What ChatGPT, Gemini, Perplexity and Claude say when a founder asks about a tool like Brew, and the fix; nobody has checked yet, because the product has no chosen name, so this page sets up the same weekly check the product runs for brands, to start the week the name is picked.
Today
What we know about ourselves
Fact
Source
No one has asked any AI tool about us; there is no record of a check
The product's name is open, so there is nothing to search for yet; Brew is the internal name
memory brand-directions-2026-10-25
Our landing has the title Brew and no description, so an AI tool has little to read
brew/public/index.html:2
We serve an llms.txt for each brand's own Brand Book, never one for ourselves
brew/server.js:937
The method we already have, built for brands
Piece
What it does
Source
The weekly check
Asks a fixed set of shopper questions every week to ChatGPT, Gemini and Perplexity through treg, and to Claude with web search; reads whether the brand is named, which rivals are and which sites are cited
brew/live-watch.js:4-6
Its cost
About $0.50 a run; a failed run waits six hours
brew/live-watch.js:22-24
The questions
Earned by evidence: Google suggest, search volume, threads from listening and best sellers; the model only picks and phrases
brew/ai-questions.js:1-12
The standing
Who the answers name and for which questions, counted where the shopper named no brand; a source that failed says so, never zeros
brew/ai-standing.js:1-6brew/SCOPE.md:16
One honesty rule
Under Gemini the Claude ask is skipped, so a Gemini answer is never labelled Claude
memory brew-gemini-provider
Tomorrow
Item
Version
Tag
Our own question set, earned the same way: what founders search before they pick a marketing tool for a small store
v1
New
Branded questions once the name exists: what is it, is it safe, what does it cost, does it post for me
v1
New
A weekly record here: named or not, what each tool says we do, and which pages it cites
v1
New
A fix list: each wrong or missing line traced to a page we own, and that page rewritten from claims we can make
v1
New
An llms.txt and a plain description for our own site
v1
New
Check that answers say it posts only after your yes, and never before v2 ships
v2
New
Check that answers do not credit us with results before v3
v3
New
What a good answer says, the target
It reads your store and what is said about it, then drafts this week's marketing for you to pick and hand off brew/SCOPE.md:11-20
Every number shows its source memory brew-proof-plain-make
It does not post or measure in v1 handbook content/marketing-and-growth/getting-found-ourselves.md
How we do it
Wait for the name and the domain.
Build our question set with the same method as a brand's: suggest, volume, threads. No question is made up.
Ask Param before the first paid run; each run costs about $0.50 memory no-spend-while-building.
Run it weekly and log one row a week here: date, questions asked, times named of the total, the wrong lines, the cited pages.
For each wrong line, find the page it most likely came from, fix that page, and note the fix beside the row.
Never write a page to game an answer with claims we cannot back.
Share of unbranded questions where any answer names us
Our weekly check
Proposed
Right description
Share of branded answers that describe v1 correctly: reads, drafts, hands off, never posts
Our weekly check, read by a person
Proposed
Our pages cited
Share of answers that cite a page we own
Our weekly check
Proposed
Open
Do we run a check before the name is picked? Proposal: no; run only the unbranded questions once, free in a chat by hand, to learn which tools get named.
Which market's answers do we check? The brand check asks from India. Proposal: India first, since every partner is Indian, and add the US when we open past India.
Who reads the answers each week? Proposal: Sai, ten minutes on Monday, and Param approves any page fix.
