Start incorporating programmatic approaches and rule-based automation into your SEO

Lazarina says: “Incorporate programmatic approaches. I’m specifically referring to rule-based automation for SEO best practices and AI-driven automation for process enhancements.

Mainly, this means taking some time to better understand different programmatic approaches and learning how to incorporate them in a responsible way – both for the brand voice of the company you’re working for and content quality. However, you’re not doing that just to put out content. You’re doing it to satisfy search intent and be helpful to users.

I’m going to talk about two approaches that would be very useful to understand. The first is rule-based automation. This is very good for creating predefined instructions or conditions that dictate how specific tasks and processes should be executed without having someone intervene or monitor them.

You can apply this kind of automation for well-known, specific best practices that we are given by Google. This could be making sure that all pages have titles and headings, headings have keywords, pages have meta descriptions, images have captions, and internal links and breadcrumbs are inserted whenever certain phrases are mentioned or to promote semantically related pages. It could also be something a little bit more complex, like triggering structured data generation at the mention of certain phrases such as ‘how to’ in the title.

We are seeing a rise in AI models, which can be extremely helpful for a variety of tasks, but we’re also seeing misuse of these models. It’s important to emphasise that AI-driven automation can be extremely helpful for tasks that are not mission-critical. These tasks can be things like automating content generation, content enhancements, summaries, or content translation, but these tasks absolutely need to have someone monitoring the output and optimising it.

I’m a huge proponent of these algorithms being used, but they should be used for improving processes and not for something that is served to the end user.”

How do you know when to use a rule-based automation approach?

“Rule-based approaches are things that we should seek out from Google’s guidance and the guidance of other search engines. In the past, SEO as an industry didn’t really have that level of guidance on how it should be done. Now, what works and what doesn’t is a lot more well-documented.

Everything that we have asset guidance for – which has systematically proven that it is helpful to users, improves rankings, and is best practice – these are the things that we should be thinking of automating through rule-based automation.

The examples that I gave are very commonplace now: having a title, having headings, breaking down the text into paragraphs to ensure that it reads well on a mobile device, etc. Of course, these can be built into the processes of the people creating the content, and you can create your own checklist, but it can also be a lot more automated. You can build a system to automatically flag when something is off. These are very good examples of a rule-based automation approach.”

What SEO tasks would you use a programmatic approach for without utilising AI?

“The tasks where you can apply rule-based automation without utilising AI are more about the lengths that titles should be, how many headings you should choose, and generic advice. You don’t necessarily need AI for that. You can use AI as part of the process, but you don’t need it in order to make that successful.

You want to make sure that you have the best practices and processes implemented before you incorporate AI to make the output better, as opposed to relying on AI to throw out a page and make it optimized for SEO by telling ChatGPT to act as an SEO consultant. The difference is that you should lead with SEO best practices when you are designing the content patterns of your pages, how they should be structured, and what is needed in order to cover a topic comprehensively. Then you can incorporate AI to help speed up the process, whether it’s for generating content, translating it, or repurposing it.

The caveat here is that, if you have a very good rule-based approach, you might not need that much intervention at the final stage. When you have an approach that is entirely based on AI, you absolutely should have someone checking the output, because otherwise, there are a few things that could happen.

The first thing is that, over time, you will lose your individuality in terms of brand voice. You might see content being published that is not necessarily aligned with best practices, both in terms of how cohesive it is and how accurate it is. If you are relying on AI tools for content automation, you might see information being generated that is not factual at all.

For other tasks like translation, AI is not really a good localization exercise. Besides brand voice and a lack of cohesion, you might also see other things like losing touch with the user because they can see that the content is not based on reality, experience, or authority. Overall, while you might think that you are following a strategy and publishing content regularly, it might not be the quality content that both search engines and users are looking for. That’s why you need to have human intervention when you are relying on such tools.”

What are some examples of when and how you can start thinking programmatically?

“The difference with programmatic SEO, for someone unaware of this term, is that it is based on the principle of providing new information to satisfy the user, but doing that through programmatic means.

For instance, if you look at a keyword during keyword research, you might be looking at things that have a particular pattern. This could be for your category or product pages because you are organising content in a particular page template and pattern. Another example could be in blog posts and other niche site entities. When we talk about entities here, these are things that are defined with different attributes. For instance, a dog would be an entity and the food that they eat would be an attribute.

If you pivot that to programmatic SEO, you might have a keyword phrase like ‘the types of food that dogs eat’. With programmatic SEO, you can take these examples and use templates to go through the different variables that your attributes have. In this case, those variables could be different dog foods or different types of dog foods, such as canned food or kibble. This would give you a very good indication of which particular elements of the page you can template.

To give a practical use case example of a programmatic issue and a rule-based approach, I recently reviewed a case study for G2 and how they do their category pages. They are a reviews website, and they have different category pages, like CRM software or ERP software. Within that, they have different ways of incorporating both rule-based automation and AI-driven automation for generating very small sections of the content.

They use rule-based approaches for templatising their headings. They might have the headings ‘What is ERP Software?’ and ‘What Are the Benefits of Using ERP Software?’, then they simply change out the category. They also use rule-based approaches for inserting particular links and organising the page template for all of these types of category pages.

When it comes to implementing this approach, multiple different businesses and niches could benefit from this. If you’re an e-commerce store working on category pages and product pages, you can take that approach backwards to see what best practices you aren’t currently implementing and how you can brew those into your processes for publishing pages. Also, you can use it to think about the content that you are missing or the different content elements like captions for images, meta descriptions, and titles, and how those can be improved.

Then, you can use very tailored machine learning models to help you in the execution of those tasks, but not as the final approach that you use for doing everything on the page.”

How do you double-check that what you’re producing actually makes sense and the entities and attributes match up correctly?

“Something that is very important when you are doing programmatic SEO and trying to create content using a programmatic approach is having a very cohesive database and strategy.

The strategy is very important because you want to know everything that you are planning to cover with this approach but it also allows you to sense-check the different avenues that you can take with your keyword targeting. The database is important because you want to make sure that you are not replicating content that is already available somewhere else just because you want your pages to look better or manipulate rankings. You want to provide new information.

Every programmatic SEO project starts with doing what you see your competitors doing. What separates successful examples from unsuccessful examples is going that step further and adding unique information to your pages to ensure that you are providing user value.

At G2, they have a graph on their category pages that shows a perception map of all of the ERP software that they have in their database and how they rank based on different characteristics. That is a great example of using your own database to provide value to the user that they could not find elsewhere. This perception map is something unique that they have added to the page, which might be a quality signal to both users and search engines that they should prefer that page.

To make sure that you are providing value for users, you should start by creating a very well-organised database and a cohesive keyword targeting strategy, and that strategy should be aimed at covering the topic in a way that is complete and robust. You should also be focused on making sure that you provide unique value to those pages and not just replicate what everyone else is doing.

That applies to all pages that you build: blogs, products, etc. Don’t just replicate what’s already there. We know very well that there are sustainability issues related to that, and there are also web quality and user experience issues that come with that. You definitely don’t want to be a website that is full of pages that aren’t unique or helpful in any way.”

If an SEO is struggling for time, what should they stop doing now so they can spend more time doing what you suggest in 2024?

“Stop doing things manually that you could be doing programmatically. I’m talking about things that are becoming low-importance factors in terms of rankings and value to users, but they are still considered best practice.

I’m talking about things like manually writing meta descriptions and image captions, especially for very large enterprise websites. These are things that a lot of people are struggling with and putting a lot of time and effort into. We have very advanced and robust APIs that can do the job to at least 90%. Having someone simply reviewing and improving those would be a much better use of your time.”

Lazarina Stoy is an SEO and Data Consultant, and you can find her over at LazarinaStoy.com.