Collaborating with AI for SEO

How to collaborate with AI to improve your SEO game 

Practical examples for Technical SEO, Content and Digital PR 

Now the dust has settled on the dawn of the AI and the ChatGPT revolution, it’s time to share the more practical side of what we’ve learned after almost a year of testing.  

We’ve talked previously on how AI can help drive efficiency – from AI as part of content creation to the impact of AI on digital PR – but what we’ve not covered as much is how to really put this into practice. 

The key thing to keep in mind before we go any further is collaboration. The focus should be on harnessing the power of this tool to help you in your day-to-day – not replace your day-to-day.  

This resonates even more following the latest Helpful Content update from Google, which finished rolling out at the end of September. Firstly, let’s acknowledge that Google did update the Helpful Content guidelines from “helpful content written by people, for people” to just “helpful content for people”. But that doesn’t necessarily mean it’s time for an AI free-for-all.  

While we’re still figuring out the impact of this update, some SEOs – such as Marie Haynes – believe the update is denoting sites with content “lacking in real-world experience”. And that is something AI can never truly provide. 

That said, there are many areas prime for collaboration, so we’ve asked our heads of Technical SEO, Content and Digital PR to give us the lowdown on key areas to consider. 

Technical SEO & AI: Practical examples 

Whether it’s spotting the patterns behind broken links, redirect chains or why page templates aren’t being crawled and indexed properly; Technical SEO can often boil down to dealing with lots of data and spotting outliers within that data. This is where tools like ChatGPT can be your friend.  

Being based on a Large Language Model (LLMs), it has been trained on millions of data points, so excels at pattern recognition. This can help speed up Technical SEO tasks and make more efficient use of your time, focusing on areas where AI isn’t quite up to scratch. Here are some of the Technical SEO tasks we have been using it for: 

Data Filtering 

When searching through crawl data, Google Search Console or trying to compile reports within Google Analytics or Looker Studio, we often find ourselves wanting to examine only a particular subset of data, such as pages or queries. Unfortunately, this data rarely sits neatly into an easily filterable or searchable structure and thus requires some more complex data wrangling.  

One useful feature of ChatGPT is that you can feed it a list of URLs or queries and ask it to create the Regular Expression (or regex), which would help uncover these terms. It’s important to note that different tools use different forms of regex, so you’ll need to clarify which tool you’re using to get the desired results.   

Data Clean Up 

Another great use for ChatGPT has been the ability to ask it for help cleaning up data. Take some ranking data which still includes fragment URLs or parameters, and which also includes some cases of ‘no rankings’ or some Core Web Vital data which still includes the ‘s’ or ‘ms’ after the score. Often these small inconsistencies prevent you from completing even simple arithmetic on these datasets, never mind anything more complex. In the past, data cleaning often required several different formulas, filtering data to find outliers which became time-consuming when dealing with hundreds of thousands or rows, but again ChatGPT’s Advanced Data Analysis can spot irregularities and suggest fixes after an initial scan of the data.  

Of course, you shouldn’t be inputting any sensitive data into these tools, but you can ask it to produce dummy data in a particular format or give it an idea of the data structure you are using and ask it to produce the code it would use to analyse the data. Then as ChatGPT also outputs the code it used, you can use a Python notebook to copy the code and run it. By saving it in a separate notebook it also allows for the further automation of processes which are repeated on a regular basis, such as regular crawls or monthly reports.  

Technical Jargon 

Another use case for ChatGPT is in transforming technically complex topics and notes into simplified formats. Often, I find myself reading through documentation for a particular CMS platform, plugin or developer notes, which use lots of acronyms, jargon or assume a lot of prior knowledge. I’m guilty of this myself, where my initial draft technical SEO audit is written in short and which isn’t always the most legible for the end-user. I then spend a lot of time reworking these documents to reduce jargon, explain acronyms, or tailor talking points based on the audience.  

Using ChatGPT, I save time transforming these notes into formats suitable for a non-technical user, a developer or another SEO, with each audience having their own concerns and areas of expertise. It also allows me to rewrite my notes in different formats depending on the client's request, for example, presentations, documentation, emails, user issues or JIRA Tickets. This is useful for meeting notes where lots of technical jargon or acronyms are used, as you can rewrite the notes in the format that best caters to all participants. 

Useful First Drafts 

ChatGPT has hundreds of use cases which could be explored, but we’ve found it best to treat the responses it produces as a drafting tool rather than the finished product. When testing the use of ChatGPT for generating robots.txt rules, htaccess rules, hreflang, XML sitemaps and JSON-LD schema markup, we’ve found that it required careful prompt engineering and custom instructions to get reliable, reproducible results. 

There are still issues with hallucination which result in additional data being included, such as creating a fake author when none exists for schema markup. So, while a definite time saver for generating draft templates, we’ve found that results applied across hundreds or thousands of pages result in small inconsistencies that have needed to be double, and triple checked, resulting in little overall time saved. 

In general, we’ve found using AI tools for Technical SEO tasks has been extremely useful in our day-to-day work and has helped us elevate our work by helping to automate or reduce the time spent on some of the more tedious tasks. 

Content & AI: Practical examples 

Creating meta data at scale 

If you’ve got a number of page titles and meta descriptions to update, GPT for Sheets is your friend – it’s all the power of AI in a handy table format that’s easy to copy into multiple rows. Here’s what to do in 5 easy steps: 

1 | Download and install the GPT for Sheets Google Chrome extension: Download it here. 

2 | Set up a new Google Sheet with the following columns and formulas:

• URL: Simply enter the URL of the page you need the meta data for. ChatGPT will use the URL as an indicator of the page’s context. If you have more time you might want to add a brief summary into cell A4 instead to use this as inspiration for ChatGPT. 

• Page title and meta description prompt: This is the instruction you’re giving ChatGPT. The example above is probably the bare minimum you want in there for page titles and meta descriptions. This is a great time to get creative, so consider adding tone-of-voice recommendations to your prompt for a more refined result. 

• Output: This is the page title output you’ll get from your formula. 

• Count: While we’ve asked ChatGPT to stick to a specific character count, we’d still recommend adding this column in to make sure it’s within the limit you’ve requested. 

3 | When you’re happy with your prompts and initial outputs, simply drag down to copy across as many URLs as you need. 

4 | Please take the time to QA the outputs – while we can rely on AI for a first draft, it takes a real-world touch to ensure the copy is clear, concise and will encourage those all-important click-throughs. 

5 | Lastly – once you’re happy with everything, get it implemented on site. 

For more tips, there is guidance at GPT for Sheets. 

Prompt building for a better output 

You’ve probably heard a lot about prompt engineering and ChatGPT – it’s true that the more refined your prompt is, the better result you’ll get. If you need a page outline, or potentially an initial copy draft, and you’re unsure of how to prompt ChatGPT to get the best output, the ‘prompt builder’ can come in handy. It essentially turns ChatGPT into your prompt engineering assistant.  

To start, simply copy and paste this directly into ChatGPT: 

I want you to become my Expert Prompt Creator. Your goal is to help me craft the best possible prompt for my needs. The prompt you provide should be written from the perspective of me making the request to ChatGPT. Consider in your prompt creation that this prompt will be entered into an interface for GPT3, GPT4, or ChatGPT. The prompt will include instructions to write the output using my communication style. The process is as follows:  

1. You will generate the following sections: 

 " Prompt: >{provide the best possible prompt according to my request} > > >{summarise my prior messages to you and provide them as examples of my communication style}  

Critique: {provide a concise paragraph on how to improve the prompt. Be very critical in your response. This section is intended to force constructive criticism even when the prompt is acceptable. Any assumptions and or issues should be included}  

Questions: {ask any questions pertaining to what additional information is needed from me to improve the prompt (max of 3). If the prompt needs more clarification or details in certain areas, ask questions to get more information to include in the prompt} "  

2. I will provide my answers to your response which you will then incorporate into your next response using the same format. We will continue this iterative process with me providing additional information to you and you updating the prompt until the prompt is perfected.  

Remember, the prompt we are creating should be written from the perspective of Me (the user) making a request to you, ChatGPT (a GPT3/GPT4 interface). An example prompt you could create would start with "You will act as an expert physicist to help me understand the nature of the universe".  

Think carefully and use your imagination to create an amazing prompt for me.  

Your first response should only be a greeting and ask what the prompt should be about. 

It’ll ask you a series of questions, and the answers will help form your tailored prompt. When you’re happy with the prompt, you simply tell the tool you’re ready to use the prompt. Once you get the result, you can review and refine it. It’s as easy as that!  

Research and fact-checking 

If you’re sourcing information for a piece of content you’re putting together and need authoritative sources, give Perplexity.ai a try. 

This was something seriously lacking with ChatGPT, and while it was announced recently by OpenAI that ChatGPT can now “provide you with current and authoritative information, complete with direct links to sources”, in the tests we’ve carried out, it hasn’t yet delivered on this front.  

For fact-checking, we prefer to use Perplexity. It works in a similar way to ChatGPT; simply ask it a question, and it’ll provide a detailed answer, along with a range of sources you can use to back up the information.  

Here’s an example of how it works: 

Digital PR & AI: Practical examples 

Imagination and automation 

AI platforms such as ChatGPT can be an incredibly useful starting point when it comes to generating campaign ideas for brands, but are currently just that – a starting point. 

For now, at least, AI isn’t actually creative. It doesn't experience emotions or understand the cultural nuances key to creating the perfect PR campaign. Nor is it capable of any of the unique eureka moments that often define human creativity. 

That said, it can certainly help facilitate creativity when it comes to ideation by giving ideas, topics and themes that might potentially spark a killer idea – it just can’t do it all for you.  

Here’s an example of a prompt you can try for travel ideas: 

Some okay ideas in there – but it’s fair to say most of the above lack the originality and creativity required for a campaign to cut through the noise. However, most importantly, the prompt has given us topic ideas and potential angles to explore, which could lead to some brilliant campaign ideas.  

An expert assistant, but not an “expert” 

Copywriting is perhaps one of the strongest use cases for advanced language models like ChatGPT and Microsoft’s Bing. With the right prompt, it can be a useful tool for generating workable first drafts and helping to overcome stubborn writer’s block.  

One such way that the copywriting capabilities of AI can be utilised for Digital PR is in assisting with the drafting of expert commentary for reactive PR opportunities. 

It’s at this point where an important caveat needs to be made, however. AI isn’t a substitute for actual expertise. And should never be used in place of an actual expert when drafting comments.  

What AI can help with is in the surmising and compiling of points made by the experts from brands that you work with, providing strong first drafts of quotes with the right prompts. Here’s how. 

Imagine you’re responding to a request from a journalist for a quote on why young adults should never opt out of their pension. You already have insights from your expert spokesperson around this subject and have examples of existing quotes from that spokesperson, so you can capture your client’s tone of voice.  

You can then input the following prompt:   

Here’s the result: 

Not the finished article by any stretch, but not a bad starting point for a first draft. From here, we can manually review and edit to ensure our ToV is correct, add important key messages or product mentions and remove some of the ‘fluff’ ChatGPT likes to add to copy! 

As the above example demonstrates, AI can be an incredibly useful tool for first drafts, combining points and speeding up the writing process. But from a PR perspective, it will always require manual review and input to ensure brand messaging is as tight as possible and copy always has the target audience in mind. 

Our approach to AI and SEO  

The evolution of tools like ChatGPT might have changed the way we approach SEO forever. But it’ll never replace the real-world experience of SEOs, content creators and digital PRs. It’s simply a new tool to leverage for SEO – whether it’s analysis, tactics, or execution.  

To find out more about how we can help up your SEO game, get in touch.