Substack partnered with Pangram to let readers run a "Scan for AI text" check on any post or note over 100 words, published from July 21, 2026 onward. Writers get a disclosure statement, a pre-publish self-scan, and a per-post opt-out — it's a transparency layer, not an AI ban.
2026-07-21 · by Moe Ameen
Substack rolled out an AI-detection feature on July 21, 2026, built in partnership with the AI-text detector Pangram. It adds a "Scan for AI text" option that lets a reader check a post or note and see an assessment of how much of the writing looks AI-generated. The scan works on text longer than 100 words and applies to content published on or after the launch date — older posts aren't retroactively scannable. At launch it's available on web and iOS, with Android support said to be coming later. Treat the exact mechanics as a snapshot from the launch announcement and Substack's own help docs, which will evolve.
Substack framed the move as being about transparency rather than prohibition — the guiding line the company used is that people should know what they're getting. It is explicitly not a ban on using AI, and Substack acknowledged up front that there are limits to what Pangram can detect, so the score is an estimate, not a verdict. Co-founder Chris Best had earlier laid out the reasoning in a post arguing against passing off machine-written text as your own personal voice without saying so.
Writers get a set of controls around it. You can scan your own draft before publishing to see what a reader would see. You can add a statement — a "how I make this" style note about whether and how you use AI — that surfaces when someone scans your work. On the publish screen you can run the scan, open the report, and disable detection for that specific post or note, after which readers won't see a Pangram analysis on it. And if you believe a scan of your work is wrong, you can report and remove the mistaken result. The net effect is a labeling and disclosure system with the writer holding most of the switches, not an enforcement crackdown.
The honest way to read this: Substack's detector rewards a real, disclosed voice and penalizes anonymous, flattened AI text — so the play is not to dodge the scanner, it's to make your AI-assisted work genuinely yours and say so. That's the posture Kompozy is built around. Kompozy is a content generation and publishing engine; when it drafts a Blog Article, an Email Newsletter, or Text Posts, the Persona Brief governs the voice and a banned-word filter strips the generic AI-tell phrasing that both readers and detectors react to. Use it as a first-draft and repurposing engine — turn a talk, a podcast, or a long post into a newsletter and a week of social — then edit in your own voice and pair it with a clear "how I make this" statement on Substack. Disclosure plus a distinctive voice is the durable move here, and Kompozy is aimed at producing on-brand drafts you finish, not text you pretend a human wrote.
The bigger point is surface area. Pangram only reads text, and it only reads text on Substack. Kompozy generates the things a text detector can't touch — Persona Shorts and HeyGen avatar video, Clipped Shorts, Carousel Posts, Quote Graphics, Photo Posts — and its Autopilot schedules and publishes across nine social platforms plus your blog and a Mailchimp newsletter from one queue. So instead of a business that lives or dies on one newsletter's authenticity score, you run a multi-format, multi-platform presence where your Substack is one disclosed, well-voiced channel among many. That's the resilient shape this news is pushing every creator toward, and it's the workflow Kompozy is designed to run.
Substack added a "Scan for AI text" feature, powered by the detector Pangram, that lets a reader check a post or note and see an assessment of how much of the writing looks AI-generated. It works on text longer than 100 words, applies to content published on or after July 21, 2026, and is available on web and iOS at launch with Android to come. Substack describes the score as an estimate and notes there are limits to what Pangram can detect.
No. The feature is a transparency and disclosure layer, not a ban — AI-assisted writing is still allowed. Substack's framing is that readers should know what they're getting. Writers can add a statement explaining whether and how they use AI, and they can disable detection on a specific post if they choose.
Yes. On the publish screen you can run the scan, open the report, and disable detection for that individual post or note, after which readers won't see a Pangram analysis on it. You can also scan a draft before publishing to preview what readers would see, and report and remove a scan of your own work that you believe is mistaken.
Disclose it and keep a real voice. Use AI as a drafting and repurposing aid, then edit the copy into your own voice and add a clear "how I make this" statement so readers know your process. An engine like Kompozy helps by governing the draft with a Persona Brief and a banned-word filter to reduce generic AI-tell, and by generating on-brand video, images, and social posts you publish across many platforms rather than relying on one newsletter channel.