Sometimes the AI does answer about you, but about the wrong you: another company with the same name, or your story from ten years ago. The brand questions in the battery show this. The fix is to make clear, in text and in structured data, who you are, and to give the AIs sourced cards they can cite.
1Read the battery's brand answers as an identity diagnosis
2Disambiguation page and the full brand name
3Person and Organization with a fixed @id and sameAs
4The catalog of cards with source, status, and date
Goal: diagnose name collision and outdated bio in the AIs and build the catalog of cards that fixes your entity.
01 · From identity error to measured fix
Click each block.
re-measure the brand questions the following month
Click each block in the diagram to see what it does.
Diagnosis through the battery, fix on the site, check in the following month's battery.
02 · Fix it in four steps
Tap each step.
1. Read the brand answers
In the spreadsheet from lesson 3, filter the brand questions. Note: did the AI talk about you or about someone else? Is the description current? Which sources did it use?
The wrong sources show where the confusion comes from.
At INEMA (07/10/2026)"What is INEMA?" got 4 out of 4 answers about INEMA-BA, the environmental agency. "Who is Nei Maldaner?" brought up the old bio of an IT businessman, with no AI. But "What is INEMA.club?" was answered correctly.
2. Full name and disambiguation
If the short name collides, always use the full name (at INEMA: INEMA.club) and publish a page that says who you are and who you should not be confused with.
On the founder page, link the old career path to the current one, in text: text is what the AI reads.
3. Entity in JSON-LD
Person and Organization with a fixed @id on every page and sameAs pointing to the profiles you control (GitHub, YouTube, LinkedIn). Course 2, lesson 1, teaches this in detail.
The structured data has to match the visible text.
4. Catalog of sheets
Write short sheets based on the battery questions: these are questions with proven demand. Each sheet with a real source; without confirmation, mark it PENDENTE.
At INEMA, after the July round, 54 sheets were published at /conhecimento/.
03 · Sheet and entity to copy
The sheet follows the catalog format of the AIV method (MIT). The JSON-LD goes in the head of every page.
catalogo/institucional/quem-somos.md
---
slug: quem-somos
tipo: institucional
titulo: What <Full Brand Name> is
resumo: <Full Brand Name> is <what it is, for whom> since <year>. Not to be confused with <namesake>, which <what the other is>.
status: revisado
fonte: <URL or document that proves it>
confianca: alta
atualizado_em: dd/mm/aaaa
relacionados: [quem-e-fundador, servicos]
---
<Direct answer in 1 paragraph, with data only you have.>
## PENDENTE
- <what still needs confirming: does not go live>
Entity Person #nei and Organization #organization published on every page of the portal on 09/24/2026. Catalog of 54 sheets published at /conhecimento/ after the July measurement.
The next round of the battery is what tells whether the identity changed in the AIs: without re-measuring, it's just hope.
Never invent data: no source, PENDENTE.
A public sheet never goes out with PENDENTE.
An honest gap is better than an invented sheet: if the question asks for something you don't offer, the sheet says so.
Ask the owner early what they really sell and what is free: it changes the text of everything.
04 · Practice: your entity
Use the brand questions from your battery.
E1 · Identity diagnosis
30 min
Ask "what is [your brand]?" and "who is [founder or person in charge]?" in 3 assistants. For each answer, note: did it talk about you or someone else, is it current, which sources did it use.
Done when: a table with 6 answers (2 questions × 3 assistants), each marked "correct / collision / outdated" with the sources cited.
E2 · Five cards with sources
45 min
Write 5 cards in the template format: who we are (with disambiguation, if there is a collision), the founder, and 3 answers to questions from your battery. Each one with a source and a date.
Done when: 5 .md files with every front matter field filled in, and no data without a source outside the PENDENTE section.
05 · Test what you learned
The explanation appears after your answer.
Question 1 of 4
"What is [brand]?" comes back describing a public agency with the same name. What is the first fix?skill
Correct
The short name may never be yours. A consistent full name and a page that says who you are (and who you are not) give the AIs something to cite.
Not this one
The short name may never be yours. A consistent full name and a page that says who you are (and who you are not) give the AIs something to cite.Review: Fix it in four steps
Question 2 of 4
What is sameAs for in JSON-LD?
Correct
It links the same entity across pages and profiles. It has to match what is written on the site.
Not this one
It links the same entity across pages and profiles. It has to match what is written on the site.Review: Sheet and entity to copy
Question 3 of 4
A catalog card has a price the owner has not confirmed yet. What should you do?
Correct
Nothing without a source goes live. Made-up information in the catalog turns into a wrong AI answer about you.
Not this one
Nothing without a source goes live. Made-up information in the catalog turns into a wrong AI answer about you.Review: Sheet and entity to copy
Question 4 of 4
How do you know whether the fix worked?
Correct
Same questions, same assistants, 3 runs: only comparing with the previous round shows change.