AEO 201: How to turn AEO into a system

On every intro call, we ask how the prospect found us. Lately, about 80 percent give the same answer: ChatGPT or another LLM.

In AEO 101, we walked through how to get your baseline: what AEO actually is, where you stand today, which prompts to track, and how to read the data once you have it. If you've done that work, you probably have a spreadsheet or a dashboard full of numbers and one big question. Now what?

When we started at Iron Horse, we tried a lot of different things. But, that's how you learn in a space that's new and constantly evolving. But one-off pushes are random, confusing to track, and you can't repeat them. You ship a page, the visibility score moves (or doesn't), and there's nothing connecting the two.

What actually moves the numbers over time is a system: a loop you run every sprint that tells you what to try next, whether it worked, and what to keep doing. We developed our MAPS (Measure, Aim, Produce, Systemize) framework for AEO to help B2B marketers create a system tailored to your specific audience and business.

Here's what's in this article:

• Why you should design your system before you have results
• Iron Horse’s MAPS framework for AEO, stage by stage
• How to run it in two-week sprints
• Where agents fit
• How to connect AEO to pipeline
• Who should run it

Like part one, we'll keep updating this as we learn.

When should you build your AEO system?

You don't have to design your whole system before you start. If you test with a consistent method, the system takes shape as you go.

Start with a few experiments and see what works for your team. Every test starts with a hypothesis, and the result tells you whether to keep going or drop it.

What matters is running each test the same way: track the same metrics, check them on a set schedule, and give every test an owner. Do that and you end up with a record of what you tried and what moved your numbers. That record is the start of your system.

When something works, it becomes part of the loop. When it doesn't, you drop it. Expect the system to keep changing, too. What works for our team might not work for yours.

What is the MAPS framework for AEO?

Iron Horse’s MAPS framework provides marketers the building blocks for creating a sustainable system for AEO. It’s the foundation for how we deliver AEO results for our B2B customers and ourselves. It stands for Measure, Aim, Produce, and Systemize, and it runs as a loop.

You measure where you stand, aim at the gaps, produce the content and fixes, then check whether any of it worked. Whatever you learn goes back into Measure, so your next baseline reflects what you just shipped.

That last step is what keeps MAPS from being a checklist you finish once. The loop closes at Systemize, and then it starts over.

Here's what each stage looks like when you're starting from the ground floor.

Measure: where do you stand today?

Measure sets your baseline. It's where you pull together the competitors you're actually up against, the personas you're targeting, and the prompts you're tracking. Then record where you stand before you change anything. That becomes the benchmark you track everything else against.

We covered how to build prompts and read the core metrics in part one. What's different here is that you measure in three layers: AEO metrics, web metrics, and pipeline. We'll get into how to read them together further down.

AI answers change every time you run a prompt, so don't read too much into a single answer. Look at your prompts as a group and watch the trend over weeks. If most of your buyers use one platform, weigh your measurement toward that one.

Aim: what should you go after?

Aim turns your baseline into a roadmap. Look at the prompts you have low visibility in, see who's ranking for those topics, and build your strategy around those gaps.

Don't stop at your own website. In our September webinar, Nick Lafferty, founding marketing engineer at Profound, shared that only about 5 to 10 percent of citations in an average category come from a brand's own domain. The rest come from somewhere else: publishers, affiliates, LinkedIn, YouTube, Reddit, even your competitors' sites. Which of those matter for you depends on what's actually getting cited for your prompts.

Once you know your gaps, prioritize them. The Momentum, Defense, Offense, and Alarm grid from part one is a good place to start. Then break your priorities into sprints. Each one gets a hypothesis, the action you're taking, and what you expect to change.

By the end of Aim, you should have a roadmap you actually believe in, split into sprints you can run.

Produce: what do you actually ship?

Produce turns your roadmap into shipped work: new content, updates to existing pages, technical fixes, and off-site work like getting into listicles and third-party articles. Every piece should trace back to a gap you found in Aim.

A few things we build into everything we ship:

• Depth on a topic: AI looks for depth and consistency, so cover a topic across several connected pages and link them to each other instead of cramming everything onto one page.
• Named entities: AI picks up on named frameworks, methodologies, and products. That's part of why we call ours MAPS instead of "our approach to AEO."
• Proof AI can check: AI validates your claims against other sources, so give it something to check: author bios with LinkedIn links, case studies with real client names, and schema that spells out who you are and what you do.
• Original research: When AI cites a stat, it cites the source that published it. Give it numbers nobody else has, like surveys, benchmarks, or data from your own client work.
• Pages AI can read: Content that only loads through JavaScript is hard for AI crawlers to read. Keep your key content in the page's HTML.
• A clear claim: Generic messaging gets lost. Say exactly what you do and who it's for.

Then check what AI is actually saying about you. If an answer is pulling from an old page or outdated positioning, that's what buyers read. Fixing stale or inaccurate content counts as Produce work too.

Systemize: what do you keep?

Systemize is the check. Go back to the prompts you targeted and see whether what you shipped actually changed your visibility, position, or citation share.

Then make a call: keep it or kill it. If it worked, keep doing it. If it didn't, dig into why before you move on. Maybe the page isn't getting cited at all, a competitor's content is ranking higher on the same prompts, or AI is still pulling from an older page. Knowing why tells you what to try next sprint. Both answers are useful, because both tell you something about what AI responds to in your category.

Whatever you keep should get easier every round. Turn it into briefs, templates, and agent instructions so you're not starting from scratch the next time you run the same play. That's how the loop gets faster over time.

How do you run MAPS in sprints?

We run MAPS in two-week sprints. A big 90-day AEO goal is hard to act on. A sprint breaks it down into what you're doing this week.

Here's how one runs:

1. Pick one priority from your roadmap: Pull it from the list you built in Aim.
2. Write down your hypothesis: Something like: if we do X, we expect Y to move on these prompts. Record the current numbers for those prompts so you have something to compare against.
3. Ship the work: Create the content, make the fixes, or do the off-site outreach you planned.
4. Check what moved: At the end of the sprint, look at visibility, citation share, traffic, and bot traffic against the numbers you recorded.
5. Make the call: Keep it, kill it, or dig into why it didn't work, then pick the priority for your next sprint.

Two things make this structure work for us. First, testing one hypothesis at a time means that when something moves, you know what caused it. Second, two weeks is long enough to ship something real and short enough to change course quickly. AEO also moves faster than SEO. SEO changes can take months to show up in rankings, but we've seen AI visibility shift within two to three days of shipping a change. If you wait a quarter to check, you miss that signal and end up reacting late.

That's the loop: test, learn, adjust, repeat.

How can agents speed up your AEO system?

You can run your AEO sytem without agents, but agents make it faster and catch things you'd miss doing it on your own.

We build our agents in Profound. The ones we lean on most:

• Opportunity Identifier: Pulls our visibility and competitor data together and hands back full opportunities, including topic ideas and gaps where competitors show up and we don't. Our content team takes it from there. This one feeds Aim.
• Content Refresh: Looks for what's gone stale, like outdated stats or old positioning, and recommends how to fix it. This one handles the accuracy check from Produce.
• Social Listening: Tracks what people are saying about us and our space off-site, so we know where conversations are happening that AI might pull from.

There's no set list of agents you need. Build them around what works best for your team and your industry.

How do you connect AEO to pipeline?

This is the hardest part. In the post registration survey for our September AEO webinar, 45 percent of respondents said proving pipeline impact was a blocker. Among teams already scaling AEO, that jumped to 60 percent, more than any other blocker at that stage. You can't see a buyer's full conversation with an AI before they land on your site, and AI-referred conversions rarely tell the whole story.

So we read the three layers from Measure together:

• AEO metrics: Visibility, position, and citation share are your leading indicators. They tell you whether the work is landing before it shows up anywhere else.
• Web metrics: AI-referred traffic and conversion rates tell you what's actually driving visits.
• Pipeline: Leads, meetings booked, and opportunities. This is where you see whether the work is turning into actual business.

For one client, we pulled all three into a single dashboard: Profound for visibility and share of voice, plus GA4 and Webflow for AI traffic and conversion rate.

Self-reported attribution fills in what your analytics miss. Add a "how did you hear about us" field to your forms, and listen back through sales calls for mentions of ChatGPT, Gemini, or Claude. It isn't perfect, but it catches the buyer who asked an LLM before they ever hit your site.

Who should run your AEO system?

At Iron Horse, one person owns the loop day to day and pulls in our content team to ship each sprint. Agents cover a lot of the repetitive work, which keeps the loop moving without a big dedicated headcount.

Whether it's one person or a whole team, make sure these functions are covered:

• Data: Setting the baseline and KPIs, and checking the numbers a few times a week.
• Content: Shipping the new pages, updates, and fixes each sprint.
• Agents: Building and managing the agents, and updating their instructions as you learn what works.

However you staff it, someone has to own it. The loop only works if someone keeps it running.

Where do you start?

Every AEO engagement we run comes back to four pillars. The first two get you discovered, and the last two get you chosen.

The fourth pillar is what this whole post is about. Without a system, the other three stall after one round.

If you're on the ground floor, you don't need all of it on day one. Get your baseline, pick one priority, and run your first two-week sprint. What you learn from that sprint tells you what to run next.