Beautifully Merging SEO with LLMs & AI

Subscribe to my newsletter for more insights from industry experts: Subscribe

In this issue, I'm chatting with Caitlin Hathaway , who loves data, AI and SEO automations.

We talked about using LLMs and AI in SEO, the best and worst ways to use LLMs for SEO and everything in between.

So, let's see what she has to say!

I’m gonna start with my standard first question: How did you get into SEO?

After graduating from university, I began my career in the less traditional route - as a sales rep at an enterprise SEO software company.

I had the rare opportunity to speak to established in-house SEO teams every single day as part of my role, which was hugely insightful into the type of work and projects involved in an SEO career.

Safe to say, it wasn’t long before I wanted to become an SEO myself!

Fortunately, this experience helped me transition to a role as an SEO specialist at an agency, where I could focus on working with clients within multiple industries and honing my skills in the field.  

Eventually, I transitioned to an in-house role, focusing on managing SEO for a single company.

Currently, I work at MVF, and I am now a part of the Audience and SEO team.

My current role is similar to working in an agency since we manage multiple websites across different verticals such as software and business services.

I get to keep one major benefit of being in-house, as these websites are within our own MVF portfolio. 

What are some key LLMs you use daily as part of your work?

I’m a big fan of the OpenAI API as I’ve been able to integrate GPT directly into Google Sheets to use on my workflows for larger projects.

I’ve relied on this method for content projects such as grouping content and tagging themes for an entire site. 

I frequently use ChatGPT to create my own solutions through GPTs, which has been super convenient in my workflow.

There are lots of nuanced and complex tasks involved in SEO, particularly with content quality, and understanding how content adheres to Google’s E-E-A-T guidelines - which can be tricky to measure. 

Some of them I include as part of my daily workflow include:

Entity Analyst GPT which performs entity analysis on search queries (identifying nuances of queries, entity names, types, and the relationships between these entities) using Google’s patent ‘Methods, systems, and media for interpreting entities in search queries‘.

I’ve built my own content briefing GPT solutions to help create much higher-quality content briefs. It has all of the important information I need baked in, saves tons of time, and makes it more clear and helpful for writers.

Search Quality Rater GPT: Brings the official Google Search Rater Guidelines to life in GPT form. The GPT can evaluate your content as if it were a search quality rater. This has been really helpful for page-level audits. 

Separate to GPT, I’m exploring more of Google Cloud APIs such as Vision AI to build more of my own solutions for some use cases I have. I will definitely share more soon about this.

What is the most underrated or, let’s say untapped way to use an LLM for SEO and digital marketing in your opinion?

One untapped way I’ve been using AI is to get ahead of the curve and understand future developments on the horizon in my field.

I schedule time each month to analyse Google's newly submitted patents. I've found some interesting developments about potential changes to Google Search by having an LLM summarise and explain these patents in simple language.

This is a recent Search-related discovery I’ve found: 

Radar Recognition-Aided Search – to understand not just verbal search queries but also non-verbal cues like gestures or physiological states to refine search outcomes.

This is how it looks in action:

Gesture Recognition: As you speak, you also naturally point towards the end of the street you’re heading towards. The smartwatch, equipped with radar-based gesture recognition, detects this pointing gesture. The system understands that you’re not just looking for any café but specifically one in the direction you’re pointing.

Physiological Data Analysis: Simultaneously, the smartwatch monitors your physiological data and notices an elevated heart rate, possibly indicating that you haven’t had your morning coffee yet and are in a hurry. It might also recognise the slight tremor in your hand gesture, further emphasising your urgency.

Refining Search Results: Using this combined information—the verbal query “nearest café,” the direction you’re pointing in, and your physiological state indicating urgency—the system refines the search results. Instead of showing a list of all nearby cafés, it prioritizes one that is directly ahead on your path and known for quick service.

Your smartwatch then tells you, “There’s a QuickCup café about 200 meters straight ahead. They’re fast, and you’ll have your coffee in no time!”

If you’re interested in analysing patents yourself, here are some resources:

• My Patent Simplifier GPT (needs ChatGPT Pro subscription) See Google’s submitted patents, sorted by Newest and US (or find your patent of choice)
• If PDF can’t read text, run through OCR enabled PDF to text converter

This may “sound” weird, but: What is the worst way to use LLMs for SEO?

In my opinion, the worst way to use LLMs for SEO is to completely outsource content generation to these models without any human oversight or review. 

The introduction of LLMs such as ChatGPT and Claude have undoubtedly made it easier to generate content quickly.

In the last year, I’ve been hearing more about businesses who have downsized their content teams or agencies because they’ve opted to save costs by completely outsourcing content to LLMs… with little to no human review.

As we're all seeing in the SEO industry, the web has been flooded with unreviewed AI-generated content. 

Entirely offsetting your content processes to AI can be harmful, as the content can be factually incorrect, misleading, and plagiarised from original sources. 

At the current time of writing, AI content has been contributing to poorer quality of search results.

Search engines haven’t developed sophisticated enough systems to decipher between AI content and authentic content written by humans.

Now, unfortunately, lots of spammy, irrelevant content is ranking in top positions.

A few specific tactics I’ve seen include automating ChatGPT generations onto sites, answering hundreds of PAAs using the OpenAI API and publishing onto sites as new pages with minimal review.

A recent evolution is due to the huge prioritisation of Reddit in search for commercial queries right now, Reddit discussions are now becoming artificial discussions with fake users, using AI-generated content as answers, and pushing affiliate links.

Google is now ranking those in prime positions - some even in YMYL queries. 

While Google says it's been better at identifying and tackling AI-generated content, it's still got a long, long way to go. 

While LLMs can be valuable tools to assist with content creation, human review and editorial oversight should always be baked into the process.

LLMs are not built for accuracy and make factual errors, engage in plagiarism, or produce biased content. 

Human editors should carefully review and validate the outputs before publishing anything using AI to assist to ensure accuracy, originality, and relevance. And to make the web a less worse, AI-ridden place.

Apart from prompting, what other things or skills are required to get the maximum benefit of LLMs as an SEO?

Understanding the underlying technology of what an LLM is and how they work will help you understand how to use them a lot better. You’ll know their capabilities, use cases and current limitations. 

In my experience, different LLMs excel at different tasks, so it's beneficial to familiarise yourself with several various models and any new ones you see. 

For instance, OpenAI / ChatGPT has been better for coding and data analysis tasks (when well prompted).

I’ve gained so much value from using ChatGPT to upskill on scripting, formulas, as well as learn programming languages including SQL and Python, to help me build in-house monitoring systems, automate processes, and build my own custom tools.

I recommend learning how to connect LLM APIs to other tools and sources, such as Google Sheets or within GPTs. While this may require some initial upskilling and setup effort, it will make you far more productive and level up your workflow.

What is an SEO task/process that you completely automated with AI?

I love exploring how AI can be used alongside my current processes, in an SEO workflow and beyond.

I always test new tools or ways of working to see how I can automate parts of processes or entire processes

I’ve built monitoring systems for all of our content with the help of LLMs, which has helped us spot opportunities and issues more quickly, and work more proactively across sites. 

I’ve found unique and helpful ways to extract valuable insights into areas such as audience interests, sourcing valuable communities your audience resides in, their content preferences, and surfacing trending topics to write about.

One of my favourite use cases was using AI to help build a content inventory.

AI helped me mark it up according to content themes, understand what the best-performing themes were by session and conversion-wide with your audience. 

AI has also reduced the time it took me to conduct a content gap analysis and give me insights into the type of content that a site is winning in that particular space, and should be doubling down on.

I’ve written about these in detail in my guide to building a content strategy using LLMs.

I’ve also written up on some of the processes improved and the challenges our team faced while working in a media business.

We used AI to support us with building solutions, such as automating auditing of content SEO issues, and better algorithm impact tracking systems. You can find the article on Sitebulb!

AI is evolving fast. Which is the best way to not lose track of its capabilities and stay up to date?

I highly recommend scheduling a small amount of time in your week to read up on the latest AI news and developments. It really will go a long way!

My favourite newsletter is The Rundown.ai - it does a brilliant job of summarising the AI news and developments of the day into an easy-to-read newsletter.

I usually catch up with these during my commute time or lunch break as it doesn’t take long at all.

I also find a lot of value in the Trending AI Tools section within the newsletter.

If there are any tools of interest, I schedule further time on the weekend to try them and see if they can somehow fit into my processes. 

What do you think is the future of SEO?

I don’t think traditional search as we know it is going anywhere, but there are certainly some emerging trends and changes to be aware of.

While people have been complaining about the quality of search results for some time now, especially with Google's recent core updates, it's important to remember that Google still dominates the search engine market share so we must still remain Google-first.

However, Google’s search engine market share has declined slightly over the last few years as users, particularly younger generations, are finding new ways to search and consume content.

We’re seeing more trends of generations finding new ways to search using different platforms due to wanting to consume more visual content, as well as the diminishing quality of search engines.

One notable trend is the rise of AI-driven search engines like Perplexity.ai, which offer a conversational search experience backed by citations from reputable sources.

I’d be interested to see the adoption of AI-powered search tools and if they gain more traction in the short-term, particularly for certain types of queries.

Additionally, platforms like YouTube and TikTok continue to grow in popularity, as users increasingly turn to visual content and seek a more human-centric experience.

Moving forward, I’d look to diversify traffic sources and focus on engaging their audiences authentically across multiple platforms, rather than relying solely on traditional search engines like Google which has proven to be a rather risky approach.

In the far future, a current or new Google competitor could potentially integrate various elements like AI search, visual content, and a human-centric approach to win a large amount of market share.

However, I believe the future will be more distributed among many ways to consume content rather than through one dominant engine.

What would you advise SEOs and Content Marketers who want to start using LLMs in their workflows?

My advice remains the same, whether you’re a student or an established professional - you’ll learn the most by just getting started!

A lot of people are quite hesitant to use LLMs because they look and sound confusing - but I promise you they’re not :)

If you’re using ChatGPT, Gemini or Claude, start by asking simple requests (prompting) about tasks in your work, and seeing the outputs that each gives you.

You’ll soon learn how to prompt more effectively and know how to request (or argue with GPT) to get exactly what you’re looking for.

Some of these are good resources to learn how to prompt better:

Text prompting:

• LearnPrompting.org - a huge resource on how to prompt, broken down into difficulty levels.
• PromptGenie - use a free trial. Enter a prompt, and it will help you craft more detailed prompts. You can use this to improve your prompting.

Image prompting: 

• Midjourney: just join the Discord channel and see how people prompt on channels. You’ll learn how they’ve generated their outputs with their exact prompts and models they’ve used, all for free.
• Lexica - a huge library of AI-generated art, complete with prompts. It’s brilliant for image generation ideas. Look at other images you like the look of and see how the artists prompted.

You can then start looking at how to integrate AI as part of a current manual process.

I ask myself these questions when looking at how I can optimise a task or a current workflow:

• What is my most manual task?
• How long am I spending on this task?
• Can AI possibly help me to automate one part of the task, or even completely automate it?
• What’s stopping me from experimenting and building my own solution?

I discussed this in my interview with Charles Floate, but I wholeheartedly believe that marketers who are actively exploring and experimenting with LLMs and AI tools will gain a significant advantage compared to the many who don’t.

By getting hands-on experience, you’ll hugely expand your understanding of how to smartly use AI in your workflows and discover new ways of streamlining processes and increasing efficiency.

From what I’ve seen, the SEO industry can be set in its ways and reliant on mainstream tools, but AI opens up creative avenues for accomplishing work more effectively, which will set you apart.

Even taking out free trials on tools can prove valuable, especially if you save time and effort compared to more manual methods. 

If you like a particular tool, consider paying for a pro plan or API credits. View it as an investment in yourself and your development!

Some platforms such as ChatGPT, and Claude are cost-effective and you’ll get access to features that could take your prompting to the next level, including Web-browsing and GPTs.

I definitely recommend looking at those as starting points! 

OpenAI credits are a little more advanced but there are amazing ways you can integrate it into your workflow.

If you’re looking to add OpenAI to Google Sheets, I recommend this video by Tech With Hitch to help you get started through Apps Script.