I Was Banned by Substack’s Algorithm: Here’s What Happened and What It Reveals!
Of Getting Caught in The Algorithm, Safety Measures, Trust, Design Missteps, and Protecting Your Brand.
Hello All!
My apologies for the delay in posting these past few weeks—I encountered an unexpected issue with Substack's algorithm that temporarily disrupted my ability to share updates.
Towards the end of May, Substack’s algorithm incorrectly flagged my account as spam, leading to a temporary ban. This meant that no one could access my content, and I was unable to publish new material or interact with you.
As some of you may know, I write two publications—this one, and
—both of which were affected by the ban.This incident took nearly a week to resolve, involving multiple appeals and emails.
Eventually, I received this response from the support team:
“… here from Substack Trust & Safety. Thanks for taking the time to submit this appeal. It appears that one of our platform safety detection mechanisms incorrectly flagged your publication as spam. I've reactivated your account so you should be able to start publishing again immediately. I apologize for the inconvenience.”
Thankfully, we're back on track now.
To show my appreciation, I’m offering a limited-time promotion: enjoy a 50% discount on upgrades until June 24. Don’t miss out!
I am writing today’s post on this issue as it may affect others as well and to illustrate Substack’s design choices, which I first wrote about in January that impact user experiences.
Using this event, I will share why Substack needs to revisit its approach to trust and safety and consequent user experience when offering to be the primary revenue platform for writers.
I will highlight design issues at each step and suggest alternate solutions.
Initial Warning Signs
The first indication that something was wrong with my account was when I could not like another writer’s post or comment on it.
I had just posted an article on May 26: Wisdom through parables! on helloIndia! which was well-received and commented upon the same day, by its readers.
A day later, I returned to check Notes, read articles, and see new feedback on my post, as well as to work on my draft post for this publication.
Given the generic ‘you are blocked’ message I received when I tried to leave a comment, I checked the Dashboard, where I then read the below banner message displayed in green.
Design Issue 1: Confusing Banner Messages and Colors
My publication had been removed, the message said.
The banner image background color used by Substack was the same green color and font size I had set as the theme for my publication.
A user’s expectation may be valid if they were to see such a serious message displayed as a system warning message in either yellow or a blazing red.
When I checked the second publication, it had the same message in blue, which again is the default publication theme color.
In addition, depending on the action, the messages shared with the user were different. For instance, when I tried to update the post to remove the image suspecting it to be the root cause, the message in green indicated an issue with my email list. See below.
Note the green color of the banner message signifying everything is ok when it is clearly not!
Design Issue 2: Lack of Timely User Notifications
Except for these inline messages, visible only after logging in, I received no email notification from Substack about this critical event, despite having opted for email notifications from the platform.
This is especially concerning given that I have paid subscribers to both publications, making me a paying customer of Substack, albeit in a small way.
Timely notifications for critical events are essential to ensure users are aware of issues that may affect their accounts and audience. It also serves as a trigger for them to take corrective action promptly.
Would you prefer to be notified automatically by the platform by email that your publication is banned ?
Design Issue 3: Sub-optimal Appeal Process and Use of Available Data
My first step was to submit an appeal to the Trust and Safety team as suggested by the message.
The link provided in the Dashboard banner led to a generic form without pre-filled user information. The platform was not leveraging available user data including specifics of the action taken against the user to make the form a breeze to enter.
The form's structure and wording also assumed the user was at fault and had to prove their innocence. This is counterproductive. The appeal form is below:
Here are the form’s suggested attestations, though listed as optional, and the required questions for the appeal to be submitted:
Asking a customer to ‘consider’ whether their content violates guidelines in an appeal form is unnecessary when the platform's algorithm has already flagged the content and should know what has triggered the event.
This approach can undermine trust with genuine customers seeking to resolve false positive actions.
In addition, would a truant actor answer differently ? The purpose of this attestation remains unclear, especially if ‘optional.’
Design Issue 4: Lack of Data Integration and Transparency
The form asked me to explain why I believed my content did not violate guidelines without providing any data or context for the flagged content.
My first question was: what was the root of the issue?
Without knowing the specifics of the issue, and lacking information from the platform, it was challenging to explain anything in the form.
For my part, I had not undertaken any action that I had not done a hundred times before—publish a post, use Substack image tool to add images, share the link in Notes, check feedback, engage, rinse, repeat.
I was working in the dark - a less than optimal space for any user.
Design Issue 5: Shifting the Burden to the User - Root Causes
Of course, in order to ‘explain’ anything in the form, without any data offered by the platform, meant that I had to revisit recent actions:
Was it the Unsplash image from the Substack image tool I had used in my last post an issue?
(Pro-tip: Yes, images and videos can contain embedded virus links. So, be careful before opening randomly forwarded ones on your mobile phone.)
Was someone posting as me? On Medium, and now even on Substack, bots are posting comments with Whatsapp phone numbers by creating fake accounts duplicating legitimate user names with slight variations. These are reported as spam by other users but may affect the original user as well though they are oblivious to it.
(Pro-tip - check the profile handle to spot odd names. More on this in a future post.)
Was it the vague random DMs1 I got from users who I presumed were subscribers but turned out to be strangers with random investment messages and hellos that I discovered only after I answered their hello? Were any of the DMs from malicious users?
Did someone hack my account and post explicit images that was seen by readers and followers without my being aware of it ?
My best guess was it was either 1 or 2. But when I contacted Unsplash, they confirmed (within 24 hours) that the image was clean.
This analysis could have been avoided with transparent communication from Substack about the specific issue that triggered the action, taking their own customer into confidence for resolution which would serve the platform’s goal to continue to build a trusted relationship with its audience.
Pro-tip: Design issues can never be fixed in support.
Design Issue 6: Preparing the Appeal Form
The way the appeal form is designed meant:
I had to submit the same information for 2 publications that was tied to my account.
I had to also submit the form to overturn the ban on my profile/account so I could resume Notes, likes, comments and chats.
In effect, I had to submit the form thrice.
A more effective design would have used all available data on the account, including the associated bans imposed, and enable a single form to be filed for the appeal to be reviewed.
This would reduce the burden on the support team by eliminating the need to process three separate transactions.
It would also prevent guesswork on the part of the user regarding what needs to be appealed.
See below for the form submitted for overturning the profile ban:
Design Issue 7: Lack of Acknowledgment of Appeal Submissions
Upon submitting the appeal form, users receive only a brief pop-up message stating “Thank you for the submission.” However, there is no subsequent communication from Substack. No auto-generated acknowledgment number is provided, nor is there an email confirming the submission details or providing an estimated resolution time.
The help center’s unrelated articles vaguely suggest a typical resolution time of 72 hours, but it remains unclear if this timeline applies universally to all cases.
This lack of follow-up leaves users in the dark, relying on guesswork to determine the status of their appeals, despite the platform having the capability to provide clearer communication and support.
Design Issue 8: Resolution Time is Dependent on Appeal Submission
The issue took approximately 72 hours to resolve after I submitted the appeals.
Had a user not logged in for a week, and delayed submitting the appeal by a week, the resolution would have taken over a week, plus an additional 72 hours.
The absence of auto-notifications or acknowledgments for received appeals leaves users uninformed, delaying resolution time, impacting publication schedules and brand reputation.
Moreover, the platform’s response was limited to reactivating my account without providing any explanations regarding the incident's cause, process, or preventative measures for the future.
Despite my attempts to seek clarification from the support team on why the issue was triggered, I did not receive a response even after three weeks.
During the period when my account was banned, I missed out on daily sign-ups of new subscribers. After it was restored, I discovered that all chat threads with reader interactions from the last few months had been lost.
Visitors to the site were also unable to access content, with the platform suggesting that they check out other Substack content instead of providing a way to contact the publication owner or set expectations for service resumption. For instance, another way a customer may arrive at your site is via direct search which would have resulted in a poor experience. See the badge sent by Google for this publication:
What Substack Did Well
The one thing Substack got right was enabling limited access, allowing users to download subscriber data, back up their content, and continue to write drafts.
Pro-tip: Have you exported your data lately? Check out how to download formatted posts.
However, the subscriber dashboard with paid subscription information was either wiped clean or hidden, without any message indicating why it was missing.
Pro-tip: In a future post, I will discuss how to use your downloaded email list to communicate with users outside Substack, without migrating, until the situation is resolved.
Conclusion: Analysis and Best Practices
Substack, like other social media platforms, automates its trust and safety mechanisms. However, Substack’s algorithm should revisit standard social media platform business logic and automated algorithms, given its unique role as a marketplace for writers and readers.
Before blocking a revenue-producing customer, a platform needs to consider its algorithm design, taking into account various use cases and edge cases to avoid false positives, or better handle them when they occur, for impacted customers.
Using data to predict harmful actions and leveraging historical transactional patterns to enhance protections and prevent false positives should be a priority—an industry-standard best practice.
Owning your audience is not the same as being able to own communications with them, despite the platform's founders promoting this capability as a feature.
Here is a note posted on June 3 that I found particularly ironic given this experience:
The reality is that controlling communication with your audience remains within the domain of platforms like Substack, which can arbitrarily decide whether communication is allowed.
This might lead you to reconsider consolidating all your content on a single platform and creating an unwanted ‘single point of failure.’
I will offer some solutions in upcoming posts on how to remain on Substack while also developing alternative communication channels with your readers. Stay tuned!
Did you find this post helpful?
As a token of gratitude for supporting this publication, I am offering a special promotion—get a 50% off forever discount if you upgrade now.
Check out this related post:
My DMs are now limited to paid subscribers only. Alternatively, please write to hellotechmadesimple@proton.me.
Wow I can’t believe how calm you are. If that happened to me I am afraid of what I might do or say. Thanks for sharing.
Sounds like an absolute nightmare. I wondered what had happened.