Ethan Smith, CEO of Graphite, a leading AEO agency, outlined a four-step framework for implementing AEO in your business:

1. Question Research

2. Question Tracking

3. Have a strategy for each platform

4. Run experiments

This post covers steps 3 and 4. If you haven’t read Part 1 yet, start there — it covers question research and tracking before you touch anything on this list.

The short version of where we left off: you’ve identified the questions your target customers are asking AI engines, and you’re tracking where and how often your brand appears in the answers. Now the work begins.

3. Strategy for Each Platform

As covered in Part 1, there are two categories of sites to think about:

1. Onsite: your own website

2. Offsite: external websites

There are four types of offsite platforms:

Based on your question tracking results from step 2, you need a specific strategy for each platform where you want to show up. Ethan shared two concrete examples.

Onsite: Help Center Optimization

Many questions customers ask LLMs follow a predictable pattern:

• Can X do Y?

• Does this product or service support my use case?

• How do I do X with Y?

These are capability and fit questions. They don’t belong on marketing landing pages — that’s not where AI engines look for them, and that’s not where customers expect to find them. Your Help Center or FAQ is the right home for this content, and it’s your best opportunity to win long-tail AEO. (The same logic applies as in SEO: unless you’re a giant corporation, mid-to-long-tail is where you can actually compete.)

Three actions to optimize your help center:

1. Subdirectories over subdomains: Move help content from help.company.com to company.com/help. Subdirectories inherit your root domain authority; subdomains don’t. This benefits both AEO and SEO.

2. Optimize internal cross-linking: Link between related articles within your help center. AI engines infer topical authority from site structure — strong internal linking signals that you cover a subject deeply, not just broadly.

3. Create content for specific use cases: Target long-tail questions others won’t bother writing for. A question like “What is the best payment verification API that uses GraphQL?” is narrow enough that you could own it entirely if you answer it well.

Offsite: Reddit Optimization

AI search engines favour human voices over brand voices. The reasoning is straightforward: UGC is harder to game, includes real edge cases, and naturally handles follow-up questions because it comes from people with actual experience.

Reddit is the highest-leverage UGC platform for most B2B use cases. The approach is simple but requires genuine effort:

1. Identify the threads where your product or service is relevant

2. Add authentic, high-value comments using a transparent structure: “Hi, I’m [Name]. I work at [Company]. Here’s what I know about this.”

The transparency is the point. Ethan is clear that you don’t need to scale this — a handful of genuinely useful comments in the right threads outperforms a hundred generic ones. Quality is the signal AI engines are looking for.

4. Run Experiments

Here’s something Ethan is direct about: most AEO “best practices” circulating right now are wrong, incomplete, or not reproducible.

The field is too new and too fast-moving for conventional wisdom to be reliable. A tactic that worked for one brand in one context doesn’t automatically transfer. The only way to know if something actually works for your business is to test it yourself.

A good strategy is one that’s testable and reproducible.

How to Structure an AEO Experiment

1. Define your question set: Use the 100-200 questions from your Question Research phase as your test pool.

2. Split into control and test groups: With 200 questions, assign 100 to control (untouched) and 100 to the test group. The control group is your baseline.

3. Sub-split the test group by strategy: Divide the test group so each subgroup gets one primary optimization. For example: 25 questions for Reddit comments, 25 for help center expansion, 25 for YouTube content, 25 for affiliate or review mentions.

4. Make the strategy precise: Don’t write “post on Reddit.” Write “post 5 high-quality comments to existing Reddit threads.” Vague strategies produce uninterpretable results.

5. Wait: Give it at least 4-8 weeks. AI engines take time to index, crawl, and incorporate new signals.

6. Measure and compare: Track mention frequency, thread visibility, and position in AI answers. Most AEO tracking tools can capture these metrics. Compare test vs. control to identify meaningful lift.

7. Reproduce the result: Run the same strategy again. If you can’t reproduce it, it’s not a strategy — it’s a coincidence.

The reproducibility step is the one most people skip. It’s also the most important one.

That wraps up the AEO Strategy series. There’s more to explore in each of these areas — particularly around affiliate strategy and YouTube optimization — and we’ll go deeper in future posts.

If you want to talk through any of this for your specific situation, reach out directly.