Is Syft's AI detector free?

Yes. Running a Syft Detect analysis here costs nothing. A free Syft account opens and keeps the result.

Can Syft detect ChatGPT writing?

Syft looks for patterns associated with machine-generated writing rather than fingerprinting a particular model. Text from ChatGPT and similar assistants often carries those patterns, but a result describes the writing in front of it, not which tool produced it.

Can an AI detector be completely accurate?

Syft reports a signal band and the passages behind it, so you can see exactly what the reading is based on. No detector, this one included, is completely accurate: detection is probabilistic, careful human writing can carry machine-like patterns, and edited AI text can lose them. That is why the evidence sits beside the band.

Does Syft show which parts look AI-written?

Yes. Rather than one unexplained number, Syft marks the passages carrying the signal so you can read them in context and judge the result.

What should I do after detecting AI-like writing?

Read the passages Syft marked. Where the writing is stiff or formulaic, Humanize can rewrite it while preserving your meaning, numbers, quotations and citations. Then run the check again on the new copy.

Syft

Free AI Detector

Check text for AI-generated writing and inspect the passages behind the result.

About AI detection

Syft looks for statistical patterns associated with AI-generated writing. A detector result is not proof of authorship.

Questions

Is the check free?: Yes. Running the analysis is free. A free Syft account is required to open the result and keep it, alongside everything else you inspect.
Do you upload my text?: For this check, yes, once. The text and its result are encrypted on our server and held only until you open them from your account, or until they expire unopened. Then they are deleted. They are never shared, used for training, or sent to analytics.
What happens after I sign in?: The analysis you already ran opens in the Syft workspace as a normal item. Nothing runs again and there is nothing to paste again. From there you can read the AI map, run every other lens, and refine the writing.
How should I read the result?: Syft combines a signal band for the whole text with passage-level evidence, so you can see what drove the detection. Uniform, formulaic human prose can carry some of the same patterns, which is why the passages matter as much as the band.
Can Syft prove something was written by AI?: Syft identifies AI-writing patterns and shows you exactly where they appear. That is strong evidence about how the text was written; it is not, on its own, proof of who wrote it.
How much text should I paste?: About 80 words or more. Below that there is not enough variation to read anything from, and Syft says so rather than inventing an estimate.

SyftPrivacy Terms Security

Limited beta

Built for essay-like academic prose in English.

Limited beta. Syft Detect is built for essay-like and academic prose in English of about 160 words or more. It reports a signal band and the passages behind it, and gives no reading for lists, code, bibliographies, quotations, repeated text or very short texts. A detector result is not proof of authorship.

Built for essay-like prose: The reading is designed for essay-like and academic prose in English of about 160 words or more. Between 80 and 159 words it is limited and is never reported as high.
No reading for some inputs: Short notes, lists, code, bibliographies, quotation blocks and repeated text get no reading. Syft says so instead of scoring them.
A signal, not proof: A detector result is not proof of authorship. It is a statistical estimate of how strongly the writing carries patterns associated with machine-generated text.
It can be wrong in both directions: Human writing is sometimes reported as elevated or high, more often for reviews, forum posts and summaries than for essays. Edited or paraphrased AI text is often missed. Text from some models is missed more often at the high band, including Llama-family models, older GPT-3-era models and, at the margin, GPT-4.
Populations we watch: In Syft's evaluation, ESL essays (English as a second language) were reported as high slightly more often than native-speaker student essays, though still rarely; short scientific abstracts were not tested in enough numbers to state a rate. Both are monitored during the beta, and a high result on such writing deserves extra care.
Marked passages show where the signal concentrates: A marked passage carries stronger AI-writing signals. It is not a claim that the passage, or the person, used AI, and a human paragraph inside a mixed text can be marked.
Results change with the model version: Every result records the detector version that produced it. A later version may read the same text differently.
Never the sole basis for a decision: Do not use this result as the sole basis for academic, disciplinary, employment, or legal decisions.
If a result is contested: A reading is not proof, in either direction. What can be reviewed is the record of how a text came to be: drafts, revision and version history, notes and source material, and the writer’s own account. Wrote this yourself? How to show your process.

How it works

How Syft's AI detector reads your text.

Syft Detect reads the text passage by passage, identifies how strongly each passage carries patterns associated with machine-generated writing, and reports a signal band for the whole text with the passages behind it.

An index, a band, and the evidence

Syft reports a 0–100 signal index with the band behind it, and marks the passages that carry the signal, so you can read the evidence and judge for yourself. The index is not a probability that AI wrote the text.

What happens to your text

Text needed for detection is sent once over an encrypted connection to Syft's own servers, is never used to train a model or shared with anyone, and is deleted once you open the result or it expires unopened.

No account to try it

Paste and check. An account only matters once you want to keep the work, open documents, or revise with Humanize.

Read the evidence

The passages are the part you can check.

Syft marks the sentences carrying the signal and names the features behind each one, so a reading can be read against the writing instead of taken on trust. A detector result is not proof of authorship: careful human writing can read as machine-like, and lightly edited AI text can read as human.

Weigh a reading alongside the record of how a text came to be: drafts, version history, notes and sources. It is evidence for a conversation, never a decision on its own.

How Syft measures AI-like writing Why an essay gets flagged

More ways in

The same check, for the situation you are in.

Checking a Word document, a submission, or your own draft before it goes out: each has its own entry, with the same analysis and the same honesty about what a result means.

Can AI detectors detect Claude?DeepSeek AI detector PDF metadata viewer

The loop

Detect, revise, re-check.

Checking text is the first step. The same workspace lets you act on what you find and confirm the result.

Detect

Run the check and see the overall signal.

Inspect

Read the passages carrying it, in context.

Humanize

Rewrite stiff or formulaic passages, keeping your meaning.

Re-check

Run the new copy through again and compare.

Common questions about AI detection.

Is Syft's AI detector free?: Yes. Running a Syft Detect analysis here costs nothing. A free Syft account opens and keeps the result.
Do you upload my text?: Only what the analysis needs. Text needed for detection is sent once over an encrypted connection to Syft's own servers, is never used to train a model or shared with anyone, and is deleted once you open the result or it expires unopened. See Security and Privacy for how processing and retention work.
Can Syft detect ChatGPT writing?: Syft looks for patterns associated with machine-generated writing rather than fingerprinting a particular model. Text from ChatGPT and similar assistants often carries those patterns, but a result describes the writing in front of it, not which tool produced it.
Can an AI detector be completely accurate?: Syft reports a signal band and the passages behind it, so you can see exactly what the reading is based on. No detector, this one included, is completely accurate: detection is probabilistic, careful human writing can carry machine-like patterns, and edited AI text can lose them. That is why the evidence sits beside the band.
Does Syft show which parts look AI-written?: Yes. Rather than one unexplained number, Syft marks the passages carrying the signal so you can read them in context and judge the result.
What should I do after detecting AI-like writing?: Read the passages Syft marked. Where the writing is stiff or formulaic, Humanize can rewrite it while preserving your meaning, numbers, quotations and citations. Then run the check again on the new copy.

Keep the work, not just the result.

An account saves your checks, opens documents and PDFs, and adds Humanize so you can revise what the check marked.

Create a free account