Substack's New AI Detector Raises Stakes for Indian Writers
Substack is rolling out a Pangram-powered tool that scores how much of a post may be AI-written. For India's freelance writing economy, the catch is false positives.
The News
Substack is adding a built-in way to judge whether the words on a page were written by a person or a machine. In a blog post published on Tuesday, the newsletter platform said a new detection feature can scan posts, notes, replies and comments and return an estimate of how much of the text may have been produced with the help of AI.
The capability is not Substack's own. It is powered by Pangram, a specialist detection firm whose model does the scoring behind the scenes. The feature is arriving first on the web and on the iOS app, with an Android release promised for later.
Crucially, the checking is not reserved for editors. Readers themselves will be able to run a piece of writing through the analyser and see a percentage estimate, turning every subscriber into a potential auditor of the person they are reading.
Why It Matters
For a platform that sold itself on the idea of the direct, human voice, bolting on a machine to police machine writing is a notable shift. Substack spent years positioning newsletters as a refuge from algorithmic feeds and content mills. Inviting an automated referee into that relationship signals how quickly generative tools have unsettled the economics of trust in publishing.
The move also lands in contested territory. Detection remains a famously unreliable science. OpenAI quietly retired its own AI-text classifier in 2023 after conceding it was too inaccurate to be useful, and academic reviews have repeatedly warned that these systems produce both false negatives and, more damagingly, false positives. The last time a detection wave swept through writing, in the months after ChatGPT arrived, universities rushed to adopt tools later shown to wrongly accuse real students. Substack is wading into the same current, this time with money and livelihoods attached rather than grades.
Indian Angle
That question of wrong readings is where the story turns sharply towards India. The country supplies a vast share of the world's freelance writers, ghostwriters, copy editors and newsletter operators, many working in English as a second or third language for overseas clients and platforms. Substack is one of those platforms.
The problem is well documented. A 2023 study by Stanford researchers found that popular AI detectors flagged more than half of TOEFL essays written by non-native English speakers as machine-generated, while rarely misjudging native writers. The systems latch on to simpler vocabulary and predictable sentence structure, the very traits common in competent second-language prose. For an Indian writer building a paying audience, a stray high score visible to readers is not an abstract risk. It is a potential hit to reputation and income delivered by a tool the writer never chose to install.
There is a policy layer too. India's Ministry of Electronics and Information Technology has been pushing platforms towards labelling synthetically generated content, following its deepfake advisories and proposed amendments to the IT Rules. Substack's approach runs the other way, inferring origin after the fact rather than tagging it at source. For India's creators, the safer bet is documented process, drafts and edit histories, not a probability score they cannot appeal.
FAQ
When does the tool go live?
Substack said the feature is rolling out now on the web and its iOS app, following the Tuesday announcement. An Android version has been promised but without a firm date, so users on Google's platform will have to wait before they can run checks themselves.
Who provides the detection technology?
The scoring is handled by Pangram, an outside AI-detection company, rather than a model Substack built in house. That means the accuracy, and any bias, of the results depends on Pangram's system rather than on Substack's own editorial judgement.
Why should Indian freelance writers care?
Because AI detectors have a track record of misclassifying non-native English writing as machine-made. Stanford researchers showed the effect in 2023. For Indians writing for global platforms, a false positive shown to readers could unfairly dent credibility and earnings.
Where can I read the original announcement?
The rollout was reported by The Verge, based on Substack's own blog post. The link to the full coverage is in the attribution paragraph below.
This story was reported by The Verge. Read the full original coverage at The Verge.