The 5-Step AI SEO Blueprint For Crypto Projects

Many crypto founders and marketing leaders feel the need to do something about AI SEO. We all use ChatGPT, Claude, or other LLMs. And so do our potential users and partners.

How can crypto products get recommeneded by AI assistants?

This post will give you a clear blueprint for how to approach AI SEO.

First, let's get clear about the definition.

By AI SEO, I mean the strategy of optimizing for getting mentioned and recommended in ChatGPT and other AI assistants.

This is also known as AEO, GEO, or LLMO.

Now here are the steps that worked for me.

This is an extremely important step that cannot be ignored.

In traditional SEO, it's easy to jump past this. We can all open Ahrefs, find some keywords, and start creating content. In search engines, people were just typing keywords and getting the some results for their queries.

But ChatGPT doesn't care about keywords. Users are asking specific questions, which can be very detailed and include lots of context, often from memories of earlier chats. AI assistants have deep understanding of the user and are programmed to assist them.

Our goal as marketers is to help AI assistants do their job. Help them help the user.

That's why we need to be very clear about the target audience:

For this step, you can use traditional marketing personas. You can also supplement it with actual data you have. User interviews, support chats, and reviews are all great sources of data.

You can also just feed everything you have into LLMs and ask it to create a target persona.

From these marketing personas, you can derive target prompts.

These are the prompts that the user can ask.

It could be something simple like: "What wallet should I use to stake Ethereum?"

Or it could be something more advanced: "I am building XYZ product. What's the easiest way to implement ABC feature?"

The key here is to go beyond keywords (e.g, "stake Ethereum") and focus on prompts that focus on some result. About 10-25 prompts is a good starting point.

All other steps are based on this.

Once you have prompts, you can track what ChatGPT is citing and recommending now.

The simplest way to do this is to just ask your prompts in new incognito browser tabs. Then just records responses.

But this isn't scalable if you have many prompts and to track them regularly.

Lots of tools emerged to help with this — I'll cover them in a separate post.

To avoid wasting time on choosing the tool and setting it up, do your research manually at first.

Then look at what ChatGPT gives you:

This last question is the most important.

When you see some sources cited, you should explore them in depth.

This is your opportunity to influence what ChatGPT says.

When a source is cited, this means ChatGPT ran a web search to answer the user's question. These searches are done in real time and are meant to bring fresh information to supplement the training data.

Your goal is to create better content than what ChatGPT cites now.

This is more art than science and could include:

Sometimes you can win by publishing fresh, new content. This will naturally overtake articles that are 2-3 years old.

Which leads us to another step.

If you already have existing content that is ranking in search, you have a head start.

This similar to the previous step except you update existing content to help answer the prompts.

The easiest way to do this by adding an FAQ to your posts and landing pages. These questions should mirror what user prompts ask.

There is one trick to find what your users are prompting.

Go to the Google Search Console and query enter "?". This will give you all queries that have question marks.

These longer keywords are actual prompts that users now also ask in Google's AI Overviews and AI Mode. Be cautious about queries with >100 impressions. Those are likely some bots. Instead, queries that have just a few impressions are gold. That’s real people asking things. They likely ask ChatGPT in the same manner.

Creating specific answers to these questions and putting them on existing pages is an easy way to get more citations.

Traditional SEO already has this concept of trust signals through external validation.

ChatGPT and other LLMs rely on the web to understand what can be trusted.

There is no way to directly measure trust. So instead, AI assistants use online popularity as a proxy for trust. If something is frequently mentioned across the web in a positive light, this likely can be trusted.

Of course, this is very naive and biased. It can and will be gamed.

Your goal is to be mentioned wherever it's possible:

PR or Public Relations is coming back. But this time you care mainly about AI assistants.

Going back to Step 3. Saw some sources mentioned in ChatGPT's answers? Well, unless they are your competitors, you can just go and ask them to include you. This would cost money, of course, but it's the easiest way to influence AI assistants.

In this post, we saw specific steps to starting your AI SEO journey.

It all starts with understanding your audience. Then you research and see what ChatGPT et al already say. After that you create and optimize content to better help their users.

Of course, each step has many substeps. And it's a field that evolves very fast, I'm learning new things every day but just doing stuff. But I strongly feel that these steps are foundational and will not change.

It's easier to learn this together. Subscribe now to find out what works.