ChatGPT, it can take minutes.
Example prompt:
You are a search engine optimization specialist whose job it is to figure out the most common entities and subtopics featured in the following urls. An explanation for entities can be found here: https://www.searchenginejournal.com/entity-seo/277231/
The URLs you are to analyze are:
https://www.g2.com/categories/medical-billing/small-business
https://www.capterra.com/medical-billing-software/s/small-businesses/
https://www.getapp.com/healthcare-pharmaceuticals-software/medical-billing/org/small-business/
Example output:
Pro tip: Share this with your PPC team (if you have one). You can also feed the keywords your PPC team is bidding to ChatGPT and ask it to create content topics for you based on that data.
02. Use ChatGPT to create comprehensive content briefs at scale
Why: Although ChatGPT can write the content, it’s best to have experienced writers that know the subject matter create the content. The language structures humans tend to use are drastically different from the language structures ChatGPT uses. For industries where “Experience, Expertise, Authoritativeness, and Trustworthiness” (E-E-A-T) content is needed to rank well, human writers are still worth their weight in gold. But to save time, have ChatGPT put together comprehensive briefs for them to get started.
In the example below, I’ve divided the task into two prompts, with the second building on the first.
Example prompt 1:
You are a search engine optimization specialist tasked with creating engaging content for your audience. Your audience are independent, small family law firms looking to improve their customer retention. Create 10 blog post topics that will catch their eye
Example output 1:
Example prompt 2:
Of the topics generated above, create an outline that summarizes the main message of the article, what subtopics a writer should include, and keywords that should be integrated into the blog post
Example output 2:
You will still need to take the briefs and match them to target keywords, but this strategy can still remove roughly 70% of the time it takes to produce this kind of detailed direction for your writing team.
Industries are responding to ChatGPT
People everywhere are navigating the potential impact of easily accessible generative AI. Here are a few of the changes (and potential changes) that are occurring:
Search algorithm updates: Google’s already addressed how it handles AI-generated content. You can expect search engines to become more and more advanced in how they detect and rank content that uses generative AI.
Social media algorithm updates (eventually): If history tells us anything, it will take years for social media platforms to adjust and successfully moderate AI-generated content. Odds are, the algorithms will reward marketers that can identify content frameworks that generate shares, clicks, and traffic, and scale those frameworks with generative AI. Expect echo chambers and repetitive blog post templates until platforms find ways to deal with the influx.
General content trends: Content development will multiply. Blog posts are super easy to mass produce now. But, will this efficiency lead to echo chambers? Probably. Will quality, unique content ultimately outperform these echo chambers? I hope so.
General market trends: Lower-priced general marketing services may get commoditized. Companies and marketers that can layer data and nuance on top of generative AI will have a huge advantage.
Ethical AI use is proper AI use
Generative AI is here to stay. There’s no putting this genie back in the bottle. And let’s be honest: We marketers have a history of abusing new technology to further our business needs, and often we don’t look at the potential consequences—like racist and predatory ad targeting practices, algorithmically created echo chambers, and invasive ad strategies.
So, as we use ChatGPT, let’s remember we marketers have a greater responsibility, now more than ever, to use this technology responsibly and educate consumers who will undoubtedly be the targets of this new tech.
Jack has over a decade's worth of experience in SEO. He's CEO of Crescendo Consulting, which specializes in marketing early and mid-stage startups in highly regulated industries (think Fintech and CBD startups). He's a fan of pineapples on pizza and Star Wars Episode I.