Five AEO Skills on GitHub: What Each Optimizes For

By

A year ago, everyone was sharing prompts, and now they have moved to skills. Skills have a lot of advantages over prompts, but it’s easy to use them without understanding what they actually do.

When you get a prompt from someone on the internet, the text is right in front of you, even if it’s complex. In my experience, many people install skills without reading any more than the about page on GitHub.

I analyzed five popular answer engine optimization (AEO) skills on GitHub, and what I found is that, while they have similarities, they optimize for very different things.

Why did I look at these AEO skills?

Skills are a quick way to give your agent a deeper or more specific understanding of a topic. But with that, you also outsource choosing what is important. You inherit more than a structure. You also inherit someone else’s judgment about what matters.

These AEO skills show the impact this has, especially in a new field that is constantly changing.

I reviewed five skills:

I chose them due to each of them having a presence on GitHub and activity showing people are actively choosing to use them.

Each tool was built for a different job

If you take one thing away from this article, I want you to know what outcome you want from a skill before you download it.

All five of these skills are meant to help with AEO, but they approach it in different ways. If you don’t consider your goal, you may end up with recommendations that don’t move the needle.

What did all the AEO skills have in common?

They all agree that SEO basics still apply, but your content also needs to work when an AI system extracts individual sections. That means keeping each question close to its answer.

You should use descriptive headings, including directly using questions, but also write your sections so they can stand on their own. If the AI agent needs to understand the full context of your article, it’s less likely to use you as a source.

A clear paragraph on your FAQ page that states the product, price, and benefits may be easier for a large language model (LLM) to extract and cite.

With that said, they all agree that core technical SEO is still vital. You need to make sure your content is accessible, make sure the technical structure is sound, and make sure your meta tags and titles are all optimized.

These skills all take those principles and turn them into repeatable workflows to help AI agents inspect your site/content, find issues, and suggest changes.

Where do the AEO skills differ?

There is a clear distinction. While similar, they were built for different jobs.

Corey Haines starts with the marketing questions: What are your customers asking, where does your brand show up, and are you actually getting recommended? You can get cited for explaining a topic while a competitor gets recommended as the company to buy from. Those are different outcomes, and Corey’s skill makes that distinction part of the strategy.

ReScienceLab gives you a more straightforward checklist for reviewing your site, researching keywords, improving content, and checking SEO basics.

Daniel Agrici takes a broader SEO approach, bringing technical checks and, when connected, your own analytics into the work.

Zubair Trabzada focuses on a comprehensive generative engine optimization (GEO) audit, using multiple agents to review different areas and turn their findings into scores and a prioritized report.

Aaron He Zhu focuses more on improving existing content, keeping the facts about your business consistent, and making it clear which numbers were measured and which were estimated.

These differences affect what the skills tell you to prioritize.

Take llms.txt.

In the versions I reviewed, Zubair treats a missing file as a high-priority issue and includes it in the scoring. Daniel gives it no citation weight at all. Aaron treats it as an optional resource that may be useful as these systems develop.

That means the same website could get different recommendations depending on which skill you use.

Before acting on one, you need to understand whether it addresses a verified problem, follows a writing preference, or depends on a tactic whose impact is still uncertain.

The scoring models also diverge considerably.

Zubair’s passage scorer rewards things like structure, word count, and language that looks like supporting evidence. Aaron explicitly separates making content easier to cite from getting it discovered and cited naturally.

A higher score doesn’t automatically mean more visibility, better recommendations, or more customers. When comparing these skills, look at what they checked, what evidence supports their findings, and whether the recommended changes move you toward your actual goal.

Skills are not plug and play

It’s tempting to find a skill in an area you need help with and install it immediately, but I highly suggest reading it over and understanding how often it is updated. I like AEO as a topic for the post because the knowledge and understanding are continuously changing, so the skills needed to keep up also need to be updated.

Not every skill is like this. I have been working on a skill to write better hooks. While I am updating and improving this skill over time, what makes a good hook doesn’t change regularly. AEO/GEO is still a rapidly changing field, so choosing the wrong skill could have you optimizing for the wrong thing.

At the very least, you should read through the skill files to understand:

  • What job does it do?
  • What does it reward?
  • What evidence supports its advice?
  • Is that evidence current?
  • How will you verify the findings and measure the outcome you actually want?

What AEO skill is best for the solo marketer?

For a solo marketer trying to work out where to focus, I’d start with Corey Haines’s ai-seo skill.

Based on this comparison, I like that it starts with business context and the questions your customers are asking. Those simple questions inform the whole strategy and take you from generic AI feedback to a personalized plan.

It also distinguishes between getting cited and getting recommended. A citation for a definition might be useful, but it means something different from being recommended when someone is deciding what to buy. This distinction deserves its own article.

I’d use it to help identify gaps and prioritize work.

I’d give it context about the company, customers, offer, and goals, then ask it to explain the evidence behind its recommendations. I’d still check specific claims about platforms, schema, and expected improvements before making changes.

The broader lesson applies whichever skill you choose.

These skills contain someone else’s judgment about what matters. That judgment shows up in the questions they ask, the things they score, and the work they recommend.

If you don’t understand those choices, you can end up spending time improving a score without knowing whether you’re improving anything that matters to your business.

So before downloading a skill, read what it’s asking the AI to do. Look at what it rewards, where its advice comes from, and whether that fits your goal.

You don’t need to become an expert in every technical detail, but you should understand enough to question the output.

A skill can help you make better decisions. You still need to decide whether its advice makes sense for your business.