KailxLabs publishes first party research on how AI apps name brands, and every study ships with its method, its limits and a data file. Two studies are published, both on clinics, from May 2026. Five more are planned: app against API, ChatGPT shopping, law firms, and two on how ChatGPT plugins are discovered. The plugin work has a first pilot with open data.
Both studies are correlation, not cause, and both come with an open data file.
8,385 answers from API versions of ChatGPT and Perplexity, plus Gemini grounding and Google results, across 75 metros. Correlation, not cause. Open data file.
233 clinic websites checked for the basic structure we look for. No AI app was queried in this study. Open data file.
Each planned study shows what we already know, or a preview of what it will look like. No empty boxes.
Public research puts the overlap at 12%. We are measuring it on 50 buyer questions, app against API, source by source.
Public figure: 12.0% of cited domains overlap between the ChatGPT API and the ChatGPT app on the same questions (Gemini: 14.8%). The ChatGPT and Gemini apps share 5.4%.
arXiv 2609.18729, academic audit · 1,536 answers · 16 Sep 2026
Who gets the carousel, which pages feed it, and how often the price shown is wrong.
Preview · fictional products
Homepage focus against review counts across well reviewed firms in five US metros. Publishes with its data file.
Publishes with its data file · no findings shown until then
A new surface: ChatGPT can suggest software during a conversation. We keep what OpenAI documents, what we counted and what nobody knows on one page.
What OpenAI documents, a directory pilot from 9 Oct 2026 with two open data files, and four planned studies.
Open the hubWhich plugins hold the directory shelves, and which appear for category words. 100 fixed queries, daily for 14 days. A nine query pilot is out with its data files; the study itself has no findings yet.
60 prompts across five kinds of request, 10 runs each, on an account with nothing installed. Not started.
Every study states its questions, the AI apps and surfaces it read, the dates, and how answers were counted, in line with our methodology.
What the study cannot show, including where it read an API rather than the app, and that a pattern is correlation, not cause.
Every study ships with its data file, so anyone can check the numbers. No data file, no study.
Yes. Cite the study page and its date. Every study ships with its data file, so you can check the numbers yourself.
It was run in May 2026, before we moved our measurement to the consumer apps, and its summary says so. New studies measure the apps buyers use, and label any API figure.
When its data file is ready to publish with it. We do not publish findings without the data behind them.
No. Our published studies find patterns, which are correlation, not cause. Proof of cause needs a controlled comparison, which is what the worked and untouched halves in client work are for.
20 minutes with Kailesk Khumar. You leave knowing where AI names you today, and what it would take to become the answer.
Pick a time that suits you. The calendar invite arrives straight away. Prefer a link? Book on Cal.com.
