Sage SEO's Owl, Owly, doing keyword research
The Promise and Perils of AI in SEO Keyword Research
February 12, 2023
Peter Yeargin

I’m frustrated.

The reason I’m frustrated is because of the exponentially increasing number of articles on the internet talking about using AI to do keyword research but not understanding how keyword research actually works.

Any good keyword strategy starts with researching your customer. Recently, the buzz is all about ChatGPT as an ai writer and how it solves all of your SEO challenges.

The problem is that ChatGPT isn’t good at keyword research. It’s good at coming up with something called semantic keywords. But that’s just a college-essay way of saying “similar topics or ideas”. It doesn’t mean they are easier to rank for ,or when put together into a blog post, that they will generate a front-page ranked article.

Today, I want to dive into this topic in more detail.

Before we get started

…a bit of background.

Prior to November of 2022 (ChatGPT’s launch), there were any number of articles about the best SEO tools or keyword research tools all over the internet. The reviews would cover topics like keyword ranking and keyword analysis, they’d dive into the keyword difficulty of a search term. There were many free keyword research tools like Google Ads keyword planner or keywords everywhere.

You could use a free keyword research tool like Google Autocomplete or Google’s People Also Ask.

Google's People Also ask feature as a free seo tool
A snapshot of “People Also Ask” in a search result

Paid tools are also in abundance; tools like and are the gold standard when it comes to keyword research. But they are also built primarily for marketing professionals.

The result of this is that the average small business owner or early-stage startup typically might turn to paid ads instead of SEO because they didn’t have time to learn how to use the tools.

Enter ChatGPT (and my frustration). Look at all the things it ChatGPT can do in mere seconds!!

  • You can brainstorm content ideas quickly
  • You can write content within minutes instead of hours
  • You can research keywords within seconds

Unfortunately, that’s not accurate. And here is why…

AI Tools don’t know how to do keyword research

This isn’t to say AI is bad at components of keyword research. But AI SEO by itself is much less effective than even basic human common sense.

This is why AI SEO tools by themselves will always fail to deliver good results.

Here is a quick example of how plausibly good ChatGPT is at coming up with keywords for a new article on dating apps for black professional women. I gave it the prompt give me a list of 10 keywords on the topic of dating apps for black professional women in 2023

Starting with ChatGPT to generate keyword lists is great but find other data tools to complement the process

This is a great list. It gives you some good ideas of topics and keywords to write about. But it has little influence on how rankable the article you write will be. Will these keywords get you started writing? Absolutely!

Will you be ranked on the front page of Google by using keyword research from ChatGPT?

Not a chance.

Here’s why.

If I tweak my prompt just a bit and add the phrase easier to rank for to it, let’s see what it spits out.

chatgpt generates plausible lists of keywords but has no data to back them up as good targets for keyword research

Now, let’s look at this list of keywords in a keyword research tool like

ahrefs - if it isn't searched, then it's not worth targeting for a keyword

Only one of the keywords is even in the database and has almost no search volume monthly. ChatGPT and other AI tools like it have no access to search analytics data. They can’t recognize search intent, and therefore, they have no idea about keyword optimization. What they are good at is finding those semantic keywords we mentioned above.

But semantic keywords have nothing to do with keyword research. It has to do with human language and how we associate different words. The AI isn’t doing keyword research to find those words. It’s using predictive analytics to find patterns in human language. It’s a meaningful component of keyword research and aligns well with Google’s north star of search intent. 

Unfortunately, what it doesn’t do is help you at all with the core goal of SEO: ranking on the front page of Google search results.

One more example to illustrate the problem

Let’s say we decide to have ChatGPT generate a list of keywords that are a bit broader and are searched more often in Google.

ChatGPT generates semantic keywords effectively for SEO purposes

These are much shorter keywords, but the problem remains. Of the 30 keywords it “suggested,” only 14 were searched enough on Google to warrant an entry in the Ahrefs database:

ahrefs showing data for low searched terms

And of the 14 that were, only three of them (best dating apps for black women, black women dating, and black woman dating site) had any notable search volume.

The problem remains that AI without keyword data is relatively useless when it comes to doing keyword research.

ChatGPT can solve all my SEO problems

Since its launch, the world has exploded with new articles talking about the power of OpenAI’s software to automate really complex parts of SEO. The first problem many of them tackled…naturally one of the hardest parts of good SEO, is keyword research.

Many marketing professionals, including more SEO professionals than I can count, think it can solve the keyword research problem alone, or with a bit of human help. But keyword research is not a problem of language. It’s a problem of data.

We’ll talk more about that in a future article.

However, some of the great things it can help with for your SEO workflows are:

  • Writing a content brief for a new article you want to write about, including potential sources of research
  • Creating outlines for long-form content
  • Mitigating the problem of staring at a blank screen
  • Writing a blog post introduction to get you started
  • Putting together a bulleted or numbered list for a particularly complex procedure you want to include in your article

What SEO problems can ChatGPT help with?

Writing content briefs for an article

ChatGPT is astoundingly good at generating what content marketers call content briefs. Given the right prompt data, it can generate content briefs like this for an article one of our customers is writing.

You can see it shows both a target audience as well as target keywords and dives into details for the article content. It even cited some sources that were from January of 2023. Looks like it’s getting more and more current.

ChatGPT can create great content briefs for your writers

Creating outlines for long-form content

If I were to ask it to create an outline for a blog post on the challenges of doing SEO keyword research, ChatGPT is more than capable:

ChatGPT generates blog post outlines

You can see the quality and depth of the outline is generic. But it gets you started and helps you ideate and expand on any of those topics.

Enumerating mundane procedures with bulleted and numbered lists

This is one of my favorite use cases for ChatGPT. Oftentimes in the SEO world, we have to do a bunch of mundane things like updating DNS records or finding reports in Google Analytics. ChatGPT is amazing at helping us create documentation to teach our customers the right procedures.


In this article, I delved into the limitations of AI in keyword research and its use in SEO workflows. We talked through the nunaces of how AI tools, including ChatGPT, can generate semantic keywords like an SEO pro, but when it comes to actual keyword research, they fall short.

You see, keyword research requires more than just language skills, it requires access to search analytics data and an understanding of search intent – things AI can’t seem to grasp (yet). But, on the bright side, ChatGPT can help with other SEO tasks like writing content briefs, creating outlines, and even enumerating procedures.

While AI may not be the silver bullet for all your SEO problems, it can still be a helpful sidekick. And, let’s face it, SEO is already confusing enough; no need to make it any more complicated!

Share this

Twitter Facebook LinkedIn Email

Related Posts