How to use AI for SEO - 5 methods
Connect on Linkedin
Create content with AI
Before you run, I want to be clear that I am not advocating a way of working where AI content is published directly. Humans still want to share other people's opinions and thoughts, not an AI's, nothing has changed here.
In addition, Google has been very clear in its E-E-A-T guidelines that first-hand experience, which a language model does not possess, is something that is rewarded.
That said, AI can still be a great resource here, so let me go into how I work with copywriters and AI to create content that ranks.
I start by doing research on the keywords I want to rank for. I do a SERP analysis, evaluate its search intent, and look at the competition.
Once I have a good idea of the type of content I want to create, I ask the AI to create a structure for the article with headings and subheadings based on the keywords I have selected. This always requires some manual actions to get it the way I want it.
Once the structure is ready, I use AI to generate content for each section. Tools like the AI story ending generator can support creative writing tasks, but the content typically needs human refinement to match the brand's voice. To achieve this, you can input existing brand content and have the AI adapt it to maintain a consistent tone.
Now I send this first draft to the copywriter. The structure is really the important thing here, the actual body content should serve as inspiration. It happens that the body content needs to be completely recreated, but it often makes it much easier for the copywriter to get started. By having something to work from, you minimize the risk of "writer's block".
Bulk create titles and meta descriptions
Regarding titles and meta descriptions, I personally think you can let the AI go a little more free. We're talking about 60 and 160 characters to work with, respectively, which means that aspects such as first-hand experience, tone of voice, expertise, and credibility can't take up as much space.
If you read Google's AI content guidelines, they mention that automated content doesn't necessarily have to be bad. I think titles and meta descriptions are a good example of where this comes into play.
Since the space is so limited, titles and meta descriptions are often very information-dense, which means that they lose some of the human touch.
Since character limit is so tight, titles and meta descriptions are often very information-dense, which means that they lose some of the human touch.
So how do you get started?
Start by listing all the URLs you want to create meta data for in a spreadsheet. Tools like Screaming Frog make this very easy, but you can also export from Google Search Console, which you can then filter.
After that, you can either download a plugin like GPT for Sheets that lets you prompt directly in the spreadsheet, or you can choose to go through the interface of your favorite AI.
In this example, I have chosen to do this for the sneaker e-commerce site Footlocker, directly in the interface. Note: I have no affiliation with Footlocker, this is just an example.
Customize the following prompt according to the brand you are working with and individual preferences:
The AI will now confirm what you have asked it to do. If anything has been misinterpreted, you have the opportunity to steer it in the right direction. If everything looks good, you can now give it a few URLs.
You will then receive a table with URL, title, and meta description. See this as a first draft that you have the opportunity to give feedback on. There may be word choices or phrases that you don't feel quite align with the brand.
Point this out and, if possible, give suggestions on how you would have adapted it yourself. You are now training the AI to be able to do the job flawlessly later on.
After a couple of rounds of proofreading, I got back the following, which I was satisfied with:
Structured Data
As many agree, the rise of AI has been revolutionary in programming. As an SEO, it is rare that you need to code yourself, but there are many areas within technical SEO that require a basic understanding of the most common programming languages.
Debugging and extracting structured data is an area that for me personally has become incredibly more efficient since I started using ChatGPT.
Previously, I used online structured data generators, which is also a big step up from coding myself, but ChatGPT has made it even more efficient.
Debugging structured data example
Let's use another page from Footlocker as an example of how to use ChatGPT to troubleshoot incorrect structured data.
Let's say you have used Google's Rich Results Test and it has confirmed that your page has problems with product reviews.
The tool flags that the mandatory field "ItemReviewed" is missing.
Step 1 is to locate the markup in the source code and provide it to your AI tool. Describe the problem and ask how to solve it.
My prompt was:
The schema markup for my product page is getting error messages due for "Missing field 'ItemReviewed'" Can you analyze the code and tell me how to correct it to fix the issue?
This is what I got back:
Okay, now we know the problem and which values are missing. Now we also want to solve it, so I will explain which value I want in the different fields. The prompt was:
Here is the prompt I used in ChatGPT:
Can you update the structured data accordingly so I have the code finished for use?
Name: adidas Basketball Tee
Description: Elevate your everyday style with the adidas One BB T-Shirt. Its smooth fabric feels soft against the skin and keeps you cozy and comfortable throughout the day. Look your best with the adidas One BB T-Shirt.
SKU: 317338718440
After that I received a new markup that put back into Rich Results Test for validation and all of a sudden!
Generating ALT text to images
The function of ALT text is to inform search engines and users with screen readers about the content of an image. First-hand experience doesn't make a difference here, so I don't see the problem with letting a language model handle this monotonous task.
For this, I use two different tools:
SheetGPT (Chrome addon) to integrate ChatGPT with Google Sheets.
Screaming Frog to get access to a page's images.
Start by opening Screaming Frog and crawl the page containing the images you want to create ALT text for. If it doesn't apply to any specific page, you can choose to crawl the domain and then export the image URLs you want to generate ALT text for.
Next step is to paste the image URLs into a sheet.
Column A (Text) This column contains the first part of the prompt. It is what I want the AI to assist with without referring to any image. In this column, you enter the following: Generate an ALT text based on SEO best practices to the following image:
Column B (Image) This column contains the image URLs.
Column C (Prompt) This column uses the Concatenate formula to combine the text from column A and the image in column B. The final result is a complete prompt that refers to an image. If your document is structured like mine, the formula is: =CONCATENATE(A2,B2)
Column D (Output) Finally, you put the formula =GPT(C2) in column D, which returns the ALT text as instructed.
Data analysis
A large part of an SEO's work involves interpreting data in order to draw conclusions and then take action. When it comes to interpreting and drawing conclusions, AI can be an incredibly powerful tool.
One thing I like to do is export reports from tools like Google Search Console, Google Analytics, and SEMrush and then ask the AI for insights.
When you work in these tools on a daily basis, it's easy to become blind to the obvious, and there may be problems or potential right in front of you that you don't notice. By asking ChatGPT, you get a new perspective that often leads to new insights.
Data analysis example
Start by going to Google Search Console, select the date range to the last 3 months compared to the previous period or year.
I like to take a little longer periods to get more data and thus be able to draw reliable conclusions.
After that, you choose to mark in the respective metric: Clicks, Impressions, CTR and Average Position and then export as a CSV file.
You should now have six different files with clicks, impression, CTR and average position, divided into the following dimensions:
Now give the file you want to start analyzing to ChatGPT. In this example, I start with Countries.csv which is dimensioned by country with the prompt: I'll provide you with a few Google Search Console exports. I want you to analyse the files and provide useful insights. Take the role of an SEO analyst.
ChatGPT will now analyze the file and then provide you with insights. Here is an excerpt from what I got back
You’ll now decide if you feel satisfied with the insights, follow up manually or ask follow-up questions. If you want to look at it from another dimension, you can always add another export and ask questions related to it.
As an example, I could have uploaded the Queries report with the prompt: which Swedish search terms have decreased the most in clicks?
Summary
In conclusion, there are many opportunities for SEOs to streamline their workflows with the help of AI. In this article, I have covered five different applications, but there are of course many more possibilities.
Despite this, one should still maintain a certain skepticism about the technology and not rely on it completely. AI has a tendency to hallucinate and produce information that sounds correct but is actually not correct at all.
Ultimately, it is up to the individual to ensure that the content remains accurate, relevant, and valuable to its audience. By using AI as a tool rather than a replacement, one can deliver results that are not only effective but also retain their human touch.
