AEO 101: the guide we wish we'd had when we started

The way B2B buyers research a vendor has changed, and marketing teams are trying to catch up. What used to start with a Google search and a stack of open tabs now starts in an AI search engine like ChatGPT, Gemini, Claude, or Perplexity, where a buyer is asking a question or a series of questions and treating the names that come back as a shortlist. Fifty-one percent of B2B software buyers now start their research with an AI chatbot more often than Google, according to G2's 2026 research, up from just 29 percent a year earlier. AI search has become the new front door, and the place buyers go before they ever reach your site.

We’re seeing this change with our own prospects. In every discovery call, we ask the prospect how they found us, and more often than not we’re hearing ChatGPT, Gemini, and Claude. If your brand isn’t in the answer an AI gives at that moment, you’re not in the room at the exact moment a buyer is deciding who to reach out to. 

About a year ago, we started asking ourselves the same question we ask our prospects: how do we show up in that answer? We dove deep into thought leaders in the space and the early testers of AEO, ran our own experiments on Iron Horse and Iron Horse Studio, and spent weeks trying to dissect and unravel information that was, honestly, pretty convoluted. There wasn't a guide written specifically for us, so we built our own by trial and error. This is that guide, the one we wish someone had handed us when we first started.

This is a comprehensive guide. Here's what's in it:

• What AEO actually is
• Where your business stands today
• Whether you need a platform or can start manually
• How to write prompts that match your actual goals
• What to track, and whether the data behind it is real
• How to find opportunities once you're tracking
• What to do with years of old content
• Who should own the work
• How to turn it into a system instead of a one-off project

AEO moves as fast as the models behind it, so we'll keep updating it as we learn.

What is AEO?

AEO is the practice of optimizing your brand to get discovered, mentioned, and cited by AI answer engines. Think of it as an additional layer to SEO. You still need the SEO fundamentals, but now you're optimizing for how an LLM puts together an answer.

There’s a real technical piece to this: schema, structure, how your pages are built. But the bigger shift is that you're now writing for a machine to parse and repeat while still having to sound like a person wrote it. Your brand can't turn into generic AI output just because you're optimizing for an LLM, buyers can tell.

One thing we heard a lot when we started is that AEO levels the playing field. A brand with no domain authority and no backlink history can get cited right alongside a company that's been building SEO authority for over a decade. There's some truth in that, but it doesn't mean SEO stopped mattering. Good AEO still sits on top of good SEO. You're just adding a second set of signals on top of the first.

Here’s how we compare the two: 

What AEO stage are you at?

You can typically place your business in one of three stages: ground floor, building, or compounding.

Ground floor

This is where you have nobody owning AEO, and you're not tracking a single prompt. Your content is still built for search, not how AI actually reads and cites a page.

Building

In this stage, the tools and the tracking exist, but content and off-site work haven't caught up. You have a handful of pages optimized, but a lot of them are untouched, and any off-site mentions happen by accident. There's still no dedicated owner.

Compounding

This is the most mature stage. You have a dedicated owner, a named framework, and off-site outreach that's active instead of accidental. Sprints are how you do it: every cycle tests one hypothesis with a specific experiment, not a vague initiative, so the wins compound instead of resetting.

Do you need a tool, or can you start manually?

We recommend starting manually. What you can do is open up an incognito window, work through the LLMs your buyers actually use, and ask the prompts you want to track. Don't spend money on a platform before you've done the manual work to know what you're even looking for.

While you're doing that manual test, see where you're ranking. If you're ranking first, great. Anywhere else, look into what your competitors are doing, what's making their content crawlable?

Once you've got a baseline, you need somewhere to put it. We built an AEO strategy planning worksheet for exactly this: a Google Sheet that walks you through the landscape, prompt research, a prompt audit, an audit summary, and an action plan, built around the full buyer journey from broad awareness queries down through comparison and decision.

This is how we started, too. But eventually, we outgrew the manual version and moved to looking for a platform. We landed on Profound, partly because the team there thinks about marketing the way we do, and partly because they have real prompt volume data pulled from where people are actually searching. The bigger difference is what happens after the data shows up. A lot of tools will hand you a dashboard and stop there. With Profound, you can build customized agents that turn that data and insights into actual action! It makes those actions and steps a lot clearer instead of leaving you to figure it out.

How to write prompts that match your business goals

The biggest mistake we see is fully outsourcing this to AI without giving it any real context. We're not saying don't use AI, we're saying feed it the right inputs, sales notes, keyword research, your actual goals, so what comes out matches what you're actually trying to accomplish.

You'll never find the exact words a buyer typed into an LLM, that data isn't available to any of us. But you can get pretty close. Mine your own sales calls for the questions prospects actually ask, dig through Reddit threads and forums in your space, and write prompts that sound like a person talking to an AI.

Your prompts should span the full buyer journey: broad awareness questions, then narrowing context, comparison, and decision. Our planning worksheet walks through exactly how to structure that progression.

You'll group these prompts into topic clusters. A topic cluster is a set of prompts built around one buyer problem. If you sell warehouse automation software, a cluster on "warehouse labor shortages" might run from "why can't warehouses hire enough workers" (awareness) to "is [Vendor] worth it for a 50-person warehouse team" (decision).

Start with 3 to 5 clusters covering your core use cases. For volume, 8 to 10 prompts per cluster is a reasonable place to start, higher if you've got a lot of products or pages, lower if your offering is straightforward.

What should you be tracking?

Unlike the manual test, this is about tracking through a platform, and three metrics matter.

Visibility score is your overall presence, the share of your tracked prompts you show up in at all.

Ranking is where you land within a single prompt when multiple brands get mentioned: first, second, third.

Citation share is why you're being mentioned in the first place, the share of the sources backing an answer that point back to your domain. It gets harder to read in crowded categories.

Start with your overview: your visibility score and ranking across every prompt. Then dig into individual prompts to see what's actually happening, and check citation share on the ones that matter most. As a marketing agency, we compete against martech platforms like HubSpot, not just other agencies, so we narrow our competitor tracking to the agencies we're actually up against.

Do this once before you touch anything, that's your baseline. Once you start running experiments and shipping changes, check back to see whether anything moved. Sometimes a dip isn't about you: a competitor gaining ground on the same prompts can look identical to your own visibility dropping, so check their numbers too before you assume you did something wrong.

Check this 2 to 3 times a week if you can't do it daily. The answers can change pretty quickly.

Is the data in these tools real?

Whether the data is real depends on what you're looking at. Where you stand on a given prompt, that's real because the tool actually ran it. Prompt volume is real too, but only shows part of the picture: it comes from a smaller panel of people who've opted in to be tracked, not the full universe of everyone asking so we treat it as directional. For example, if a tool shows a prompt getting searched 200 times a month, that's 200 people in its own panel, not everyone actually asking that question across ChatGPT, Perplexity, and Gemini. The real number is almost certainly higher, you just can't see by how much.

In our own numbers, we're hearing from people on our contact form that they found us through an LLM, something the tool isn't capturing at all. That gap is worth watching for in your own data too.

How to find opportunities from your data?

We usually start by looking where you're ranked lower to uncover opportunities. If you're using Profound, click into the topic to see which individual prompts are climbing and which are falling behind, the topic average hides that. If you’re checking manually, the signal is simpler: Did you fall off a list you were on last time?

For an individual prompt, look at your position and your citation rank, then run it through this framework to respond:

That's exactly how we caught one of our own opportunities. We saw ourselves climbing for "enterprise growth marketing agency," read it as Momentum, and wrote content specifically to build on it.

What do you do with years of old content?

Start with your highest-intent pages first, service pages, pricing, product overviews, the pages people actually land on to learn more about you. Those should come before anything else. 

If you've got a resource center with hundreds of posts, you can't tackle it all at once, so start with a topic. Once you have a baseline, defend what's already working, and put your energy into the high-intent pages that aren't performing the way they should. 

Picking which topics to prioritize should tie to your business goals first, then let prompt data point you to the specific opportunities inside those priorities.

Who should be doing your AEO work?

As more buyers lean on AI, having someone own this internally matters more, or having a partner who goes beyond strategy and actually owns execution. A strategy deck alone won't move your numbers, someone has to be constantly watching the data and making changes because of it. Whether that's someone on your team or an agency doing the implementation itself, there needs to be real ownership, not just advisory. Start smaller if you need to, but this eventually grows into something as necessary as your SEO function or agency. If you already have a dedicated SEO person and a large site, either bring in someone dedicated to AEO too, or train the SEO person you already have.

If you're hiring a partner, look for someone data-driven, opportunity-driven, and action-driven. They should treat AEO as a system, not a one-off strategy. There should be a framework in place that feeds into itself instead of stopping after one round of changes. And find someone who actually works in your industry, the rules change by industry.

Where to go from here?

We've spent 25 years building B2B demand gen programs, and if there's one thing that's held true through all of it, it's that things are constantly changing. AEO isn't an exception to that, it's just the newest version of it. LLMs update constantly, which means your visibility can move without you touching anything. That's exactly why you need a person watching it.

None of this has to feel as overwhelming as it looks from the outside. The best thing you can do is get a baseline and start experimenting, most of what you'll learn comes from watching what actually moves and what doesn't. That's how you get discovered, and it's how you get chosen.

If you’re ready to map your own baseline, download our AEO strategy planning worksheet and start tracking where you actually stand.