Name: Rephrasy AI Detector
Brand: Rephrasy
Rating: 4.9 (203 reviews)

Question 1

Can Rephrasy bypass Substack's AI detector?

Accepted Answer

Substack's detector is Pangram. Rephrasy rewrites AI text to change the structural and stylistic patterns Pangram's classifier scores on. Because Pangram is a deep learning model rather than a perplexity checker, no tool can promise a 100% bypass — but Rephrasy goes deeper than surface paraphrasing and meaningfully reduces the AI signal in your draft.

Question 2

How does Substack detect AI-written content?

Accepted Answer

Substack does not run its own detector. It partnered with Pangram in July 2026 and passes eligible text to Pangram's neural classifier, which returns an estimate of how much was written by hand versus with AI assistance. Pangram was trained on roughly a million human documents plus output from GPT-5 and other frontier models, and uses hard negative mining and active learning rather than perplexity and burstiness — so it is explicitly built to catch text that has been through a paraphrasing tool.

Question 3

Can I turn off AI detection on my Substack?

Accepted Answer

Yes, but only one post at a time. On the Publish page of a draft — or from the three-dot menu on the analysis box — you can choose to disable AI detection for that piece, and readers who try to scan it see "AI detection unavailable". Substack does not currently offer a publication-wide toggle, which has been the most common complaint from writers since launch. Bear in mind that an unavailable scan is itself visible to readers.

Question 4

What content on Substack can be scanned?

Accepted Answer

Posts, Notes, replies and comments longer than about 100 words, published on or after July 21, 2026. Any reader can trigger a scan from the three-dot menu in the Substack Reader on web or iOS. Content published before launch, very short text, video, audio, emails and posts read on custom-domain sites are all outside the scan.

Question 5

Can I scan my own post before publishing it?

Accepted Answer

Yes. Substack lets writers run the same Pangram check on a draft, so you can see how your piece reads to the detector before subscribers do. That makes the workflow simple: draft, humanize with Rephrasy, scan the draft, publish once you're happy with what comes back.

Question 6

Is the AI score shown publicly on my profile?

Accepted Answer

No. The result is shown only to the reader who ran the scan — there is no public AI badge attached to your post, your publication or your profile. The risk is reputational and one reader at a time, not a permanent label.

Question 7

Does Substack ban or penalise AI-assisted posts?

Accepted Answer

Substack has said the feature is about transparency, not prohibition — it is not designed to penalise AI-assisted writing, and there is no automatic ban. Chris Best framed the problem as "Claudefishing": the mismatch between what a reader expects and what they get, rather than AI use in itself. The practical risk is losing subscriber trust, not losing your account.

Question 8

Does the scan work on short Notes?

Accepted Answer

No. Text under roughly 100 words returns "not enough text" rather than a score, and detector accuracy drops sharply on short passages in general — a documented limitation acknowledged in Pangram's own research. Longer posts give the model far more signal to work with.

Question 9

Does Substack's AI detection work in other languages?

Accepted Answer

Pangram advertises support for more than 20 languages, so a non-English newsletter is not automatically outside the scan. Detector accuracy varies by language, and non-native English writing in particular is documented to draw false positives.

Question 10

What if the scan flags my genuine writing?

Accepted Answer

You can report a scan on your published work that you believe is mistaken and have it removed. False positives are a known problem with every AI detector — Pangram claims a rate of roughly 1 in 10,000, but independent reporting, including by The Atlantic, has questioned how reliable these tools are in practice.

Question 11

Should I use Substack's "How I make this" statement?

Accepted Answer

Substack gives writers a dedicated space in Settings › Details to describe their process, and readers see it when they scan your work. If AI is part of how you write, stating that openly is usually the stronger play with paid subscribers — Rephrasy is there to make sure the writing itself reads like you, not to hide your workflow.

Question 12

Is Rephrasy free to use?

Accepted Answer

You can try Rephrasy with a free trial — no credit card required. Paid plans unlock higher word limits and access to the advanced humanization models.

Question 13

Is it ethical to humanize AI text before publishing?

Accepted Answer

Rephrasy improves the quality and readability of AI-assisted drafts so they read in your own voice. How you use the output is your responsibility — be straight with your readers and follow Substack's publishing policies.

Bypass Substack'sAI Detector.

Bypass Substack's
AI Detector.