OK, so I love using AI to automate and speed up a lot of SEO workflows. For the past year, I’ve been using it for almost nearly everything I touch including data analysis, ICP content mapping, content outlines and much more. However, during that time, there’s been one thing that’s been escaping me.

Keyword Research.

You think this is a fairly basic task that tools like ChatGPT would be easily able to help with. However, in general the tests I’ve tried have turned out pretty lackluster. While you would think that simply giving a large list of potential search terms would be rather straightforward it isn’t.

I think Mark Williams Cook did a good job of summarizing ChatGPT’s weaknesses in a LinkedIn post:

 

An Example: Domo

So let’s use the site Domo as an example. Domo is a business intelligence platform that allows businesses to analyze and connect large sets of data they use. It’s a more complex product and one that might require an newer SEO more upfront research to better understand their keyword landscape.

So let’s say that you’re using ChatGPT to find keywords related to Domo product offering. A lot of times, SEO might use a single-shot prompt that simply asks ChatGPT to find keywords related to a company. I did this for Domo and the results are pretty underwhelming to say the least.

Sure, it’s fast but these queries are mostly branded, not super descriptive and I don’t think go into the amount of detail that an SEO would need in order to get started. This means that using ChatGPT for proper query research is going to require a bit more advanced functionality.

Improve ChatGPT Keyword Research With “Deep Research”

Honestly, ChatGPT’s Deep Research might be my one of my favorite functionalities of the platform. You can use it for so many things including analysis of your site, competitors, general web research and so much more. By using Deep Research, you’ll allow ChatGPT to actually crawl through a large portion of your site and identify key opportunities.

For a long time, one of my favorite ways to find deeper query opportunities for clients was to analyze landing page copy. This can give you insights into deeper/lower funnel ideas based on actual copy that a site is using. For instance, for Domo we can see their “Governance” page contains a few interesting ideas including “Self Service Analytics” and “Embedded Analytics”:

These are the types of opportunities that ChatGPT’s Deep Research is excellent at identifying at scale.

How To Use ChatGPT For Keyword Research

Now, let’s start the process by using ChatGPT’s Deep Research. Open up a new chat and click the “+” dropdown and you’ll see the “Deep research” icon to select.

This will allow you ingest A LOT more data into ChatGPT then you were before. Now you can configure your instructions to ChatGPT in order for it to get started. In this example, I used:

I want to do keyword research for this site. Please go through their home page, key product and solution pages and blog articles and more. Along the way, identify potential keywords that might be important for SEO.

At the end, provide a big list of keywords you found. Give me as many as you can. Format them in a big list separated by commas: https://www.domo.com/

ChatGPT will then begin to go through all the different pages of your site, analyzing the terminology that the site is using along the way. You can see the exact steps that it’s taking in the “Activity” section:

 

Now, go grab and coffee because this will take a bit longer than your basic single-shot prompt. However, in the span of just 10 minutes, ChatGPT is going to give you some MUCH improved and more varied keyword opportunities to consider. From this prompt, here is the final output that it gave me.

 

You can see that it gives a lot more potential ideas that are actually based on what the site is talking about. This includes topical areas that we might have missed before such as “Embedded Analytics”, “Collaborative Analytics”, “Self Service Business Intelligence” and more. This is already much improved from what our starting list was.

In another example, I actually asked it for 300+ keywords. Deep Research ended up actually giving more significantly more query opportunities and clustered them into different topical areas based on the business function for Domo.

Of course, no we need to verify that these keywords are actually searched by real people. The next step is to validate the list through an SEO tool. I’m using Ahrefs, but you can use the tool of your choice. I’m simply going to take this list and plug these terms into the “Keyword Explorer” feature.

From there Ahrefs will at scale analyze whether or not this is something that users are actually searching for or not. It’s an extremely effective way of validating the data at scale:

 

If you wanted to take it a step further, you could also have ChatGPT’s Deep Research analyze your key competitors as well. For Domo, if we’re interested in learning about Qlick’s SEO strategy, we could have Deep Research go investigate their site and core target queries.

 

Once again, this will allow you to ingest a lot more data into ChatGPT based on actual search terms they’re using in their content.

Hopefully this process is helpful for everyone! I think it goes to show that ChatGPT can be an extremely valuable for keyword research when used properly. The key here is to use ChatGPT to analyze and ingest data at scale and using the model to identify search terms that already exist.