Inside the Work with Emboss: What my Writing Process Reveals

On May 12, 2026, the literary journal Granta published a short story, "The Serpent in the Grove" by Jamir Nazir, in partnership with The Commonwealth Foundation Short Story Prize. This prize is a big deal—it has lots of eyes on it—and as such, it didn't take long for people to start scrutinizing this short story. In no time, the AI accusations were flying. Unfortunately, this case is not entirely unique. As Alyssa Matesic said in her video covering the situation, the literary world has no idea what to do with the ever-growing flood of AI-generated writing. Just think back to the Shy Girl debacle. (Note: Nazir's story was not the only one to receive AI accusations, but it did garner the most attention, and went on to win the Commonwealth Foundation's prize.)

In thinking about these situations, one thing is quite clear—there's a stigma against AI-use in the writing world. And that's good! There ought to be. LLM-produced writing is low-quality, trained on billions of stolen words, and shows no love for the self poured into the craft. As has oft been repeated in this new digital age: Why would I want to read something that someone didn't even take the time to actually write? And that's to say nothing of AI's environmental impact.

But while the idea of grifters trying to pass AI-generated content as original creative work is scary, there's another fear that consequently arises. Are writers going to have to field AI accusations forever? Even Substack is aware of the problem; the platform recently announced a new partnership with Pangram, a prominent AI-detector (and among those used to check the Commonwealth Foundation stories). Either Substack users opt out of the feature and draw suspicion to their work, or they opt in and allow AI to train on their writing. It's a complete lose-lose either way.

So, what's the pro-human solution? Is there any way for writers to prove their humanity without the use of AI? Will there be a constant need for authors to defend their work when they never touched an LLM or generative model in their entire process?

Well, not necessarily. Odds are good that writers will always find solutions—we're creatives, after all—but there's one excellent solution available to all of us right now: Emboss. Ellipsus's new Emboss feature is one option writers have to ward off AI accusations.

When an artist wants to prove their pieces are human-made, there's no better way to do that than to show the process. Multiple visual artists now record their steps to the finished piece or save drafts to show their work's progression. Emboss similarly gives readers insight into the writing process.

The word "emboss" in the traditional sense means to create a raised pattern on a material. Historically, the carved designs were used as a signal of trust between parties, as in the case of notary embosses. In the same way, Emboss by Ellipsus is a way to share the chronicle of your writing journey—the many pauses, tweaks, and drafts made to reach the end—and so inspire trust in your readers that your writing is human written.

When you go to share a reader link to a document with Ellipsus Plus, you have the option to emboss it and share your Writing Journey and/or version history. You can see your writing metrics any time you want, and you can choose to share those metrics with the audience as well.

As you can see in the screenshot below, when you go to share a reader link with your audience, you can toggle the Emboss features on and off at will. Writing Journey will let your readers see metrics such as word count, time spent hitting keys, editing choices, and more, all covered later in this post. The Version History lets readers see every draft of the document that Ellipsus autosaved throughout your writing session(s).

To be clear: Emboss doesn't ‘predict’ if your work is AI-generated based on the finished product—that’s the business of “AI-verifiers,” which are known to be rife with errors and false positives. AI-verifiers are a(nother) desperate, unreliable attempt by the AI industry to exert control over human writing.

By contrast, the tech underlying Emboss is AI-free and surprisingly simple. Ellipsus already syncs and autosaves your real-time writing: Emboss simply allows you to share those metrics in digestible form, whether in the form of public version history, or your writing journey, or both.

Instead of analyzing how "AI-sounding" your cadence and language choices sound, Emboss features show the progression of your pieces from start to finish as you write—thereby demonstrating every human judgment in a document's recorded history. The same processes that make Ellipsus a functional writing software now help demonstrate your humanity—your pauses for thought, your trials, your mistakes—every jot and title.

Let’s see this in action.

Let's take a look at this document containing three of my poems. Click on the clock icon at the top right corner of the page (just to the left of the settings gear) to view my version history. When making that click, you can see I created the document at 12:55pm on June 10, then stopped working on it at 3:12pm that same day.

Between starting and finishing, there were 137 drafts of the document saved. Other word processors also autosave the drafts you're working on—that's standard practice for writing software—but none of them are making those saved drafts accessible to you in case you want to go back, and none use that version history as a tool to showcase the hand of the author behind the work. And remember, this is all opt-in as far as sharing with your audience goes, but the same is true for every choice you make. Everything that the average word processor records can be made available to your readers with Emboss.

Taking a look at my document again, you'll see three of my own original poems. We'll read through them a little more later, but as you can see, I have three rough drafts here. If I go back into my version history and recover the draft from 1:58pm, you'll see one and a half poems, with the draft saved right as I was typing the word "The." And, if you take a look at the version history there, you can see I turned that draft into its own document on June 17 and haven't made any change to it beyond that.

Now, back to the final draft. If you look at the top left of the screen, right next to the document's title, you'll find my Writing Journey.

You'll notice that the actual amount of time spent putting words on the page was only eight minutes. A two hour session and only eight minutes of actual writing time (i.e. time hitting the keys)! If you're a poet, I'm sure you'll get it. If we look under the cadence section, we'll see that I took 69 pauses throughout writing. Nice. These were pauses spent thinking up the next word, phrase, or line, the kinds of pauses human writers make all the time, and the very same pauses that LLMs don't make.

And under the editing section, you'll see I made 139 edits in this 478-word document, making my word-to-edit ratio 3.4. And that makes sense. My followers know that I think the art of poetry is one that's very precise, so of course I aim for precision, even in my rough drafts, and that inevitably means a lot of editing as I go.

We can compare this to a script I wrote for a video that's a whopping 9,055 words long.

I spent an hour and 33 minutes of putting words to page across nine different writing sessions over three days. Throughout this document, I made 178 pauses and 506 total edits. My word to edit ratio was 17.9 in this case, which makes total sense to me; of course it would be higher for a script than it would be for a few poems, especially on a first draft.

With Emboss the focus is on the human labor behind a work—because that’s the only value that matters!

Now, I briefly want to talk about the Writing Insights feature, as it's closely tied with the Emboss feature. Just as Emboss takes a look at your pauses and edits, Writing Insights analyzes metrics like your word choice and sentence structure. With no AI whatsoever, Ellipsus can lay out quirks of your writing like your vocabulary diversity, rhythm, paragraph density, and more. The vocabulary diversity is especially helpful for me because one of my biggest weaknesses as a writer is the unintentional repetition of words, even in close proximity. Ellipsus can take all of these insights and give your writing an overall complexity rating.

I think, oftentimes, we're so close to our own writing that we can't see our own worst habits. And, obviously, the best thing for that is to have someone else read your writing, and we'll get into that soon enough, but these writing insights are great for learning about your own habits while also keeping known habits in check.

If we look back at my document of poems, we can see that I used 478 words total and 298 unique words among them, which I think is a decent ratio for poetry—especially poetry that doesn't intend to use repetition as a device. In looking at the word frequency, the most-used word is "headspace." After that, the most used word is "thoughts," followed by "old," "storage," and "lip." If you take a look at the poems for yourself, you'll see exactly why those words spring up so much.

The adverb checker is also very helpful because we can often unknowingly use way too many adverbs, the majority of which are redundant. In the case of these three poems, I only used one adverb, which was "flavorfully," which is rather unique.

While AI accusations are unfortunately impossible to avoid at this point, AI-verifiers are not the solution that tech companies would have you think they are. Even if AI-detectors were 100% accurate, they would still require feeding original work into AI, which is exactly what we're trying to avoid! When LLMs are trained on swaths of stolen content already, why would we willingly give up anything more to them? At the end of the day, Substack’s Pangram partnership is another step in the wrong direction.

We can't reasonably rely on AI to detect itself, nor to prove our humanity. As long as creativity has existed, humans are the only ones who can demonstrate and prove their own creative processes. Even in the case of the Commonwealth Foundation prize, AI-verifiers alone did not sway the public nor Commonwealth Foundation members.

The foundation's director-general, Razmi Farook, told the Guardian that:

... rather than surrender our judgment to AI-detection software, we asked our winners to show their working drafts, outlines, the evidence of an artistic journey. That software, it must be said, is not infallible: it returns inconsistent verdicts and, in doing so, corrodes the very trust on which a prize depends.

In other words, those at the top of the Commonwealth Foundation literally used the authors' "artistic journey" to prove what AI couldn't.

Turns out, when it comes to human authorship, the drafts, outlines, and evidence—the writing journey that Emboss catalogues on your behalf—have the final say. Humans: 1, AI: 0.

Lastly, no matter how you choose to support Ellipsus, be sure to check out their Discord! You can find other Ellipsus users with whom you can collaborate. Even if you don't find someone to comment on your work, it's always good to have an accountability partner—someone with whom you can write at the same time. That way, you're sure to get something done.

If you want to learn more about Ellipsus Plus, you can read all about it here, and if you'd like to access Plus for yourself, you can do so here.

Jonah Wardell is a writer from Salt Lake City, Utah. He will graduate from the Vermont College of Fine Arts with an MFA in Fiction and Poetry in January 2027, assuming the program doesn't kill him first (and it's doing its damnedest). He's not a real intellectual, but he plays one on his YouTube channel, Roughest Drafts.