The job
What the Portuguese AI detector does
You have a piece of Portuguese writing and a question about it: a student submission, a freelancer's draft, or your own text that something else has already flagged. Paste it, run the check, and you get a written analysis of the patterns in it.
Not a score. The analysis instruction rules out claiming certainty about authorship, about plagiarism or about which model produced something, and asks instead for observable signals described with their limitations. So the output is sentences you can take back to the text, rather than a number you can only forward.
The workflow
How to run the Portuguese AI detector in three steps
How much text you paste matters more than anything else you do here.
- Step 1
- Paste the Portuguese text, ideally several hundred words. The limit is 12,000 characters and the meter shows where you are.
- Step 2
- Optionally add an instruction: ask for the findings in Portuguese, or ask it to concentrate on connectors or sentence rhythm.
- Step 3
- Read each finding and verify it in the text. A claim you cannot locate in your own document is a claim to drop.
Before you act
How to use the findings responsibly
The output is a starting point for reading, not a conclusion. Work through it like this.
- Locate it
- For each signal named, find the sentences it refers to. If you cannot, the finding does not stand.
- Count it
- If connector repetition is flagged, count the connectors. Patterns like this are checkable by hand.
- Consider the writer
- Formal, technical or second-language Portuguese looks mechanical to any detector. Ask whether the style explains the signal.
- Never decide on it alone
- One of the Portuguese detectors in this space warns in its own FAQ against using detection alone for career or academic decisions. That is sound advice.
By eye
Portuguese phrases worth checking yourself
A lot of this needs no tool. Guides on spotting AI Portuguese converge on the same connector stack, used far more heavily than a person writing naturally would: além disso, por outro lado, portanto, em resumo.
Then there is the grammatical sentence that carries no information. One Portuguese guide quotes two as typical of the form: este item é ideal para quem busca qualidade e durabilidade, and a inovação tecnológica transforma o futuro de maneira positiva. It also flags confidently wrong detail about Portuguese history, citing garbled accounts of the Treaty of Tordesillas, and cultural references that feel inserted rather than meant.
The honest part
Why the Portuguese AI detector flags real writing
False positives in Portuguese are common enough to plan around. One Portuguese-language guide puts misclassification of natural Portuguese text at roughly 15 percent and identifies the writers it hits: concise, repetitive or formal styles frequently trigger false alerts.
That is most academic writing, most technical documentation, and a great deal of writing by people working in Portuguese as a second language, whose sentences tend to be more regular by construction. If you are the one being accused, the useful move is not a counter-score. It is your drafts, your version history, your notes, and a request that whoever flagged you name the specific sentences they mean.
Variants
Brazilian and European Portuguese in one check
There is no variant setting here. One analysis prompt handles both and is tuned for neither, so do not read the output as variant-aware. Competing detectors differ on this: one advertises optimisation for Brazilian Portuguese, another does not raise the question at all.
For your own reading, mixed variants are worth noticing. Consistent você constructions alongside a European word choice like autocarro is a combination a single writer rarely produces by accident, and you can spot it without running anything.
Allowance
Free checks, text length and the output ceiling
Five checks per rolling 24 hours, counted from your first use, with no signup and no paid tier anywhere in the product. Each check accepts 1 to 12,000 characters, and the server rejects anything outside that before analysing it.
Context for those numbers: a Portuguese detector in the same results allows three free uses a day capped at 1,000 characters, with unlimited use behind a subscription, while simultaneously advising that detection works better on longer texts. Our own ceiling is on the way out instead: the findings are capped at 1,400 completion tokens, so a long document still returns a short analysis.
Straight answer
Can the Portuguese AI detector prove who wrote something?
No, and it is instructed not to imply otherwise. It cannot prove authorship, cannot name the model, cannot check plagiarism and cannot forecast what another service will conclude. There is no affiliation with any detection company and no result here predicts anyone else's.
The detectors that do return confident numbers say much the same in their own words once you read past the headline. One states that no AI detector is 100 percent reliable and calls its results indicative rather than definitive. Treating any of this as evidence rather than as a prompt to read more carefully is the mistake the whole category keeps enabling.