LLenny's Podcast
← All episodes
Ethan Smith (Graphite)14 September 2025

The ultimate guide to AEO: How to get ChatGPT to recommend your product

5Frameworks
15Insights

Episode overview

Lenny Rachitsky interviews Graphite CEO Ethan Smith about answer engine optimization and how it overlaps with, yet extends, traditional SEO. Smith argues that brands can improve their visibility by answering long-tail questions on their own sites and earning mentions across citations such as YouTube, Reddit, affiliates, and blogs. He advocates tracking share of voice across answer engines and using controlled, reproducible experiments, while warning that fully automated AI content and many asserted AEO best practices are unreliable.

Key ideas

  • AEO and GEO describe optimizing a brand or product to appear in answers generated by LLM-based services.
  • Traditional SEO remains foundational, but AEO additionally rewards repeated mentions across the sources an answer engine cites.
  • Early-stage companies can compete quickly through citation coverage and highly specific long-tail answers without first building substantial domain authority.
  • LLM-referred visitors can be unusually qualified; Webflow reportedly saw a sixfold conversion-rate difference versus Google Search traffic.
  • AEO measurement should track repeated prompt runs, question variants, multiple answer engines, citation presence, and average rank as a share of voice.
  • Teams should test tactics with untreated control questions, pre- and post-intervention tracking, and repeated experiments rather than trusting unverified best practices.
  • Help centers can capture obscure product, feature, integration, and use-case questions when they are crawlable, well cross-linked, and filled out across the long tail.
  • AI-assisted human editing is presented as useful, while fully automated AI-generated content is associated with weak search visibility and risks such as derivative convergence and model collapse.

Transcript available · source text is retained privately and is not published

Frameworks in this episode

Marketing3 steps

Head vs Tail in Answer Engines

Why AEO rewards citation frequency and hyper-specific long-tail questions

Read →
Marketing3 steps

Help Center Optimization for the Tail

Turn your support docs into the citation source for obscure use-case questions

Read →
Marketing7 steps

The 7-Step AEO Execution Plan

A repeatable pipeline to make your product show up in LLM answers

Read →
Marketing3 steps

The Authentic Reddit Citation Strategy

Win Reddit citations by being a real, disclosed user — not a spam farm

Read →
Communication3 steps

The Information-Gain Content Heuristic

Judge content by what new it adds, not how typical it is

Read →

People & resources mentioned

Attributed to the moment in the episode. Timestamps are approximate.

People · 6

  • Lenny RachitskyMentionstoday we've got another very special compilation episode something I've been pulling on more and more with the podcast and the newsletter

    “Today, my guest is Ethan Smith. Ethan is the CEO of Graphite and my go-to expert for all things SEO.”

    ▶ Watch at ~1:00approx.
  • Ethan SmithMentionsone of the most innovative disciplined first principles seo thinkers and i've met is ethan smith from graphite

    “Ethan is the CEO of Graphite and my go-to expert for all things SEO.”

    ▶ Watch at ~1:00approx.
  • Michael JordanMentionsMichael Jordan used to say that you know I've failed over and over and over and over again

    “I really like Michael Jordan documentary, Last Dance.”

    ▶ Watch at ~1:05:00approx.
  • Lance ArmstrongMentionsI like Lance Armstrong documentaries about how aggressive and confrontational he is.

    “I like Lance Armstrong documentaries about how aggressive and confrontational he is.”

    ▶ Watch at ~1:05:00approx.
  • Martha StewartMentionsthey did not have nearly as much authority as all recipes and Martha Stewart

    “they did not have nearly as much authority as all recipes and Martha Stewart”

    ▶ Watch at ~1:08:00approx.
  • Brian JohnsonMentionsIt's something Brian Johnson, I noticed, is really good at on Twitter, the longevity guy.

    “It's something Brian Johnson, I noticed, is really good at on Twitter, the longevity guy.”

    ▶ Watch at ~1:11:00approx.

Resources · 18

  • GraphiteMentionscompany

    “Ethan is the CEO of Graphite and my go-to expert for all things SEO.”

    ▶ Watch at ~1:00approx.
  • ConductorMentionssoftware · Orkes

    “the company behind Open-source Conductor, the orchestration platform powering modern enterprise apps and agentic workflows.”

    ▶ Watch at ~2:30approx.
  • VantaMentionssoftware · Christina Cacioppo

    “Vanta's trust management platform makes it quicker, easier, and more scalable.”

    ▶ Watch at ~4:00approx.
  • Great QuestionMentionssoftware · Great Question

    “This episode is brought to you by Great Question, the all-in-one UX research platform”

    ▶ Watch at ~28:00approx.
  • Google SearchMentionssoftware · Google

    “Google's slice of the pie stays the same. The pie gets bigger.”

    ▶ Watch at ~48:00approx.
  • Model collapse paperMentionspaper

    “There's a paper about this called model collapse.”

    ▶ Watch at ~56:30approx.
  • OtterRecommendssoftware · Otter.ai

    “But you could use Otter because Otter has a Zapier integration.”

    ▶ Watch at ~1:02:00approx.
  • ZapierUsessoftware · Zapier, Inc.

    “But you could use Otter because Otter has a Zapier integration.”

    ▶ Watch at ~1:02:00approx.
  • BigQueryUsessoftware · Google

    “You could send a zap of the meeting, put it into BigQuery and then do Looker on top of that.”

    ▶ Watch at ~1:02:00approx.
  • Emotional IntelligenceRecommendsbook · Daniel Goleman

    “Number one is Emotional Intelligence.”

    ▶ Watch at ~1:03:30approx.
  • How to Measure AnythingRecommendsbook · Douglas Hubbard

    “And then the last is How to Measure Anything.”

    ▶ Watch at ~1:04:30approx.
  • The Last DanceRecommendsdocumentary · ESPN Films and Netflix

    “I really like Michael Jordan documentary, Last Dance.”

    ▶ Watch at ~1:05:00approx.
  • Sony mirrorless cameraRecommendsproduct · Sony

    “getting a mirrorless SLR with a wide angle lens really transforms your video calls.”

    ▶ Watch at ~1:06:30approx.
  • OutliersMentionsbook · Malcolm Gladwell

    “there's the Outliers book about 10,000 hours”

    ▶ Watch at ~1:07:00approx.
  • LinkedInUseswebsite · LinkedIn

    “I spend lots of time on LinkedIn and I publish original.”

    ▶ Watch at ~1:09:00approx.
  • LinkedInRecommendswebsite · LinkedIn

    “So follow me on LinkedIn, add me on LinkedIn, send me a message.”

    ▶ Watch at ~1:09:30approx.
  • GraphiteMentionscompany

    “Also, just to point people to your domain, graphite.io, is that the right domain?”

    ▶ Watch at ~1:11:00approx.
  • Lenny's PodcastCoinedpodcast · Lenny Rachitsky

    “You can find all past episodes or learn more about the show at lennispodcast.com.”

    ▶ Watch at ~1:11:30approx.

Spot an error or want something removed? Request a correction or removal.

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Myth Buster· 4

Myth Buster05:30

AEO Is the Second Biggest Change to SEO, Not a Total Reinvention

Many people treat AEO as a moment where everything they've done before stops working. Ethan pushes back: after 18 years in SEO, he calls this the second biggest shift, behind Google's Panda-era crackdown on autogenerated spam pages. It's related to search, just with summarization and new inputs layered on top.

  • Ethan started in SEO in 2007 (18 years of experience)
  • The biggest-ever change was Google killing autogenerated spam pages (Panda)
  • AEO is a summarization of search with new inputs, not a clean break from it

“I think that the main thing here is it is related to search but it's a summarization of search and”

Ethan Smith · 05:30
#aeo#seo#history#myth
Myth Buster17:30

Reddit Spam Doesn't Work — Just Be a Real User

Reddit is one of the most cited sources in LLMs, so growth teams' instinct is to spin up hundreds of fake accounts and mass-post. Ethan says that obvious hustle-mindset strategy fails: accounts get banned and comments deleted because the community polices itself. The strategy that works is unglamorous and doesn't need scale.

  • Reddit is heavily cited in LLM answers and community-moderated
  • Fake-account spam gets banned and deleted, so it doesn't work
  • Winning move: real account, say who you are and where you work, give a useful answer
  • Even five genuine comments can be enough — no need for thousands

“Fortunately, that doesn't work very well, but that's the obvious strategy.”

Ethan Smith · 17:30

“Make an account, say who you are, say where you work, and give a useful answer.”

Ethan Smith · 18:30
#reddit#spam#citations#authenticity
Myth Buster54:30

Fully AI-Generated Content Doesn't Work — He Ran the Study

Ethan's team ran a rigorous study: thousands of Google searches and ChatGPT questions, run through a validated AI detector (they proved its accuracy with generated articles and a common-crawl false-positive test of ~8%). Only 10–12% of content ranking in Google and ChatGPT was AI-generated; 90% was not. His conclusion: AI-assisted, human-edited content is the future, but 100% automated content with no human in the loop does not work.

  • Validated the AI detector against generated articles and pre-ChatGPT common-crawl data (~8% false positive rate)
  • Only 10–12% of top-ranking content in Google and ChatGPT is AI-generated
  • AI-assisted, human-edited content is clearly the future
  • Purely 100% AI-generated content does not work

“around 10 to 12% of content in Google search and in chat GPT are AI generated 90% are not”

Ethan Smith · 54:30
#ai-content#study#content-quality#seo
Myth Buster48:00

Google Isn't Dying — The Pie Just Gets Bigger

Ethan flags heavy misinformation in AEO, starting with the recurring 'Google search is dying' headline. Every couple of years a new channel (TikTok search, Instagram, Facebook, YouTube) is declared Google's killer, but each adds a new channel on top rather than taking share. He cites Google's VP of search stating traffic to publishers is not down — it's up slightly.

  • The 'Google is dying' story recurs every ~2 years with a new challenger
  • New search surfaces add to the pie rather than taking Google's slice
  • Google's VP of search said publisher traffic is up slightly, not down
  • AEO has an unusually high ratio of misinformation to correct information

“Google's slice of the pie stays the same. The pie gets bigger.”

Ethan Smith · 48:00
#misinformation#google#seo#myth

Hot Take· 2

Hot Take12:30

Early-Stage Companies Can Win AEO Immediately (Unlike SEO)

Ethan's standard advice to startups asking about SEO is: don't do it, you lack the domain authority to rank for months or years. AEO flips that. Because a single citation can surface you in an answer right away, brand-new companies can start showing up the day after launch through a Reddit thread, a YouTube video, or a blog mention.

  • SEO usually can't pay off until Series A/B due to domain authority requirements
  • AEO has no such delay because one citation can surface you immediately
  • A brand-new YC company can appear in answers the day after launch

“you can get mentioned by a citation tomorrow and start showing up immediately”

Ethan Smith · 12:30

“So, early stage companies can win. They can win quickly”

Ethan Smith · 12:30
#aeo#startups#domain-authority#citations
Hot Take41:00

Your LLM Traffic Is Hiding in 'Direct' and 'Branded Search'

Because most B2B answers aren't clickable, people who discover you in an LLM often open a new tab and type your brand into Google, or go straight to your domain. You then misattribute that visit to branded search or direct traffic — badly underestimating the answer-engine channel. To measure it, you must track answer visibility and ask users how they heard about you post-conversion.

  • Users open a new tab and type your brand into Google — looks like branded search
  • Or they type your domain directly — looks like direct traffic
  • Last-touch referral traffic understates the real AEO impact
  • Measure via answer tracking plus a 'how did you hear about us' survey

“they will open a new tab and they will type in the brand name and they will go to Google and then they'll click on…”

Ethan Smith · 41:00
#attribution#traffic#measurement#b2b

Explainer· 6

Explainer06:30

AEO and GEO Are the Same Thing (And Why He Prefers AEO)

Ethan defines AEO (answer engine optimization) and GEO (generative engine optimization) as two names for the same discipline: getting your product to show up as the answer inside LLMs. He prefers 'answer engine optimization' because 'answer' is more narrowly and accurately defined than 'generative', which can also mean images or video.

  • AEO and GEO describe the same thing: showing up in LLMs as an answer
  • He prefers AEO because 'answer' is narrower than 'generative'
  • The definition of the term is whatever the community collectively agrees on

“AEO and GEO are are essentially trying to describe the same thing, which is how do I show up in in LMS as an answer?”

Ethan Smith · 06:30
#aeo#geo#definitions#llm
Explainer11:00

Getting Mentioned Most Beats Ranking #1 in AEO

In Google, winning the top blue link wins the click. In an LLM, that's not how it works: the model summarizes many citations, so the brand mentioned most often across those citations tends to become the answer. Ethan explains this is why the 'head' of AEO differs fundamentally from SEO.

  • LLMs summarize many citations rather than picking one top result
  • Ranking a URL #1 in citations does not win the answer by itself
  • The tool mentioned most across citations usually leads the answer

“the LM is summarizing many citations. And so you need to get mentioned as many times as possible.”

Ethan Smith · 11:00
#aeo#citations#ranking#llm
Explainer20:30

AEO Works on RAG, Not the Model's Training Data

Ethan separates two layers of an LLM: the core model (next-word prediction from training data) and RAG (retrieval-augmented generation, a live search that gets summarized). Almost everything in AEO targets the RAG layer, because influencing the core model is extremely hard and slow. The controllable, high-leverage surface is the search-and-cite side.

  • Core model = next-word prediction (e.g. capital of California = Sacramento)
  • RAG = do a search, then summarize the search results
  • Influencing the core model is extremely hard and may only show up a year later
  • AEO focuses on RAG because that's the controllable layer

“most of what I'm describing is about the rag piece not the core model piece”

Ethan Smith · 20:30
#rag#llm#training-data#technical
Explainer25:30

1 in 20 Landing Pages Drive 85% of Your Traffic

Ethan shares an analysis showing that the vast majority of landing pages produce almost nothing: one in twenty pages drives roughly 85% of all traffic. This is why the industry cheaply mass-produces mediocre pages — and why the smarter play is to identify the few pages that will actually work and pour resources into writing them well.

  • 1 out of 20 landing pages drives ~85% of all traffic
  • 19 out of 20 pages drive little to no traffic
  • This drives the cheap 'get a non-expert to rewrite it' economics
  • Better ROI: find the few winning pages and write them exceptionally well

“one out of 20 landing pages drive roughly 85% of all your traffic. So 19 out of 20 landing pages drive little to no traffic.”

Ethan Smith · 25:30
#landing-pages#roi#power-law#seo
Explainer57:30

Model Collapse: Why AI Summarizing AI Converges on Vanilla

Ethan explains what happens if AI-generated derivatives feed the loop. LLM answers rely on 'wisdom of the crowd' — averaging thousands of opinions beats any single expert. But if you generate derivatives, rank them, summarize, and repeat, that diversity collapses into a single opinion. Ask for the best ice cream flavor and it converges on 'only vanilla' — a small illustration of a broadly bad outcome.

  • LLM answers depend on wisdom of the crowd — diverse opinions beat one expert
  • An 'infinite loop of derivatives' shrinks that diversity to a single opinion
  • Example: best ice cream flavor collapses to 'only vanilla'
  • A referenced paper on 'model collapse' shows hallucinations when models train on their own derivatives

“if you feed in derivatives of derivatives into the model, you will basically take the wisdom of the crowd and that will shrink and you'll…”

Ethan Smith · 58:00
#model-collapse#ai-content#wisdom-of-crowd#llm
Explainer1:01:00

Optimize Your Help Center for the Long Tail of Questions

Ethan's most exciting under-discussed tactic: help-center optimization. Because people ask LLMs very specific 'can your product do X?' questions, help centers are a great place to answer them. Move the help center from a subdomain to a subdirectory, cross-link pages heavily, and fill in the tail of obscure use cases — where you may be the only citation and can win by default.

  • LLM users ask specific feature/use-case/integration questions
  • Move the help center from a subdomain to a subdirectory
  • Cross-link help-center pages to each other
  • Fill the tail from sales-call and support questions; you may be the only citation

“subdomains don't work”

Ethan Smith · 1:01:00
#help-center#long-tail#aeo#tactics

Story· 1

Story14:30

Webflow Saw LLM Traffic Convert 6x Better Than Google

Asked whether answer-engine leads are actually valuable, Ethan shares a concrete result: at Webflow, LLM-driven traffic converted six times better than Google search traffic. He attributes the quality to priming: users arrive after a multi-turn conversation that has already narrowed their intent.

  • Webflow: 6x higher conversion rate from LLM traffic vs Google search
  • Higher intent comes from multi-turn conversations before the click
  • LLM leads are significantly more qualified for B2B SaaS

“Web Flow, we saw a 6x conversion rate in uh difference between LLM traffic and uh Google search traffic.”

Ethan Smith · 14:30
#conversion#webflow#b2b#lead-quality

Tool· 1

Tool1:03:00

Ethan's Book Picks: Psychology Frameworks for Growth

In the lightning round, Ethan recommends books whose psychology frameworks transfer directly to growth work: Emotional Intelligence (the foundational research, not the buzzword), Robert Cialdini's persuasion books, Thinking, Fast and Slow, and How to Measure Anything — which argues even hard-to-quantify things (like a conductor's skill, via standing ovations) can be measured.

  • Emotional Intelligence — foundational psychology research behind the buzzword
  • Robert Cialdini's persuasion books — frameworks for getting people to sign up/buy
  • Thinking, Fast and Slow — behavioral economics applied to growth tests
  • How to Measure Anything — measuring things that aren't obviously measurable

“if you just read thinking fast and slow, persuasion, emotional intelligence, you can basically take those frameworks and”

Ethan Smith · 1:04:00
#books#growth#psychology#recommendations

Takeaway· 1

Takeaway26:30

Information Gain vs. Typicality: What Makes Content Win

Ethan describes how Google (and eventually LLMs) could stop rewarding the endless loop of non-experts rewriting each other's articles. His heuristic: reward information gain (did you say something no one else said?) and penalize typicality (are you so average you look like a rewrite?), ideally backed by original research and real domain expertise. Lenny notes it doubles as a guide for making genuinely good newsletters and podcasts.

  • Information gain: did you say something others didn't?
  • Typicality: too-average content looks like a rewrite and should rank lower
  • Original research and real domain expertise strengthen content
  • The same heuristic describes what makes any content genuinely good

“did you say something that somebody else didn't say? Two is how typical are you? Are you so typical that I think that you're a…”

Ethan Smith · 26:30
#content-quality#information-gain#heuristic#seo