Substack's Mike Cohen Says His Algorithm Just Wants To Be Your Friend, So I Ran Some Tests

The mucky mucks at Substack keep telling us they want to "connect us with the things we care about." What they don't say is that they get to decide what we care about.

I watched an interview the other day with Mike Cohen, Head of Artificial Intelligence and Machine Learning Engineering at Substack. He told the interviewer that it was unfortunate that the word “algorithm” had fallen into disrepute. Substack’s algorithm, he explained, is completely benign. He claimed the Substack algorithm that controls which articles you are presented when you visit the site is only helping users connect with the things they care about.

It’s a phrase you see all over the place on Substack’s website. Helping users find content they care about…helping readers find writers they care about…helping us build communities we care about…and so on. What we care about is something of an obsession at Substack. And it’s all for the sake of Mike Cohen’s algorithm.

Of course, it’s not only at Substack where you find this behind-the-scenes algorithmic manipulation. News feeds, playlists, wallpaper rotations, movie suggestions, ad servers—everybody’s just trying to give you what you like.

Frankly, I find it invasive and stultifying, this obsession with trying to figure out what we care about in order to show us what we care about. Don’t steer me, okay?

As a Substack reader, the algorithmic nanny is annoying, but there are ways to get around it if you are aware of it. As a Substack writer, however, the algorithm is more problematic.

What if your writing is fresh and original or what if it is about obscure esoterica or what if it takes an unflinching look at important but unpleasant truths? What if people don’t know they are interested in what you write about? If you are that kind of writer, do you want Mike Cohen’s algorithm steering readers away from you? It might not matter much to the readers if they don’t discover your writing, since they won’t be aware of what they’ve missed, but it’s a very big deal to you, the writer, if the algorithm is blocking new readers.

Because it’s an undeniable fact that, for every new writer the algorithm recommends to you on the grounds it’s figured out what you care about, some other writer has been bumped from consideration on the same grounds. And that’s the case even if Mike Cohen is telling the truth and his algorithm is 100 percent benign and only recommending writers on the basis of what you care about.

But what if it’s not 100 percent benign? What if Larry Ellison or Charlie Kirk’s donors have gained control of the algorithm? What if they allow the ACLU or the IDF or the GOP or the SPLC or the FBI to determine whom the algorithm recommends? Or doesn’t? In that case, it’s not just the edgiest writers whose concern it is, but everyone’s—everyone, that is, who opposes ideological despotism.

So, I tested that claimed impartial benignity.

First, I chose five keywords from my most popular post, “It’s Weird How It Took 20 Years for Anyone To Notice Six Million Jews Were Missing”:

1. whacky

2. yodeling

3. niggling

4. spike

5. grossly

and put them into Substack’s site-wide search. Results returned? 0. I tried it with commas and without, with quotation marks and without. I searched in “publications” and in “posts.” Always, zero results returned. Even though that post has been viewed more than 11,000 times and has been liked nearly 500 times, and does, in fact, contain those five terms, Substack’s search function informs users that there is no article on the entire site that contains those five terms.

So, I tried again, this time with only three terms:

1. niggling

2. spike

3. grossly.

This time, there was one hit, “Trump’s Tariff Pain Is Knocking at Your Door,” by Andrew Egger, Cathy Young, and Jim Swift. Their article did, indeed, contain those three words and Substack’s search found it, even though the word “spike” was in the form “spiking.” However, it couldn’t find my article.

So, I tried with two words:

1. spike

2. grossly.

There were 104 hits across Substack. My article wasn’t one of them.¹

So, I tried again with a different article, “How Jews Are Behind Border Surge and Making You Pay for It.” I chose five words:

1. gentiles

2. Gottheimer

3. pantry

4. indigenous

5. loyalty

A search again produced zero results. I tried three:

1. gentiles

2. Gottheimer

3. pantry

Again, zero. I tried two:

1. indigenous

2. loyalty

A ton of articles contained those two words. Was my article one of them? I started scrolling down to get to the bottom of the list of returned results, and kept scrolling and kept scrolling. I finally gave up after I’d gone through more than one thousand articles. Mine wasn’t among them.

So, now I have reason to question Mike Cohen’s sincerity when he claims his algorithm is benign. But, perhaps Substack just has a shitty search function and the complete absence of my articles from the search results isn’t evidence of in-house, stealth censorship.

I ran another test.

Lee Kern is a Jewish chauvinist, a shrill, incautious Zionist extremist who writes a Substack called Lee Kern Substack. I first stumbled across a post he’d written titled “Hurt—Don’t Heal the Antisemite.” (I write about it here.) In that post he blatantly calls for violence against people who don’t like Jews, apparently oblivious to the fact that wanting to hurt people for not liking you is a good reason not to like you.

But, outside of the calls to violence, Lee Kern and I agree on the fundamentals. We both agree that the Holocaust story is useful propaganda, not history. We agree that the Holocaust story’s whole purpose is to shape the way gentiles view Jews. We also agree that WWII was fought to punish Germans for their antisemitism and not for the noble-sounding reasons we got from our history teachers.

So, given that Lee Kern and I agree on the fundamentals, while he openly calls for violence, and I don’t, the search function at Substack should fail to produce his articles like it fails to produce mine if there is no in-house censorship at Substack.

Right?

So I took an article of his at random, “The Ghoul of Gaza,” and searched on five key words from it:

1. memorabilia

2. Sinwar

3. grisly

4. bestowed

5. aura.

One hit. “The Ghoul of Gaza” by Lee Kern.

So, how about that? Pretty undeniably, Substack engages in stealth censorship.²

I write a lot about Jews, the behavior of Jews, and the role of Judaism in the world. I feel this subject matter is the most pressing issue facing the us today. I honestly believe the fate of our country and of our civilization hangs in the balance. I might be wrong about that, of course; Judaism may play a very insignificant role in the world; it may not be a pressing issue at all.

On the other hand, I might be right.

Either way, I make my arguments as cogently and as forcefully and as honestly as I can. Right or wrong, don’t my fellow Americans, my fellow white people, my fellow gentiles, my fellow humans have the right to consider them? What gives Substack the right to prevent you from hearing what I have to say?

No one becomes an antisemite to get rich. We tell the truth about Jews as we see it because we feel it is exceedingly important that our people hear it. We point to what they are doing to the Palestinians as evidence of that importance. We understand that Jews are a powerful, vengeful, cohesive group that controls most of the social and political levers in the West and so our effort is most likely quixotic and will come at great personal cost.

But we do it anyway. We do it because we are convinced there is nothing more important than countering the mountain of lies we’ve been told and waking up our people. When a free speech company like Substack sneakily decides who is allowed to say what to whom, it makes our task even more difficult and our effort even costlier.

It stings. Hugely disappointed in you, Substack.

1. I asked a friend who doesn’t have a Substack account to run the same test and the results were the same.

2. Back in 2020, I ran a similar test on Google. Here’s Steve Sailer’s article about that.