How to do AEO topic research

• Answer engine optimization (AEO) topic research starts with the same topic list you already use for SEO — you don't need a second calendar.
• Keyword volume is still your first filter. Treat it as an order of magnitude, then layer in fanouts, community questions, and citation checks.
• Fanout research shows you the follow-up questions a model asks itself after a prompt, which often point to pages you haven't written yet.
• Reddit and community research surfaces the phrasing your buyers use, before they ever open Google or ChatGPT.
• The new-versus-refresh decision works the same way for AEO as it does for SEO: check what already ranks, what could be updated, and what already wins elsewhere.

Your team already has a topic list. You built it from product marketing, competitor research, and a pile of keyword data. The question in front of you now is whether that same list works for the growing chunk of traffic coming from ChatGPT, Perplexity, and Google's AI Overviews.

It does, mostly. The topics don't change. What changes is the research you layer on top of them. Volume tells you whether people search for something. It doesn't tell you what questions a model fans out to, what your buyers ask on Reddit before they trust a vendor page, or whether your content is structured so an answer engine can lift a paragraph and cite it.

This post walks through the AEO-specific research method: the same topic list, a fresh layer of questions, and a clear call on what to write next.

Core concepts: what AEO research adds to volume

SEO topic research with Topic Explorer runs on three fields: keywords, volume, and competition. That data tells you how popular a topic is and how hard it'll be to rank for in Google. It's a solid foundation, and you shouldn't throw it out.

AEO research asks a different set of questions on top of that foundation. Instead of "how many people search this phrase," you're asking:

• What follow-up questions does a model generate after someone types a prompt related to this topic?
• What are people asking in communities like Reddit and LinkedIn, in their own words, before they land on a search engine at all?
• Which FAQs match the exact phrasing an assistant is likely to surface?
• Are you being mentioned in AI answers on this topic, or actually recommended?

None of this replaces volume. It sits on top of it. What is replacing keyword volume? makes the case plainly: keyword data is still useful for order-of-magnitude judgment calls, and it's being supplemented by community research, synthetic queries, and transcripts.

How to find AEO topics (hint: same list as SEO)

You don't need a separate topic map for AEO. The same 5–10 strategic topics that anchor your SEO content plan — the ones tied to your product, your messaging, and the problems your buyers are trying to solve — are the ones worth researching for answer engines too.

This is a deliberate choice, not a shortcut. Topics versus keywords explains why topics are the right unit of strategy: they're broad enough to support a library of content and specific enough that your audience actually lives inside them. Keywords are the individual phrases inside a topic. Fanouts and community questions are more of those phrases — sourced from AI behavior and human conversation instead of a keyword tool.

If you're starting from scratch, pull the same 5–10 topics you'd use for SEO topic research with Topic Explorer. Run volume and competition on them first, the way you always have. That step hasn't gone away. It's your intro filter, and it tells you whether anyone cares about a topic before you spend research time on it.

The workflow: four research passes on each topic

Once a topic clears the volume bar, run four passes to build out the AEO layer.

Pass one: fanouts and synthetic queries

When someone prompts an LLM with something like "best CRM for small businesses," the model doesn't stop at that one query. It runs a set of follow-up searches behind the scenes — pricing comparisons, reviews, implementation costs, integration questions — before it synthesizes an answer. These are fanouts, sometimes called synthetic queries.

Fanouts matter because they show you the sub-questions hiding inside a topic that a plain keyword list won't surface. A keyword tool might show you "CRM pricing" as a phrase with decent volume. It won't show you that a model researching "best CRM" is quietly also checking "CRM implementation cost for a 20-person team" on your behalf. That's a page you may not have written yet, and it's the kind of gap what is replacing keyword volume points to when it talks about synthetic queries from AI search tools.

Treat each topic on your list as a seed and ask what a model would fan out to next. Write those fanouts down alongside your keyword list. They belong in the same brief.

Pass two: Reddit and community questions

Fanouts tell you what a model asks itself. Reddit and LinkedIn tell you what your buyers ask each other, in their own words, often before they've typed anything into a search bar or a chat window.

The practical way to do this at scale is a workflow that ingests your topic list, searches Reddit, LinkedIn, and Google, identifies relevant threads, and extracts discrete questions from the discussion — including comments that raise a concern without phrasing it as a question. That's the approach behind customer questions on Reddit: an n8n-based flow that turns scattered community discussion into a clean list of questions.

The output needs a human pass before it goes anywhere. A workflow like this will export dozens of candidate questions, and not all of them are worth answering. Review for relevance and strategic fit, and cut anything that doesn't match what your brand is positioned to address. Don't drop the raw export into a content brief and call it done.

Pass three: FAQ extraction

Once you have a filtered list of real questions — from fanouts and from community research — the next step is deciding which ones become FAQ content on your existing pages.

FAQ content with AI lays out the logic: pages with FAQ schema tend to get cited more often by LLMs, and FAQs built from real questions serve two audiences at once, readers who scan for a quick answer and models that lift a self-contained paragraph to quote. The same discipline applies here that applies to Reddit research. Draft answers with AI if that speeds you up, but have a human edit before publishing. A list of unedited, auto-generated FAQ answers reads like it was written for a crawler, not a person, and that undermines the credibility you're trying to build.

Match the FAQs to the exact phrasing your fanout and community research turned up. A question like "what's the difference between data governance and data management" earns a citation when it sits on the page a model would already consider a candidate source for that topic.

Pass four: mention versus recommendation checks

The last pass is a reality check on where you already stand. Before you commit research time to a topic, look at whether your brand shows up in AI answers about it at all, and if it does, whether the model is mentioning you in a list or actually recommending you.

Mentioned versus recommended draws that distinction clearly: a mention is any appearance of your brand name, even in a warning or a leftover-category list. A recommendation is the model treating you as a genuine fit, with language like "worth considering" or "a strong choice." If you're showing up as a mention but never a recommendation on a topic, that's a signal about what your existing content gives the model to work with, and it should shape what you write next on that topic.

New content versus updates: the same three questions

Once you've run all four passes, you're back to a familiar decision: write something new, or update what you already have. The criteria don't change for AEO. You're checking the same three conditions used for SEO topic research:

1. The topic isn't ranking well. Nothing you've published shows up credibly, in search or in AI answers, for the questions you've surfaced.
2. Nothing could be updated. You've genuinely got a coverage gap. No existing page is close enough to the question that a refresh would fix it.
3. Nothing should be updated, because an existing page already wins a different, related query. Sometimes the right call is to leave a strong page alone and write something new and adjacent, rather than overload one URL trying to answer everything.

The library inventory you'd check for SEO — which page already ranks, which page could be refreshed — is the same inventory you check here. This is where a tool like the ércule app, pointed at Google Search Console and Google Analytics 4 data, becomes useful: it's the place you already look to see what's ranking and what's due for a refresh, and AEO research sits directly on top of that same view instead of requiring a separate system.

Next steps: question and long-tail coverage within the topic

Topic research doesn't end with a single new-versus-refresh decision. Inside any topic that clears the bar, you'll usually find a long tail of specific questions worth covering — some as standalone sections, some as FAQ additions to a page that already ranks.

Work through the fanouts and community questions you gathered and sort them by which existing page, if any, should own each one. A question with no ranking page attached is a coverage gap. A question that maps to a page ranking in position 8 is a different job: improve the page you have rather than spin up a new URL. This mirrors the same long-tail keyword research you'd do for SEO, just sourced from a broader set of signals — fanouts, Reddit threads, and FAQ candidates, instead of a keyword tool alone.

Conclusion

The trap most teams fall into with answer engine optimization is treating it like a separate discipline that needs its own topic list, its own calendar, and its own research process from scratch. The topics that matter to your buyers in Google search are the same topics that matter when they ask ChatGPT or Perplexity.

What changes is the layer of research you add on top of volume: fanouts that reveal a model's follow-up questions, Reddit and community threads that surface real phrasing, FAQ extraction that turns those questions into citable content, and a mention-versus-recommendation check that tells you where you already stand. Run those four passes against your existing topic list, apply the same new-versus-refresh criteria you already use, and you'll have an AEO research process that reuses the strategic work you've already done instead of duplicating it.

Frequently asked questions

Do I need a different topic list for AEO than for SEO?

No. Use the same 5–10 strategic topics you already research for SEO. Topics versus keywords explains why topics work as the shared unit of strategy — they're broad enough to serve both search engines and answer engines without splitting your content plan in two.

Is keyword volume still worth checking before I do AEO research?

Yes. Volume remains your intro filter for whether a topic is worth researching further. Treat it as an order of magnitude rather than an exact count, then layer on fanouts and community questions, as described in what is replacing keyword volume.

What's a fanout, and why does it matter for AEO?

A fanout is a follow-up search a model runs behind the scenes after a user's original prompt. A query like "best CRM for small businesses" might fan out to pricing comparisons or implementation costs. These fanouts often point to sub-topics your keyword list hasn't caught yet.

How do I find the questions my buyers are actually asking?

Search Reddit and LinkedIn for the language your audience uses when discussing your topic, ideally with a repeatable workflow rather than manual searching. See customer questions on Reddit for one approach, including why a human review step matters before those questions become content.

Should I turn every Reddit question into an FAQ?

No. Filter for relevance and strategic fit first. FAQ content with AI walks through drafting FAQ answers with AI, then editing before publishing, so you avoid dumping unedited, low-quality FAQs onto a page.

What's the difference between being mentioned and being recommended by an AI answer?

A mention is any appearance of your brand name in a model's response, even a warning or a leftover-category listing. A recommendation is the model treating you as a genuine fit for the question. Mentioned versus recommended breaks down how to tell the two apart.

How do I decide between writing something new and updating an existing page?

Check three things: whether the topic is ranking well already, whether an existing page could be updated to cover the gap, and whether an existing page already wins a different related query and shouldn't be touched. This is the same decision framework used in SEO topic research with Topic Explorer, applied to AEO topics too.

Where do I start if I'm new to AEO?

Start with your existing topic list and a baseline check of which pages are already getting cited in AI answers. Getting started with AEO covers the diagnostic steps and how to treat AEO as a subsystem of your broader content system, rather than a rebuild.

Related reading

• How to do SEO topic research with Topic Explorer — read this first for the core volume-and-competition method that AEO research builds on.
• How Can AI Help With Topic Research for SEO and GEO? — the sibling post on using AI to gather demand signals; this piece is the AEO research method itself.
• Getting started with AEO — useful if you haven't set up baseline diagnostics for AI citations yet.