AEO Playbook: 8 Decisions Plus the Reporting Layer Most Skip

Your potential buyer asks ChatGPT for the best vendor in your category. Your name is in the answer or it isn’t. If it isn’t, they’ll never even know you exist.

That’s the whole channel in one screen. Answer Engine Optimization (AEO) is the practice of getting your brand cited when prospects ask ChatGPT, Perplexity, Gemini, or Google AI Mode for the best options in your category. Citations are the new rankings. The brand named in the answer is usually the one that wins the deal.

For most B2B companies we work with, AEO stopped being a fringe channel a year ago. As of August 2026, citations routinely influence deals before the prospect ever loads a landing page. That’s a pipeline mechanic, not a brand-awareness one. It compounds quietly into shorter sales cycles and warmer demos. Tools like Lectern exist to track exactly that compounding — which prompts you appear in, which surfaces are driving share-of-voice, and which placements actually moved the needle when they landed.

Ethan Smith, CEO of Graphite, recently laid out the most useful operator-level AEO playbook we’ve heard. Graphite ran SEO for Webflow, Notion, Rippling, and Upwork. They’re now running AEO for the same kind of late-stage operators. We’ve been pressure-testing his frameworks against what we see across our own citation campaigns and reporting dashboards for the last 4 months.

We’re breaking down what we’ve learned here. First, we’ll distill the playbook into the eight decisions you actually need to make. Second, we’ll flag one thing the playbook understates: the reporting layer. The companies winning AEO right now treat citation reporting like a paid-channel dashboard, not a vanity report.

Ethan walked through this in a follow-up deep dive with Niklas Buschner, CEO of Radyant, building on his earlier appearance on Lenny’s Podcast. Watch the full conversation on YouTube. The summary below is our read mixed with our experience.

AEO vs SEO at a glance

AEO and SEO aren’t opposites. AEO is a superset of SEO with a different end-state. Both rely on similar trust signals (domain authority, editorial legitimacy, structured content), but the goal and success metric diverge:

Aspect	SEO	AEO
Goal	Rank in search results	Be cited inside AI answers
Success metric	Position, clicks, traffic	Mentions, share of voice, citation rate
Primary lever	On-site content + backlinks	Off-site citations (Reddit, YouTube, G2, editorial media placements)
Query length	~6 words	20 to 25 words, conversational
User behavior	Skim ten blue links	Read the answer; rarely click
Time-to-result	6 to 12+ months	2 to 3 months for measurable lift
Best signal	Backlinks from authoritative domains	Mentions across diverse, trusted sources

The practical takeaway: if you only chase rankings, you’ll be invisible inside the AI answer that ate your funnel. If you only chase citations, you’ll miss the on-site conversion paths once the prospect does click through. The eight decisions below are how operators run both.

The mechanic underneath: query fan-out

Every decision below sits on top of one practical fact: broad buyer prompts often get resolved through a set of related research angles, not one exact-match keyword. The system may look for comparisons, pricing context, use cases, reviews, category definitions, and third-party mentions before assembling an answer with named sources or citations.

That fan-out is the actual unit of work. You’re not optimizing for “best meeting transcription tool” — you’re optimizing for the cluster of sub-queries the model expanded that prompt into (“transcription tools that integrate with Zoom”, “Otter vs Fireflies pricing”, “AI note-taker pros and cons”, and so on). Three implications fall out of this:

• Multiple paths beat one perfect page. A single home-run blog post might catch the head prompt and miss the long tail of fan-outs entirely. The Otter example in section 4 is a fan-out play disguised as a content play.
• Off-site matters because the fan-outs hit different surfaces. Some sub-queries get answered by Reddit threads, some by G2 categories, some by YouTube reviews, some by editorial media placements. Citations have to live across the fan-out spread, not just one channel.
• Indexing and authority both still matter. If a piece of content isn’t indexed (or isn’t being linked to from somewhere indexed), it won’t show up in any sub-query, which means it won’t show up in any answer. AEO doesn’t override SEO mechanics — it sits on top of them.

Hold this in mind for everything below. Most “AEO best practices” make sense once you remember the model is searching a hundred small things, not the one big thing the user typed.

1. Decide if AEO is the right channel for your stage

Stage matters more than people admit:

• Seed or early stage: skip SEO entirely. Your domain has no authority and you’ll wait two years for nothing. Invest lightly in AEO instead. Citations from Reddit, G2, and editorial media placements can land tomorrow even with no domain authority.
• Mid-stage: AEO becomes a real channel. Probably your fifth most important.
• Late stage: AEO is one of the few genuinely incremental channels left after you’ve saturated search and ads.

Why early-stage companies can still win: citations don’t care as much about your domain. ChatGPT will cheerfully tell a prospect that you’re the best vendor in your category if Reddit, G2, and a couple of editorial media placements say so. Your homepage doesn’t need to outrank anyone.

2. Split AEO work into on-site and off-site

These are two different jobs.

On-site is traditional SEO mechanics: pages targeting prompts, technical hygiene, content with unique information gain.

Off-site is the new battleground: getting cited on Reddit, YouTube, G2, LinkedIn, Instagram, and editorial media placements. Ethan’s framing: “When you ask ChatGPT what’s the best credit card, the citations are the answer. If you’re in the citations, you’re suddenly the best credit card.”

Most companies under-invest in off-site because nobody owns it. Reasoning by analogy from SEO is wrong here. In SEO, link building is a side project. In AEO, off-site work is often higher impact than on-site.

3. Hire AEO operators for two skills (and only two)

Ethan looks for exactly two things in AEO hires.

Experiment design and analysis. Most “best practices” in SEO and AEO are wrong because nobody actually tested them. AEO is too new for established playbooks. Hire someone who can run a clean A/B and read the result.

Comfort with ambiguity. “Go figure out a Reddit strategy” is an open-ended problem. Most operators can execute a runbook. Far fewer can solve the open-ended version.

This is a hiring criterion, not a job description. Adjust your interview accordingly.

4. Get cited on the surfaces LLMs already pull from

The mental model: find the surfaces LLMs already trust, and earn citations there.

Ranked by impact:

• Reddit: massive input to LLMs, both as cited content and as training data
• YouTube: already big, and getting much bigger as models extract semantic context from video
• G2 and review platforms: clean citation source for “best X for Y” prompts
• LinkedIn, X, Quora: diverse opinions, which LLMs reward
• Editorial media placements: the reliable workhorse
• Integration partner pages: the under-rated one

The Otter example is worth memorizing because it’s a fan-out play in disguise. To show up for “what meeting transcription tool integrates with Zoom?”, Otter doesn’t write one page. The prompt fans out into sub-queries about Zoom integrations, transcription pricing, comparison reviews, and how-to walkthroughs. Otter builds a citation path into each branch:

1. A help center page about the Zoom integration (catches the how-to fan-out)
2. A feature page like /zoom (catches the integration listing fan-out)
3. A how-to article (catches the long-tail walkthrough fan-out)
4. A page about the integration on Zoom’s own site (catches the third-party authority fan-out)

Four angles, one answer. Whichever sub-query the model expands the prompt into, Otter is sitting in the top results.

5. Video is about to dominate AEO citation sources

Ethan’s bet: video will eventually overtake Reddit and X as the top input source for LLMs.

Why: most of the unique information in a video is locked. Today’s models use the title, description, and sometimes the transcript. They miss the actual content of the conversation. That’s changing. Vimeo just shipped semantic search inside videos. Once LLMs can extract concepts from video the way they extract them from a Reddit post, the corpus of unique video content becomes massive, and almost none of it is rewritten from someone else’s article.

Three practical moves to lock in citation share before that shift fully lands:

1. Start producing original video now. Founder interviews, product walkthroughs, expert panels. The asset compounds before the channel does — by the time semantic video search is the default, the operators with two years of indexed video content will have a moat that AI tooling can’t shortcut.
2. Optimize transcripts and chapters as if they were blog posts. Models that can’t read the video frame-by-frame are still ingesting transcripts and YouTube chapter markers today. Add headers, structured Q&A, and named-entity callouts inside the transcript so the chunk that gets cited is the one you wrote.
3. Cross-publish video onto your owned domain. A YouTube embed on a /resources/[topic] page with a clean transcript catches both the YouTube fan-out (where the model trusts video as a source) and the indexable-page fan-out (where it trusts your domain). One asset, two citation paths.

6. AI-generated content only works with unique input

The “AI heist” approach (point ChatGPT at competitors, generate a thousand pages, ship) fails. Hard.

But AI content itself isn’t the problem. The rule:

AI plus unique input works. AI plus derivative input doesn’t.

Graphite’s Sermo example is the cleanest version of this. Sermo is a closed community where doctors share which drugs worked for which patients. That data isn’t on the open web. Graphite uses AI to summarize it into landing pages. Those pages rank right under WebMD.

If you’re going to use AI at scale, feed it something the open web doesn’t have. Customer support transcripts. Sales call insights. Proprietary metadata. Product behavior. Then have it summarize. If your AI workflow is just a wrapper around ChatGPT with no unique input, you’re shipping noise.

7. Track three AEO metrics, not thirty

The four AEO tool features that matter: visibility tracking, prompt volume, content workflows, content scoring. Everything else is dashboard theater.

Visibility tracking. Do you appear when people prompt? Two notes that operators get wrong:

• LLM answers are probabilistic, not personalized. The model samples from a distribution. Ask the same prompt 7 to 10 times to see the real distribution before you call it visibility.
• API data and logged-in scraped data give different citations. Logged-in is what real users see.

Tools Ethan currently uses: AirOps, Profound, Surfer SEO, Evertune. We use Lectern, which we built to fill the parts of the stack we kept hitting walls on — probabilistic sampling per prompt, snapshot-to-snapshot compare across providers, and tying citation events back to pipeline lift.

Prompt volume. No LLM publishes this data. The triangulation hack we use too: pull a competitor’s top paid keywords from Ahrefs, drop them in Claude, ask it to convert each into a natural-language prompt. That’s your starter list. As a rough planning heuristic from our own checks, expected prompt volume is often far lower than search volume; if a tool is showing AI prompt volumes close to keyword search volume, treat the panel data as directional until you validate it against referrals and conversions.

How many prompts to track:

• Early stage: 10 to 15
• Enterprise: 1,000 to 5,000

Group prompts by topic, target ten prompts per page, build a visibility matrix.

8. AEO attribution is messy. Build the channel anyway.

Most AEO conversions are no-click. A buyer asks ChatGPT for the best vendor in your category, sees your name, opens a new tab, types your brand into Google, clicks your brand-search ad. The conversion gets logged as branded paid search. The actual cause was the citation.

Ethan’s advice, and ours: focus on the beginning and the end.

• Beginning: are you appearing for the prompts you care about? (Visibility tracking.)
• End: ask new customers “how did you hear about us?”

Skip the obsession with mid-funnel attribution. The data isn’t there. The channel still works.

The one thing the AEO playbook understates: citation reporting

Here’s where I’d add to Ethan’s frame.

Most companies treat citation reporting as a vanity layer. Screenshot the win, post it in Slack, move on. The companies who actually compound in AEO treat reporting as the operating dashboard for the channel itself. They run weekly visibility reports against tracked prompts, segment by buyer persona, and run before/after analysis when a new editorial media placement lands or a Reddit thread takes off.

What we saw in a directional Presscart/Lectern sample of B2B visibility checks in Q1 2026:

• Reddit appears in roughly 40% of citations for category-defining “best [X] tool” prompts — the single most-cited surface for B2B AEO.
• Editorial media placements show up in roughly 25% of citations and disproportionately drive the first-mention slot inside the answer (not just a footer cite).
• G2 and review platforms appear in about 20% of “best of” prompts but underperform editorial for thought-leadership and category-definition prompts.
• Brands cited on 3+ distinct surface types (e.g., Reddit + editorial + G2) had 2x to 4x the share-of-voice of brands cited on just one. Diversity of citation surface beats volume on any one surface.

Two things fall out of doing reporting well:

1. You learn which citation surfaces convert. Not all citations are equal. A G2 mention often outperforms a Forbes link for high-intent prompts. The only way to know is to track the visibility delta when each one lands.

2. You can connect citations to pipeline. Tag the citation event, log the resulting visibility lift, overlay the inbound demo curve. You stop guessing. Sales can finally tell you which articles their prospects read before they booked the call.

This is the analytics layer of AEO and most teams haven’t built it yet. It’s the gap we kept hitting on every engagement, so we built Lectern to close it. Lectern runs scheduled visibility checks across ChatGPT, Claude, Perplexity, and Gemini, samples each prompt repeatedly to cut through the probabilistic noise, scores mention/position/sentiment per query × provider, and lets you compare snapshots before and after a placement so you can see which surfaces actually moved share-of-voice. It’s the operating dashboard we wished existed when we started running AEO for clients.

It’s also the reason editorial placements compound: every Presscart placement is a paid editorial placement that adds another third-party source for Google and AI search to retrieve, which we’ve broken down in How Third-Party Sources Help ChatGPT Understand Your Brand. Indexing, rankings, and AI citations remain independent platform decisions.

Frequently asked questions about AEO

What is Answer Engine Optimization (AEO)?

Answer Engine Optimization (AEO) is the discipline of optimizing your brand and content to be cited when AI assistants like ChatGPT, Claude, Perplexity, Gemini, and Google AI Mode answer questions about your category. The unit of success is being mentioned inside the answer (or cited as a source), not ranking on a list of blue links. AEO leans heavily on off-site signals like Reddit, YouTube, G2, integration partner pages, and editorial media placements, because LLMs synthesize answers from the consensus across those surfaces.

How is AEO different from SEO?

SEO competes for position on a search engine results page; AEO competes for inclusion inside an AI-generated answer. SEO is won primarily on-site (page structure, on-page content, backlinks). AEO is won primarily off-site (citations on Reddit, YouTube, review sites, and editorial media placements). Every signal that helps SEO also helps AEO, but the reverse isn’t true: a #1 ranking doesn’t guarantee you’ll appear inside ChatGPT’s answer. Treat AEO as a superset of SEO, not a replacement for it.

How long does AEO take to show results?

Most operators see measurable shifts in AI visibility within 2 to 3 months. Citation work, like landing on Reddit threads, getting included in G2 categories, or earning a mention in a trusted publication, can move visibility within weeks if the surface already feeds the model. The compounding effects (citations begetting citations, models updating their training corpus) play out over 6 to 12 months. Anyone promising overnight results is selling something else.

What tools should I use to track AEO?

The four tool categories that matter are visibility tracking, prompt volume estimation, content workflows, and content scoring. Tools Ethan currently uses include AirOps, Profound, Surfer SEO, and Evertune; we built Lectern to fill the gaps we kept hitting — probabilistic sampling per prompt, snapshot-to-snapshot compare across providers, and tying citation events back to pipeline lift. Two practical rules: (1) Run each prompt 7 to 10 times because LLM answers are probabilistic, and (2) Use logged-in scraped data, not just API data, because real users see different citations than the API returns.

Do early-stage companies need to invest in AEO?

Yes, more than they need to invest in SEO. Domain authority takes years to build, so a seed-stage company shipping SEO content is wasting time. AEO citations don’t care about your domain. ChatGPT will cheerfully recommend you if Reddit, G2, and a couple of editorial media placements already do. That makes AEO one of the few channels where early-stage companies can compete with incumbents on day one.

How do I know if my AEO efforts are working without click attribution?

Track the beginning and the end. At the beginning: are you appearing for the prompts you care about? Run a visibility check across ChatGPT, Claude, and Perplexity weekly. At the end: ask new customers “how did you hear about us?” Most AEO conversions are no-click (a prospect asks ChatGPT, types your brand into Google, clicks your brand-search ad, and the conversion logs as paid search). Skip the obsession with mid-funnel attribution. The data isn’t there, but the channel still works.

Will AEO replace SEO?

No. AEO is a superset of SEO, not a replacement. The retrieval layer underneath ChatGPT, Perplexity, and Google AI Mode is still mostly Google-shaped, so the same trust signals (indexability, domain authority, structured content, backlinks) feed both. What’s changing is the end state: instead of competing for a position on a results page, you’re competing for inclusion inside the AI-generated answer. Companies that abandon SEO to chase AEO will lose both. Companies that treat AEO as the layer above SEO will win both.

How do I rank in ChatGPT?

You don’t rank in ChatGPT — you get cited inside its answers. Three moves do most of the work as of August 2026: (1) Earn citations on the surfaces ChatGPT already pulls from (Reddit, YouTube, G2, editorial media placements, integration partner pages); (2) Build help center and comparison content that captures the long-tail sub-queries ChatGPT fans your prompt out into; and (3) Run probabilistic visibility checks (7 to 10 samples per prompt) across providers so you can see which placements actually moved share-of-voice. Domain authority matters less than diversity of trusted citations.

What are the best practices for answer engine optimization?

The eight that consistently move citation rate: (1) Decide AEO is the right channel for your stage before investing; (2) Split work into on-site (content, structure) and off-site (Reddit, YouTube, G2, editorial media); (3) Build multiple citation paths per prompt to catch sub-query fan-out; (4) Use AI-generated content only with unique input you own; (5) Track 7 to 10 samples per prompt because LLM answers are probabilistic; (6) Use logged-in scraped data, not just API data; (7) Track three metrics (visibility, share-of-voice, citation source mix) instead of thirty; (8) Build the reporting layer that ties citations to pipeline.

The eight decisions, in order

1. Early stage? Invest lightly in AEO via off-site. Skip SEO.
2. Split on-site and off-site work. Off-site is where the impact is.
3. Hire for experiment design and ambiguity tolerance. Not “10 years of SEO.”
4. Get cited where LLMs already pull from. Build multiple citation paths per prompt.
5. Start producing original video. The asset compounds before the channel does.
6. Use AI content only with unique input. Otherwise ship nothing.
7. Track 7 to 10 samples per prompt. Use AirOps, Profound, Surfer, or Evertune.
8. Build the reporting layer. It’s the gap most teams haven’t filled.

Most companies will follow the checklist. The ones who win will run real experiments and build the dashboard.

That reporting layer is what we run for the companies we work with: paid editorial placements via Presscart on the surfaces LLMs already draw from, paired with Lectern — the visibility dashboard that tells you which placements actually moved the needle. The Presscart marketplace is open at app.presscart.com, and you can spin up a Lectern visibility report at trylectern.com. Or read more about how our SEO and linkbuilding program is built to score across both Google and AI search.