Use Data Or Be Used By Data!

The April 2 issue of Seotistics is here for you!

I will talk about AI once again to uncover the last bits you need to be more informed on the topic.

I've asked 2 experts to give me their opinions for this newsletter.

P.S. We will just examine the key concepts, if something is not clear, reply to this email, please!

​Read this in your browser​

Grasping Artificial "Intelligence"

AI means simulating human intelligence for computers.

How do you define intelligence? That's a tough question actually and Jess Peck discusses it in her article.

In SEO/Marketing, we simply talk about generating text or understanding patterns.

We humans can both write and detect patterns... but it's too much effort to do it for 1000s of pages (or more).

As explained by Cassie Kozyrkov in her article, you can say ML = AI in the modern industry (academics will hate you :p).

As digital marketers, we don't care much about all the differences, c'mon.

How Does AI Actually Work

AI makes predictions, it does NOT retrieve data from some database.

For example, when generating text, AI models actually calculate probabilities to see what will come next.

Given the word "the", what's more likely to come next?

Probably a noun or an adjective!

As my dear Seotistics readers know, RAG (retrieval) is a popular method to improve your outputs.

In this case, we use external sources to support our AI models and have a better output.

N.B. The retrieved data gets passed to LLMs, it's still about prediction but with a better context!

OK, Then Why We Call'Em LLMs?

LLMs stand for Large Language Models, because:

• Large: they were trained on massive amounts of data from the known web
• Language: you should get why... it's all about natural language
• Models: they model reality in a simplified way

What The Heck Are Embeddings?

You may have heard it everywhere... what are these embeddings, also called vectors?

They are representations of things, like texts, videos or images, in a format that machines can easily grasp.

Converting a text to a set of numbers is much clearer for a machine.

You can think of a vector as a list (array if you may) of numbers.

The important property of vectors is that machines can compare them...

so they can see if they are similar to each other.

The picture above shows 4 similar documents and 1 that is quite different.

So machines convert data into embeddings to find similarities between objects.

Vector Databases: More Madness

OK cool, machine learning models can create these vectors... and then?

It's better to store these vectors somewhere, like in a vector database.

You might ask... but why should I even store data?

Simple:

• LLMs have amnesia, they don't remember past inputs on their own
• LLMs need to have the context every single time

This translates to more time and money spent.

Since companies hate wasting money, vector databases exist.

They make it cheaper and quicker...

A good vector database you can test now is ChromaDB, which is also available in Python!

Oh wait... BigQuery can also be a vector database!

Is there something that BigQuery can't do?

Custom GPTs & SEO/MKTG Use Cases W/ Caitlin

Custom GPTs made a lot of buzz but didn't amount to much change.

They are invaluable for personal productivity but aren't suitable for large-scale automation.

I hope you know what they do...

in case you didn't know, it's like having a specialized ChatGPT that handles specific tasks and is finetuned with other data.

If you want some AI solutions for your company, Custom GPTs are toys.

We rely on engineers to build more complex solutions that are safe and scalable.

Anyway, there are some cool Custom GPTs you can use in SEO/Marketing:

• ​Ultimate Content Repurposer​
• ​Diagrams: Show Me​
• ​GA4 SQL (if you use BigQuery with GA4)

Caitlin talked about even more GPTs in her Moz webinar and you can download the slides here or check her article.

Among other things, you can also help your content strategy with AI if you are smart enough like Caitlin:

SEO and Marketing are already affected by AI but even if you can't code there are useful applications.

Summarizing data/files is one of the best things for lazy people:

Examples of AI

So far I've been playing with Neo4j, a promising graph tool that is well versed in AI.

I still need to test it in depth but I didn't like the User Experience which is quite confusing.

A straightforward application for your website could be a chatbot answering the questions of your users.

More complex use cases can involve asking questions to your BigQuery base or even generating text for your products.

Commercial Break

A great tool I've recently tested is QuizWizard AI which allows you to create quizzes from videos or blog posts.

This is invaluable for content websites if you start with target social media (like Facebook).

I bet you know quizzes can go viral easily...

QuizWizard is ideal for lazy people like me who HATE manual work.

[The link below is affiliate]

🔗 Try out QuizWizard AI​

Prompt Engineering

Prompt Engineering is a bit of a buzzword but there is some logic in it.

Knowing how to structure and write your prompts is what makes the difference.

When using APIs you can't afford to be vague and generic like you'd do with ChatGPT.

There are many techniques used in prompt engineering to ensure you get the best out of your results.

Some examples include:

Machines don't behave as expected if your prompts are generic.

So you should be specific and even be clear about your output!

LangChain... Mentioned Everywhere

LangChain is an open-source framework that you can use to build chatbots or cool AI applications.

To keep it super simple, it combines LLMs with external data sources, like the data you have in BigQuery.

An example of its use can be developing a chatbot to ask questions about your BQ data.

It's nothing else than a connector between AI and data sources.

Andrea shows an example below:

As you remember from this past issue, this is also an example of RAG since the machine is retrieving context from a vector database.

And I want to also remind you that Knowledge Graphs are a great alternative to vector databases.

This is often the case when you want to explore relationships and need more explainability.

[Analytics For SEO Course: Preorder]

You can pre-order my Analytics For SEO course with a 30% discount until June 1.

This course will teach beginners how to analyze SEO data and seasoned professionals more advanced workflows.

AI Agents

One of the most important concepts in AI and probably the future of our world.

AI agents are entities that are able to perceive external inputs and act based on the environment.

The most common examples are in Robotics where agents are the robots we all know (or self-driving cars).

In SEO/Marketing, an agent can be a chatbot...

the key difference with traditional automation is that AI agents act on their own.

They can prompt themselves and the only thing you need is telling them what's the goal.

Does AI Fit Into Analytics?

AI is usually considered to be part of the prescriptive step of Analytics.

After RAG, there are many interesting use cases in SEO but I wouldn't say for every business.

You see the most value in big websites and companies that can afford to test more.

In the vast majority of cases, you should test or propose AI as the last resort.

If you want to learn more about the 2 experts, you can find them at:

Andrea d'Agostino:

• ​Twitter​
• ​LinkedIn​
• ​Medium​
• ​Website

Caitlin Hathaway:

• ​Twitter​
• ​LinkedIn​
• ​Website​

👥 Join Our Community

Our Discord community offers a small place where we can talk business and SEO.

If you hate all the noise of social media, then this place is for you.

🔎 Analytics For SEO Ebook (v7)

If you want to learn about Analytics for SEO and get the best resource out there, this is for you.

It will teach you or your employees to:

👉 Prepare audits that make sense and are actionable 🔥

👉 Avoid common pitfalls that cost you money 💸

👉 Create analyses that add value and money💯

👉 Move the needle faster with efficient SEO systems ⏳

This comes with periodical updates to keep the content fresh.

📚 Recommended Reads

Some good ol' resources to learn more about AI:

• ​Analytical Skills for AI and Data Science : Building Skills for an AI-Driven enterprise​
• ​Data Literacy In Practice​
• ​How Search Generative Experience works and why retrieval-augmented generation is our future​
• ​A Guide To Artificial Intelligence in Easy English​
• ​12 Prompt Engineering Techniques​
• ​Create a Chatbot in Python with LangChain and RAG​
• ​What is an AI agent?​
• ​Guide to Building a Content Strategy using LLMs​

📞 One-Hour Call

If you have doubts about SEO or Analytics, you can book a call with me.

Have doubts about your content website or with your data?

Look no further, I can help you:

❗️ Feedback and Recommendations

If you have ideas/recommendations for the next issues of Seotistics, you can simply reply to this email.

Marco Giordano
​SEO Specialist & Data Analyst

Follow me on 🔽🔽🔽:

​​​

Bernerstrasse Süd 169, Zurich, Switzerland
​Unsubscribe · Preferences​