Krea 3 in Brazil: what Grok took from the PAA
Our own PAA data: the PT-BR seed 'krea 3 inteligencia artificial' returned 30 questions, 30 about Grok, 0 about Krea. What that says about mindshare.
On this page
Krea 3 in Brazil: what Grok took from the PAA
This article uses our own data. On 2026-10-15 we scraped the Google "People Also Ask" box for the Portuguese seed "krea 3 inteligência artificial" with the regional settings hl=pt and gl=br PAA dataset. The panel returned 30 questions. All 30 were about Grok. None were about Krea. That is the finding, and this article is the analysis around it: what the questions reveal about Brazilian demand for AI image tools, and what the gap means for Portuguese content.
TL;DR
- 30 PAA questions were collected for the PT-BR seed "krea 3 inteligência artificial" on 2026-10-15 PAA dataset.
- 30 of 30 (100%) were about Grok: free access, limits, censorship, video and image creation, origins PAA dataset.
- 0 of 30 mentioned Krea, Krea 3, or Krea 2.
- The largest clusters are free access (7 questions) and usage limits (5 questions), together 40% of the panel PAA dataset.
- Inference, not a measurement: Krea 3 has no visible mindshare in Brazilian search, and the local "free AI" demand is being served by Grok content.
How the data was collected
The collection used the people-also-ask scraper, a Playwright/Chromium pipeline that opens Google with regional settings, records the PAA panel questions at depth 0 to 2, and stores them as JSON with the answer text people-also-ask repo. For this article the relevant file is the raw JSON for the seed "krea 3 inteligência artificial": 30 questions, collected 2026-10-15 PAA dataset. The collection notes for the whole run are in the ideas file, which records this seed's Grok drift as an observation at collection time collection notes.
The seed was a real Krea query: "krea 3 inteligência artificial". Google's PAA panel for that query, in the Brazilian region, answered with Grok. The Krea 3 model exists on the Krea platform, listed as "Krea 3 Beta" with generation and editing from image references Krea. The model being real and the PAA panel ignoring it are both true at the same time.
What the 30 questions ask
The full question list, as recorded PAA dataset:
| Category | Count | Examples (as recorded) |
|---|---|---|
| What it is / capabilities / origins | 8 | "O que é o Grok 3?", "O que a IA Grok faz?", "O Grok AI é de Elon Musk?", "O que significa o nome Grok?" |
| Free access | 7 | "O Grok é gratuito?", "Como usar a IA Grok de graça?", "Como conseguir Grok gratuito?" |
| Comparison and value | 6 | "Qual IA é melhor que o Grok?", "Vale a pena usar o Grok?", "O Grok é a melhor IA?" |
| Usage limits | 5 | "Qual o limite de uso do Grok?", "Quando acaba o limite do Grok?", "Quantos vídeos posso criar por dia no Grok?" |
| Censorship | 2 | "Como tirar a censura do Grok 3?", "Como desativar a censura do Grok?" |
| Creating images and video | 2 | "Como criar imagens no Grok?", "Como criar vídeos com o Grok?" |
| App and troubleshooting | 2 | "Como utilizar o aplicativo Grok?", "Por que o Grok não está funcionando?" |
| Krea questions | 0 | (none) |
The distribution matters more than the names. Four of every ten questions are about cost and limits (free access plus usage limits, 12 of 30). That is transactional intent: the Brazilian searcher in this panel is not asking "what is this AI", they are asking "can I use it without paying, and how much can I use". Censorship gets two questions in one panel for a video-and-image model, which matches the H3 censorship cluster we tracked in the English seeds collection notes.
What this does and does not show
What it shows:
- For the query "krea 3 inteligência artificial" in gl=br, Google's PAA panel associates the topic with Grok, not Krea PAA dataset.
- The local demand around AI image and video generation is framed in free-access and limits language, in Portuguese PAA dataset.
What it does not show, and we will not pretend otherwise:
- It is one query, one region, one day. It is a snapshot, not a traffic ranking and not a market-share measurement.
- The PAA panel reflects Google's association between the query and related questions. A Grok-dominated panel is evidence of association strength, not of installed base.
- We did not scrape the same seed in hl=en for the same day, so we cannot say the panel is Grok-specific to Brazil from this file alone. The English seeds in the same run did return Krea-relevant questions, and that contrast is the one that matters collection notes.
What the gap means
The inference, clearly labeled as inference: Krea 3 has no visible mindshare in Brazilian search for image generation, and the generic "IA grátis" demand is being answered by Grok content in Portuguese.
Three consequences for content strategy:
- Portuguese coverage of Krea 3 is a gap, not a nice-to-have. The model is documented on the platform Krea, and the local panel has zero Krea questions. That is the space a PT-BR tutorial set claims first.
- The Grok questions are a demand map. "Como usar a IA Grok de graça?" and "Qual o limite de uso do Grok?" are the exact transactional questions a Portuguese Krea free-plan article can answer better, because Krea's free tier is documented in numbers: 100 units per day, no credit card Krea pricing.
- The censorship pair ("como tirar a censura", "como desativar a censura") points to the same curiosity cluster as the H3 filter article, in Portuguese. One data point per panel is weak, but it is the same cluster in two languages, which is the pattern collection notes.
Common errors and fixes
| Error | Cause | Fix | Source |
|---|---|---|---|
| "Krea 3 is unknown in Brazil" stated as market share | One PAA snapshot is not a survey | State it as: the PAA panel for the seed returned 0 Krea questions on 2026-10-15 | PAA dataset |
| "Grok dominates Brazilian AI search" | Same single-query limit | Cite the seed, region, date, and question count with every claim | PAA dataset |
| Answering Grok questions with Krea content without saying so | Panel questions are about Grok | Answer the Grok question, then bridge to Krea where the documented facts differ | PAA dataset |
| Reusing the data after models or panels change | PAA panels move with the index | Re-scrape the seed before quoting counts; the file is dated 2026-10-15 | people-also-ask repo |
| Claiming the EN panel behaved the same | EN seeds were collected separately | The EN seeds returned Krea-relevant questions; do not merge the two without saying so | collection notes |
Frequently Asked Questions (FAQ)
Is Grok free? The recorded PAA answer for "O Grok é gratuito?" says Grok has a free version with usage limits, available on grok.com and in the app PAA dataset. That is the PAA box's summary, recorded as collected, not a verification by us.
How do I use Grok AI for free? The panel contains three variants of this question ("Como usar a IA Grok de graça?", "Como usar o Grok sem limites de graça?", "Como conseguir Grok gratuito?") PAA dataset. The recorded answers describe the free tier with limits; the specific numbers are not in our collected fields.
What are Grok's usage limits? "Qual o limite de uso do Grok?" and "Quando acaba o limite do Grok?" are both in the panel, which is how Google's own box frames the question PAA dataset. The exact limit values are not stored in our collected fields.
Is Grok AI made by Elon Musk? "O Grok AI é de Elon Musk?" is one of the recorded questions PAA dataset. The PAA answer text is stored in the raw JSON; this article reports the question's presence, not a fact check.
How do I create images in Grok? "Como criar imagens no Grok?" appears in the panel, confirming image creation is part of the local demand PAA dataset. The Krea side of this comparison is documented in the Krea free plan guide.
What AI is better than Grok? "Qual IA é melhor que o Grok?" and "Qual aplicativo de IA é melhor que o Grok?" are both recorded PAA dataset. The comparison space is planned in the AI image generator landscape guide.
Sources
- PAA dataset, paa_pt_krea_3_inteligencia_artificial.json: 30 questions for the seed "krea 3 inteligência artificial", hl=pt, gl=br, collected 2026-10-15. Own data.
- Collection notes, blogpost-ideas-minimax-h3-krea3.md: methodology, seed notes, the Grok drift observation. Own data.
- people-also-ask repo: the Playwright/Chromium scraper used for collection.
- Krea: Krea 3 Beta listed on the platform. Checked 2026-10-15.
Learn how Krea 3 edits with a sentence Krea free plan: accounts and limits Is MiniMax H3 censored