Notes, learnings, and honest post-mortems from building AfterLaunch, including using it on ourselves.
Cited answers that include YouTube, by engine59.0%4.1%59.0%8.7%4.8%Cited answers that include YouTube, by engine59.0%4.1%59.0%8.7%4.8%Where AI answer engines find their sources: observed citation patterns, measured separately by engine, sample and counting method.
Read →
Answers citing Reddit44.2%Perplexity0.0%ChatGPT via OpenAI APIAnswers citing Reddit44.2%Perplexity0.0%ChatGPT via OpenAI APIReddit citation rates varied sharply across our matched question set. The study also shows why the OpenAI API and ChatGPT Search need separate labels.
Read →
All 170 domains checked, serving llms.txt45.9%All 170 domains checked, serving llms.txt45.9%We checked 170 domains for llms.txt on 26 August 2026. Here are the adoption counts, file-quality checks and limits of what a one-day census can establish.
Read →
Pew panel: searches ending in a click8%With an AI summary15%Without onePew panel: searches ending in a click8%With an AI summary15%Without oneOrganic search still carries most of the traffic. A second discovery channel is growing fast alongside it, and it behaves differently.
Read →
AI search answers naming them6.4%Young YC companies36.7%Comparison groupAI search answers naming them6.4%Young YC companies36.7%Comparison groupYoung YC companies were named in 6.4% of measured AI search answers, versus 36.7% for the comparison group. What the study measured, and its limits.
Read →Median mention rate0.8%Recent Y Combinator batches83%Household namesMedian mention rate0.8%Recent Y Combinator batches83%Household namesA ten-company pilot found five companies absent from their recorded AI answers. How the sample compares with a well-known group, and what it cannot explain.
Read →Answers naming Atlassian, by engine7%33%33%7%27%Answers naming Atlassian, by engine7%33%33%7%27%Nearly 600 AI answers about ten companies you know. The biggest name was named least, and the engines cited each company's own site 3% of the time.
Read →
Scope and purposeReadActWriteScope and purposeReadActWriteConnect a coding agent to AfterLaunch research and growth drafts through MCP or REST, with account context, scoped permissions and review before completion.
Read →The flow, end to endSubmitOne emailPollTrialThe flow, end to endSubmitOne emailPollTrialThe agent submits a site and an email. One approval email, and nothing runs until the founder clicks it. The whole flow, and why it is approval-first.
Read →How the work movesDiagnoseDecideDraftProveHow the work movesDiagnoseDecideDraftProveHow AfterLaunch measures AI visibility, investigates the sources behind answers and prepares growth work for a founder to review.
Read →How to close the gap, in order1Frame2PresenceHacker News3AnswersHow to close the gap, in order1Frame2PresenceHacker News3AnswersAsk ChatGPT for the best tool in your category and it names competitors, not you. Here is why AI search favours legible products, and how to close the gap.
Read →What works nowChannelsAuthoritySignalsWhat works nowChannelsAuthoritySignalsAI answers add another place for customers to discover a product. A practical guide to measuring that visibility alongside search traffic and conversions.
Read →The audit frameworkSurfaceAuthorityVoiceThe audit frameworkSurfaceAuthorityVoiceA practical audit of how AI assistants describe your product, when they recommend it, and which sources support their answers.
Read →Six checks, six pitfallsChecksPitfallsSix checks, six pitfallsChecksPitfallsPractical checks for AI search visibility: accessible pages, clear product information, useful sources and repeatable measurement.
Read →Two distinct layers, stacked2LinksEntitiesCapterraTwo distinct layers, stacked2LinksEntitiesCapterraHow to assess a new domain’s visibility, maintain useful product profiles and earn references without making an authority score the goal.
Read →How people actually vet softwareHacker NewsHow people actually vet softwareHacker NewsHow to get found where SaaS users actually vet software: Reddit, Hacker News, G2 and Product Hunt. What compounds, and what gets you ignored.
Read →They ask, search, read and checkAn AI assistantSearchCommunities and peersDirectories and review sitesThey ask, search, read and checkAn AI assistantSearchCommunities and peersDirectories and review sitesAn honest playbook for the first 100 users with no audience: where people actually look, how discoverability compounds, and why consistency beats launches.
Read →Overlapping work, shared foundationSearch engine optimisationAnswer engine optimisationGenerative engine optimisationOverlapping work, shared foundationSearch engine optimisationAnswer engine optimisationGenerative engine optimisationAEO vs GEO vs SEO explained without the jargon: what each acronym optimises for, where they overlap, and a clear focus order for a pre-traction SaaS.
Read →Why the four answers fragment4Why the four answers fragment4How to show up in ChatGPT, Claude, Gemini and Perplexity: a per-engine guide to how each one sources answers, why they disagree, and what to do.
Read →The first two steps2Liftable claimsClear entityThe first two steps2Liftable claimsClear entityA plain-English guide to generative engine optimisation (GEO): what it is, how it differs from SEO, and a prioritised playbook a solo founder can run.
Read →