Video summary

The ultimate guide to AEO: How to get ChatGPT to recommend your product | Ethan Smith (Graphite)

Main summary

Key takeaways

Business

What AEO / GEO means (and why it matters now)

  • The guest treats AEO and GEO as effectively the same concept: optimizing so your product appears in answers returned by “search response systems” such as ChatGPT, Gemini, Claude, Perplexity, and similar tools.
  • Key shift vs traditional SEO:
    • In LLM answers, the system often uses LLM + RAG (retrieval-augmented generation).
    • That means visibility depends heavily on what gets retrieved and cited.
  • Winning is less about “ranking a single URL first” and more about being remembered and cited across sources.

Positioning: AEO is “SEO-like” but has important differences

What remains similar to SEO

  • Everything that works in SEO also works in AEO, including:
    • content/coverage
    • topical relevance
    • citations/mentions

What’s different: head vs tail behavior

  • Head difference (top-most query phrasing)

    • Google-style answers: earliest/most prominent URL placement can matter.
    • LLM-style answers: responses may aggregate multiple quoted snippets, so being cited repeatedly can matter more than a single top placement.

  • Tail difference (long tail / specific questions)

    • LLMs handle many more specific follow-up questions.
    • The guest cites ~25 words for LLM-tail vs ~6 words for Google snippets (approximate transcript note).
    • Targeting question-like, ultra-specific queries can be more effective in AEO—especially for early-stage companies that lack domain authority for Google rankings.

Core “operating playbook” (step-by-step)

1) Find the questions you need to rank for

  • Competitor discovery
    • Review your and competitors’ paid search terms.
    • Convert those terms into AEO targets (as questions).
  • Convert keywords → questions
    • Use ChatGPT to “convert this into questions” as a practical way to generate AEO targets.

2) Track visibility in response systems

  • Use an AEO tracker to monitor:
    • how often you appear (share of responses)
    • average rating/score given by the system
  • Account for variability:
    • the same question can produce different answers across runs
    • track question variance and repeat queries
    • track across different “surfaces” (e.g., ChatGPT vs other products, different UI modes)

3) Build on-site pages to answer the question set (topic pages)

  • On-site work resembles traditional SEO, but organized for AEO:
    • create landing pages targeting topics
    • each page should answer hundreds/thousands of related questions
  • Principle:
    • The more sub-questions you answer, the better your odds of being included in responses.

4) Execute off-site “citation strategy” by source type

The guest categorizes off-site citations, such as:

  • Video citations: YouTube, Vimeo
  • Community / UGC citations: Reddit
  • Media / affiliate-like citations: major publications and third-party sites (example mentioned: Dotdash/Meredith)

Importance: in LLM answers, you can win when citations appear in the retrieved evidence set.

5) Run controlled experiments (don’t trust generic “best practices”)

Example experimental design:

  1. Choose N questions
  2. Split into test group vs control group
  3. Control: “do nothing”
  4. Test: apply one change (e.g., post on Reddit, publish videos, pursue partner citations)
  5. Track over time and compare before/after vs control

Reproducibility:

  • repeat “a few times” so results aren’t just noise from answer variability.

6) Build a team / role coverage plan

  • Typically, you’ll need:
    • an SEO-focused capability (often already covered)
    • additional ownership for video + Reddit tactics (often a marketer, growth operator, or specialist) to scale execution.

Concrete examples / case-study style details

Webflow case: AEO outcomes

  • The guest credits AEO work on Webflow with:
    • “6x the difference in conversion coefficient” between LLM traffic and Google search traffic.
  • Tactics mentioned:
    • traditional SEO-style landing pages for high-volume queries (e.g., “best website designer” and variants)
    • off-site presence:
      • YouTube and other video mentions
      • Reddit
      • other blog mentions
      • partner programs / affiliated citations

Reddit example: anti-spam credibility approach

  • Warning: avoid fake Reddit spam accounts (described as attempted by others and often blocked).
  • Recommended approach:
    • post authentically as a real user
    • use a real account profile (name, where you work, helpful answer tied to the discussion)
  • Scaling note:
    • doesn’t require “10,000 comments”—the guest suggests ~5 thoughtful comments can be enough if they match the right threads.

Key metrics & KPIs mentioned (and targets/timelines)

  • Traffic-to-conversion improvement
    • Webflow: ~6x better conversion coefficient from LLM traffic vs Google search traffic (guest asserts “6x”).
  • B2B attribution nuance
    • Measuring LLM/AEO can be harder due to long sales cycles and many touchpoints.
    • Suggested proxy: monitor referral traffic and confirm how users found you during conversion (don’t rely only on “last touch”).
  • Channel/impact timing
    • growth acceleration around January (clickability increased, more user engagement).
  • Customer “share of visibility” metrics
    • track percent of time you appear
    • track average answer quality/rating via AEO tracking tools
  • Tracking reliability
    • repeat questions because answer distributions change; rely on averages, not single runs.

Frameworks and mental models explicitly used

  • LLM + RAG distinction
    • Much AEO optimization targets the retrieval/citation layer (what appears in the evidence), not the base model.
  • Head vs tail targeting
    • Google vs LLM differ at the head; the tail is larger in AEO because more specific questions are asked/answered.
  • Topic clustering
    • Instead of one page per keyword (older SEO), create pages oriented to thousands of related questions.
  • Test vs control experiments
    • treat AEO like an experimentation problem: keep a control group unchanged.

Actionable recommendations (condensed)

  • Convert competitor paid search keywords into AEO questions.
  • Track repeated runs for:
    • share of answers
    • answer rating / quality
    • while monitoring question variance
  • Build topic pages that cover the likely follow-up question set.
  • Execute off-site citations by category:
    • YouTube/Vimeo for evidence
    • Reddit via authentic, high-signal participation
    • partner/media citations when relevant for a B2B niche
  • Use control groups and iterate—don’t assume online “best practices” are correct.

Notes on markets / investing content

  • Minimal investing emphasis.
  • Mostly focused on execution, positioning, and measurement for AEO.
  • Mentions include growth curves and tool-industry context, but no deep market forecasts.

Presenters / sources

  • Presenter/Guest: Ethan Smith (General Director, Graphite)
  • Host/Interviewer: Lenny (podcast host; references “my podcast,” “lenny newsletter,” “lennispodcast.com”)

Companies / examples cited

  • Graphite, Webflow
  • ChatGPT, Google, Claude, Gemini, Perplexity
  • Dotdash Meredith, Forbes (example), Zapier, Otter, BigQuery, Looker
  • Zendesk / Intercom (example)
  • Reforge (mentioned webinar), YC (example)
  • Open-source / Orcus (sponsor)
  • Vanta (sponsor)

Tools / resources mentioned

  • Surfer SEO (for AI-detector benchmarking)
  • AEO tracking tools / dashboards (unspecified in transcript)
  • AI detectors (Surfer SEO referenced)

Original video