Everyone’s talking about AI and how it “continues to evolve.” What they mean is that sometimes AI messes up their online strategies, and sometimes it improves them, and, as it hasn’t settled down yet in predictable patterns, we don’t always know what the outcome will be when we use it.
This is great, because it gives us all something new to dig into and test and hypothesize. In this post, we’re going to discuss:
Table of Contents
(TL;DR: One, we don’t recommend depending solely on generative AI for anything, and especially nothing longer than a paragraph. Even then, things get iffy. We need people. Two, different tools do different things well. Try them out, take advantage of free trials, and then settle down with one or two that have “potential,” as far as you’re concerned. Three, pay attention to commentary from those for and against generative AI as you decide what role you want AI playing in your company.)
How Generative AI and LLMs Work Together
What is an LLM and how does it work?
LLMs, or Large Language Models, are a type of AI program that can generate text similar to humans. GPT is an LLM.
These models are trained on enormously large amounts of text and text-like data, such as code, books, and articles. ChatGPT was trained on about 570 GB of data. To put that into perspective, one gigabyte can contain about 178 million words. That means ChatGPT was trained on around 101.5 trillion words. To compare, BARD was trained on 1.56 trillion words.
How do they work? First, they ingest data. Then, the data is passed through a transformer, which allows the LLM to understand how every word relates to every other. Finally, the programs are trained. This training is when they learn to predict the words that should come next in any given sentence and generate reliable sentences.
LLMs are used by AI chatbots to generate content, translate languages, converse with humans, summarize content, etc.
How Can Businesses Use AI for SEO Today?
Generative AI isn’t something to fear BUT nor is it something to use indiscriminately. It can help you create content, but it can’t take over the job on its own. You still need humans in place making your content relevant and authentic, whether you need metadata, ad copy, or blog content.
Tools are good at improving your workflows by helping you outline, brainstorm, and get a general jumping-off point for content creation. Let’s take a look at four of the options on the market.
Natural language processing AI model designed to understand and generate human-like language.
ChatGPT has many use cases. For the purpose of this post, we used it to:
- Brainstorm ideas
- Rewrite/format emails
- Summarize information
How It Works
Open ChatGPT and type whatever you want in the message box.
After the tool’s initial response, you can provide follow-up directions. Such as:
- Rewrite with shorter words
- Make it more informal
- Rewrite with bullet points
- Write five different closing options
The more you work with ChatGPT, the more you learn how to best direct it, and the more it learns what you want.
Pros of ChatGPT
In our initial testing, we found ChatGPT to be good at:
- Generating human-like content
- Generating ideas for brainstorming
Cons of ChatGPT
- Unintentionally biased at times because it is trained on a large set of text that contains biases or prejudices
- Content is still robotic and needs refining to make it sound more natural (aka more human)
- Can summarize text, but not offer insights
- Trouble generating long-form, structured content
Content creation: blogs, articles, outlines, etc.
We tested Rytr on the following:
- Blog idea & outline
- Wikipedia test (using the blog writing use case)
- SEO meta title
- SEO meta description
How It Works
Rytr offers different use cases. You select one, and then provide the other information requested, which will change based on the use case.
Once you receive Rytr’s variants, you can edit them with options such as ‘expand’ and ‘shorten.’
Pros of Rytr
While not perfect (refining was definitely needed), we found Rytr to be a good starting point for
- Blog outlines
- Meta descriptions
Cons of Rytr
On-page optimization SEO software
SurferSEO is designed to help you optimize your content. We tested:
- Surfer AI (to generate articles)
- Content editor
How It Works – Content Editor
You can copy+paste content into Surfer, or write from scratch within the tool. You provide guidelines and keywords. It then “reads” your text and gives you tips on how to improve SEO.
How It Works – Audit
Use this feature to audit published content based on different keywords that you provide.
How it works – Content Editor + AI
For the best draft, you will want to:
- Perform in-depth keyword research – the better, more relevant your keywords, the better Surfer AI will understand what you want
- Choose the right competitors – again, this helps Surfer understand what you want it to achieve
- Boost Anti-AI detection – this might result in less natural language, but it is less likely to be detected as AI-written – and since you need to edit all content to have your brand voice and authenticity, you would be editing language anyway
- Review Surfer’s outline and edit/add/subtract headings
- Review + edit the written article
Pros of Surfer SEO
Surfer is useful if you want to:
- Identify keywords and internal linking gaps
- Have an AI editor review your content and give suggestions as to how to improve SEO
- Write a draft of an article that you only need to edit
Cons of Surfer SEO
- If you are writing in a niche market, Surfer might not be able to pull enough related content to understand your keywords and content needs
Natural language processing AI model designed to understand and generate human-like language
We tested Google Bard on the following:
- Blog content with images
- Rewriting/formatting emails
- Summarizing information
How it works
Much like ChatGPT, Bard opens with a message box and you type in your directions or request. The tool then goes to work.
Pros of Google Bard
Check out Bard for the following:
- Generating human-like content
- Basic outlines
- Sentiment analysis – Bard will analyze text and determine if the sentiment is positive, negative, or neutral
- Up-to-date – Bard is connected to the internet and so has access to the latest information
Cons of Google Bard
- Wouldn’t provide metadata
- Unreliable data – it can pull data that is not factual from the internet
- Creative limitations – repetitive and unoriginal content; look out for keyword stuffing
- Biased content
- Not good at creating long form, structured content
Challenges and Benefits - How AI Will Affect Search Results
Companies want to use AI to improve their content and SEO. Users, whether they know it or not, depend on AI to deliver the best answers to their search queries. When companies understand how AI is working through search engines, they can better create content that will play nicely with AI tools.
AI has been a part of search results for awhile. Its main job is to read the searcher’s mind. When someone types “Apple,” into Google, what do they likely want to know? Are they looking for information on the company, or do they want pictures of the fruit? AI makes an intelligent guess by analyzing the language used, the context of the query, and previous search behavior. It takes everything it knows and then provides (hopefully) more relevant results.
Here are a few elements for brands to pay attention to:
1. Powerful Combination
AI can analyze large amounts of data while providing recommendations for SEO strategies. These tools can also personalize and streamline SEO tasks. They are not a replacement for your SEO professionals, but a new skill set for everyone to learn.
2. New KPIs
Some might consider this a challenge, but we’re labeling it a benefit. The new AI age of SEO is ushering in new reliance on certain KPIs. A few examples:
- Engagement is changing. Users can see much more information on the SERPs (search engine result pages). This means you have more ways to reach people and encourage them to click through. Grabbing attention is no longer only dependent on your meta descriptions.
- Keyword footprint: More information on the SERPs can also mean people don’t click through. They’ve already got what they need. So, how do you measure your site reach and presence? Your keyword footprint. Your keyword footprint is a measure of how many keywords an individual URL ranks for. The more, the better the page’s organic reach and visibility. Even if someone doesn’t click through, the more times you appear in front of them as a reliable source, the better for your reputation.
3. More FAQs for More Real Estate
Because of how SERPs and engagement are changing, you have the chance to gain more real estate. An easy way is to embrace FAQ sections on your product and category pages. These sections are formatted in such a way that they are easy for a search engine to grab and display. They can also help teach generative AI tools exactly what you do and how you help.
Ultimately, it’s all about impressions.
1. Competitive Challenge
No one is really sure how AI updates and algorithms are going to be integrated into search results. Like all things SEO, we’re likely going to see several different iterations of results. That’s life. The unfortunate thing will be if search engines begin to rely more on AI than on human input and feedback, reducing the transparency of SEO.
Personalized search results are based on a user’s search history and behavior. While good for users, this is challenging for companies because it means the same search query will provide different results for different people. This makes it more difficult to optimize for a particular keyword. Companies must start considering different user preferences and behavior as they research and target keywords and phrases, instead of sticking with the “one size fits all” approach.
The Future of AI & SEO
Where is this heading? Here’s what we see right now.
1. More Personalization
2. More Emphasis on User’s Intent
As we said above, AI algorithms have the potential to make search results more accurate, personalized, and efficient. AI search engines are continuing to focus solely on understanding a user’s intent based on a search query.
This means that content creators and SEO professionals need to
- Shift from focusing on keywords to optimizing content to match the intent of a user’s query
- Use tools such as RankBrain, Google’s latest AI integration, to better understand intent and deliver more relevant content
Does Google penalize for using AI created content?
Right now, no. As long as you follow Google’s EEAT guidelines, the search engine is generally receptive to AI-generated content. Note that there are also no special gains for AI-generated content.
The EEAT guidelines – which apply to everything you create, AI or otherwise, are:
Google is always going to reward quality content over everything else.
The use of generative AI and LLMs is unprecedented. Everyone’s trying to figure it out, and there’s strong pushback from certain people and groups. Some celebrities and writers are filing lawsuits against generative AI companies for plagiarism and intellectual property infringement. Groups like WGA and SGA protested. Some people believe there is a moral factor in choosing or not choosing to use AI, even if there is nothing legally wrong.
Our takeaway is that AI with SEO can bring some benefits, but nothing can compare to a strong voice, expert stance, and unique content. Test out a tool. Ask it to rewrite one of your best performing blogs. How does it do? Does it capture your voice? Is it authentic to your brand? No. Not the first draft. Maybe if you continue coaching the tool and editing the content, it will get there, but you see our point. AI is a tool to complement human output, not replace it.