Substack has introduced a new AI transparency feature that lets readers estimate whether a post or note was written by a human or created with AI assistance. The tool, called “Scan for AI text”, analyzes supported content and provides an estimate of how much of the text appears to be human-written versus AI-assisted.
The feature is powered by Pangram and is available for posts and notes published on or after July 21, 2026. While it won’t tell you with certainty whether AI was used, it offers a quick way to better understand how a piece of content may have been created.
Here’s how to use it on your iPhone and on the web.
How to Scan a Substack Post for AI Text on iPhone
If you’re using the Substack app on iPhone, scanning a post only takes a few taps.
- Open the Substack app.
- Tap the post or note you want to analyze.
- Tap the three-dot menu in the top-right corner.
- Select Scan for AI text.
- Wait a few seconds while the analysis is generated.
Substack will then display an estimate showing how much of the content appears to be human-written and how much may have been AI-assisted.
How to Scan a Substack Post on the Web
The feature is also available in the Substack Reader on desktop.
- Open a post from your Subscriptions feed.
- Click the three-dot menu in the upper-right corner.
- Select Scan for AI text.
After a brief analysis, you’ll see the same AI estimate that’s available in the iPhone app.
What the AI Analysis Means
The result isn’t a definitive judgment about whether AI was used. Instead, Substack provides an estimate based on Pangram’s analysis. The report indicates how much of the writing appears to be human-authored versus AI-assisted, helping readers better understand how the content may have been produced.
Why You Might Not See the AI Scan Option
If Scan for AI text isn’t available, there are several possible reasons.
The feature only works with posts and notes published on or after July 21, 2026. It also isn’t available for video posts, audio posts, email versions of newsletters, or articles viewed on standalone Substack websites using custom domains.
In some cases, the creator may have disabled AI detection for that particular post or note. When that happens, readers will see an “AI detection unavailable” message instead of an analysis.
Can Creators Turn Off AI Detection?
Substack allows writers to disable AI detection for individual posts and notes before or after publication. When disabled, readers won’t receive a Pangram analysis and will instead see a message stating that AI detection is unavailable.
Creators can also add a “How I make this” transparency statement explaining whether they use AI during their writing process. That statement appears alongside the AI analysis when readers scan supported content.
Substack says neither it nor Pangram uses publisher content to train generative AI models. The platform also offers a “Block AI training” setting that signals external AI crawlers not to use a publication’s content for AI training.
As AI-generated writing becomes more common online, Substack’s new transparency feature gives readers additional context without claiming to offer a definitive answer. While no AI detector is perfect, “Scan for AI text” provides a simple way to estimate how much AI may have been involved in creating a post before deciding how much weight to give its conclusions.
