Generative Engine Optimization
A lot of buying now starts with a question to an AI. Ranqia measures what those answers say about your market, finds where each answer comes from, and puts your content where the AI looks.
What's the best digital bank for a first account?
For a first account, three options stand out: Aurora, for its easy approval; Norte, which charges no fees; and Lume, with a no-annual-fee card and unlimited transfers.
Ranqia reading
Lume or Aurora: which one is worth it?
It depends on how you use it. Aurora pays more on your balance; Lume wins on fees and on the app, which users call simpler.
Ranqia reading
Is Lume trustworthy?
Yes. Lume is licensed by the central bank and rated well for support, though some complaints mention slow account transfers.
Ranqia reading
Alternatives to Aurora with a no-fee card
If a no-fee card is the criterion, Norte and Lume are the most cited alternatives. Pagus comes up for people who want cashback.
Ranqia reading
Where can I open an account today with a no-fee card?
You can open one in minutes in the Norte or Lume app. An up-to-date comparison site lists the fees side by side.
Ranqia reading
Example with fictional brands
Where the question happens
Anatomy of an answer
An AI answer looks like opinion. Behind it are searches, pages read and choices, and all of it leaves a trail. You can measure it and you can change it. Follow one question from start to finish.
01 / 08
Before answering, the AI rewrites the question into several searches. Asking for the best first account makes it look up fees, no-fee cards, complaints and interest rates. Those searches decide which pages it reads.
We log the searches the AI runs on its own. Each one is a topic with proven demand.
02 / 08
Each search returns pages. The AI reads many and cites a few. In our study of ChatGPT in Brazil, 0.3% of domains held 34.3% of the influence over the answers.
The gap between what the AI read and what it cited shows which pages are one step away from the answer.
03 / 08
Ask again and the order changes. Sometimes the brand is not there at all. A single answer proves nothing, so we ask the same question over and over and read the whole set: the chance of appearing and the most likely position.
One answer is a sample. The measurement comes from the whole set.
04 / 08
ChatGPT, Gemini, Perplexity and Claude read different sets of pages for the same question. The strongest sources show up in all of them. The rest changes from one assistant to the next.
Doing well in one assistant guarantees nothing in the others.
05 / 08
We read every answer in detail: which brands appear and in what order, whether they come with a link and of what kind, which attributes the AI ties to each one, the tone of each passage and who shows up alongside.
That is where visibility, average position, share of answers and tone come from, per question and per intent.
06 / 08
Every week the signals become direction: searches that keep coming back, attributes your brand is missing, the sources the AI reads about your market. The result is a short list of pages to write and channels to take.
Each suggestion is linked to the questions it can change.
07 / 08
We write for machines to read: clear structure, authorship, dates and schema.org markup. The page goes to the channels the AI already reads about your market. Then we count the assistants' crawler visits and check each one against the addresses the crawler's operator publishes.
A ChatGPT-User visit is a person asking the assistant about that page.
08 / 08
The next measurement shows the effect: the brand shows up in more answers and closer to the top. On the visibility and position grid, it is the path from the bottom corner to the leaders' quadrant. Then the loop starts again with the week's questions.
Measure, understand, publish, measure again.
Intents
Each kind of question shows something different about your brand, so each intent gets its own measurement.
01
"What's the best digital bank for a first account?"
What we readWhether your brand makes the list, in which position and next to whom.
02
"Lume or Aurora: which one is worth it?"
What we readWho wins the head-to-head and on which attributes, one by one.
03
"Is Lume trustworthy?"
What we readThe tone of each passage and the positive and negative attributes the AI repeats.
04
"Alternatives to Aurora with a no-fee card"
What we readWho the AI puts in a competitor's place when someone wants to switch.
05
"Where can I open an account today with a no-fee card?"
What we readWhere the click goes: your site, a comparison site or nowhere.
Readings
A visibility number on its own doesn't tell you what to do on Monday. These readings do.
| What the AI did | What it means | What to do |
|---|---|---|
| Read your page and didn't cite it. | The page is close to making the answer. | Give the AI what it was looking for: structure, data and evidence. |
| Answered without searching the web. | The answer comes from the model's memory. | Build reputation in the sources that feed training. It moves slowly. |
| Named your brand without a link. | Presence without a click. | Take the page the AI uses as the category's reference. |
| Gave the win to a competitor on "interest". | Nothing out there backs that attribute for you. | Publish the number, with a comparison and a source. |
| Leaned on one source for half the answers. | One channel decides the market. | Start there. |
| Answered differently in another country. | Each market has its own answer. | Measure and produce per country and per language. |
Public research
We publish what we measure. Each study has the ranking per category, the sources the AI uses and the method. Studies are in Portuguese.
Palestra técnica de Mateus Borges, cofundador da Ranqia, na IA Conference Alura (setembro de 2026): a arquitetura probabilística por trás das respostas do ChatGPT, a diferença entre geração …
Read the studyRelatório Ranqia sobre mais de 9 mil respostas do ChatGPT a perguntas sobre quem venceria as 19 categorias do Emmy 2026, coletadas antes da premiação. O favorito das respostas venceu em 78,9…
Read the studyRelatório amostral Ranqia sobre 1.197 respostas válidas do ChatGPT à pergunta pelas dez campanhas de marketing mais inovadoras e criativas do Brasil em 2026, coletadas em 24 de agosto. Um nú…
Read the studyEstudo amostral Ranqia sobre ESG no ChatGPT: respostas reais a 27 perguntas em 11 temas no mercado brasileiro, ao longo de julho de 2026, com ranking global e por tema. A Natura lidera com f…
Read the studyFounding team
Operators from Bravium, Orbia and Pravaler and researchers from USP, Unicamp, Insper, FGV and UMKC.
Business and operations
Co-founder and COO at Orbia. Director at Pravaler. Estudar Fellow. Insper.
Business and AI research
Data and AI Director at Bravium. Co-founder at Orbia. Forbes 30 Under 30. Unicamp.
Software engineering
Computer Science at Insper. Estudar Fellow and Behring Scholar. Works across frontend, backend and infrastructure.
Backend, infrastructure and AI
Machine Learning at USP. ML Tech Lead at Bravium. Data scientist at Orbia. Owns backend, optimization, infrastructure and machine learning.
Early access
Ranqia is in selective access. Every request is reviewed and becomes a diagnosis of how your brand shows up in AI answers.
We reply within two business days. You can also write to admin@ranqia.com.
FAQ
It is the work of getting a brand to show up, well placed and well described, in the answers of AI assistants like ChatGPT, Gemini, Perplexity and Claude. SEO optimizes for a list of links. GEO optimizes for the answer the AI writes and for the sources it writes from.
SEO competes for positions on a results page. In GEO there is no results page: the AI runs its own searches, reads a few pages and writes an answer citing even fewer. What matters is the searches the AI runs, the sources it prefers and what it says about each brand, not only where your site ranks.
Continuous monitoring starts with ChatGPT. The same measurement setup takes other assistants, like Gemini, Perplexity and Claude, as each one matters for the client's market.
Because the AI answers differently every time. A single answer can show the brand first or not at all. Repeating the question lets you measure the real chance of appearing and the most likely position, with a margin of error.
Measurement of your questions by intent (discovery, comparison, reputation, alternatives and decision), a weekly reading with the pages to write and the channels to take, production and publishing of structured content in those channels, and tracking of the assistants' crawler visits. Details are on the service page.
Mid-size and large brands selling in Brazil and Latin America, especially where questions are asked in Portuguese or Spanish, where tools built for English have thin coverage.
You send a request through the form, we look at your brand and come back with a diagnosis of how it shows up in AI answers today. If it makes sense for both sides, we move on to continuous tracking.