What The Paris Review Teaches Dating Apps About Prose as Character
The Paris Review has spent more than seven decades reading prose as if a few thousand words could tell you who a person is. Founded in Paris in 1953 by Humes, Matthiessen, and Plimpton (Wikipedia, The Paris Review), it bet its reputation on one conviction: fiction and the long-form interview reveal more than any biography could.
That conviction is the editorial filter — and it is also, almost word for word, the premise behind a writing-first manuscript on Anketta.
What Anketta adds is mechanical rather than literary. The magazine reads to publish; Anketta reads to match. Every phrase a user highlights as love this or crosses out as no thanks is stored as a preference signal — the same reading-method Plimpton's editors used to assess a story, ported to two-person matchmaking. The point of this article is not that dating apps should publish a quarterly. It is that a 73-year-old literary editorial logic — prose first, image second — already exists, and the apps that ignore it are the ones quietly running out of new things to try.
Try Anketta — write the version of yourself the Paris Review would read.The magazine's founding generation read manuscripts at café tables on the Rue Garancière and decided early that the way someone wrote was the most reliable thing about them they could measure. Prose was the test from the first issue, and it has stayed the test ever since.
The "Writers at Work" interview series began in the first issue — E. M. Forster sat for the inaugural conversation — and within a decade it had been renamed The Art of Fiction and turned into the magazine's signature franchise. The current homepage still leads with a numbered entry in the series — at the time of writing, The Art of Fiction No. 269, an interview with Yan Lianke (The Paris Review, theparisreview.org); it sits alongside the Quarterly's current fiction selections.
The editorial premise that survived seven decades is simple: a person revealed through several thousand words of considered prose is harder to fake, harder to repackage, and easier to recognise than a person revealed through a headshot and a list of preferences. The magazine has been small — its 1989 circulation was 9,700 — but its taste-making influence sits far above its readership numbers. Editors at Knopf and FSG read it. So do other writers. So do the dating apps that have started to copy its premise, knowingly or not.
The shape of the editorial mechanism matters here. The Paris Review's reading process is patient: an editor reads a story, then re-reads sentences, then forms a judgment that is partly about prose mechanics and partly about the human implied behind them. The judgment is not "is this person attractive" — that question is not even posed. The judgment is closer to what kind of mind made this. That question is the entire point of an Anketta manuscript.
The literary intuition has empirical backing. The psychologist James W. Pennebaker built his career on the finding that low-level words can give clues to large-scale behaviors (Wikipedia, James W. Pennebaker). His tool, LIWC — the Linguistic Inquiry and Word Count — measures how often a writer reaches for pronouns, articles, and function words, and uses those rates to predict sex, age, personality, and even recovery from depression. The everyday vocabulary you use, in other words, is character.
The implications for matchmaking are concrete, and they sit one layer below what people consciously write. A person's pronoun balance reveals where their attention sits. Their adjective density tells you how much they qualify the world. Their sentence rhythm reveals whether they pause to think or run forward. These signals are present in any unguarded paragraph and absent in any photo. They cannot be edited out without rewriting the person.
What dating apps have done for fifteen years is throw away exactly these signals. The bio grid is the photographic negative of the literary magazine: it asks for the headline summary and rejects the prose. The signal-rich part of a person — the part you would read three pages of, then know — is squeezed into a height-and-hometown line. The reading-method behind The Paris Review's editorial filter cannot operate in that format. The format itself is the problem.
Anketta's writing surface is a free-form document. There are no questions, no intake prompts, no mandatory length floor — a manuscript is a blank file the user organises with headings and paragraphs, the same way a Paris Review submission would arrive. The product term inside the app is manuscript in English and анкета in Russian, and the editorial body — the part other users actually read — is whatever the writer decided to put there.
The matching engine reads that body the way a literary editor would: phrase by phrase, attending to what stands out. When a user highlights a sentence as love this, the underlying preference model stores the highlighted phrase. The next batch of manuscripts the algorithm surfaces will skew toward profiles whose prose contains the same phrase, or its semantic neighbours. Cross-outs work the same way in negative polarity. The longer a person swipes, the narrower and more accurate the queue becomes — without anyone ever filling out a personality test. The mechanism is described in more depth in our piece on text-vs-photo dating and the dating-for-writers post.
There is one more parallel worth naming. The Paris Review uses moderation as a quality filter — most manuscripts the magazine receives are not published. Anketta does the same: every user manuscript is reviewed by a language model for quality and authenticity before it becomes visible to other users. The bar is not "Plimpton-grade prose"; the bar is a real person writing honestly, in their own voice, at enough length that the reader has something to read. The editorial principle — that the writing has to be worth someone's attention — survives in the product mechanic.
The bio grid was designed for throughput. A profile that can be evaluated in two seconds at the cost of a left or right swipe is a profile that produces engagement metrics for the platform. It is not, by any reading-method that has survived since 1953, a profile that produces character signal. The two goals are not aligned.
The comparison sits cleanly on a single axis: what part of the person does the editorial filter actually filter?
| Reading surface | What gets filtered | Time-to-judgment | Character signal | What survives the filter |
|---|---|---|---|---|
| Literary magazine (Paris Review) | Prose, voice, sentence-by-sentence mind | Hours to days | Very high | Writers worth reading |
| Bio-grid dating app | Photograph + headline + checkboxes | Two seconds | Low — appearance dominates | Photogenic profiles, regardless of fit |
| Anketta manuscript | Free-form prose + highlight-driven preference | Minutes per profile | High — text first, image secondary | Writers other users actually want to talk to |
The middle row is the format dating-app users have been complaining about since at least 2019. The third row is the editorial logic of the first row, rebuilt as a matchmaking primitive. The thing that has been missing from the conversation is the lineage: this is not a new idea, and it did not need to be reinvented. It needed to be ported.
Try Anketta — write the version a slow reader would actually finish.The Paris Review is the most prestigious anchor for the prose-as-character thesis, but it is not alone. Three other journals make the same editorial bet in slightly different registers, and all three offer techniques that port directly:
- McSweeney's prizes formal play — lists, fake documents, footnotes-as-essays. The editorial signal is: structure is also voice. A user's manuscript that experiments with structure (a numbered list of red flags, a fake interview with themselves) reveals more than a paragraph in standard form.
- n+1 uses long-form cultural essays as the primary character vehicle — three thousand words on a film, a city, a recurring frustration. The editorial signal is: how a person frames a topic reveals what they actually care about, not just what they list as interests.
- Granta specialises in the quiet first-person — childhood, family, place. The editorial signal is: a person writing without performance is the most discoverable version of them. Anketta's free-form editor was designed to make exactly this register easier than the performed bio.
- The Paris Review itself uses the long-form interview to let a writer's mind unfold across hours of conversation. The dating-app analogue is the 48-hour-match window — a deliberately slow first chapter where two people get to write something together before either decides.
- All four magazines share a deeper rule: moderation is the editorial position. Not every submission goes through. The implicit promise is that what does go through is worth your reading time. Anketta's manuscript moderation is the same promise, miniaturised — a quality filter, not a gatekeeping one, but a filter nonetheless.
The combined lesson is unfussy. Long-form prose is a substrate that takes any of four editorial techniques — formal play, cultural framing, quiet first-person, conversational unfolding — and produces character signal that a bio grid cannot. Dating apps that read like a magazine subscribe to one or more of these. Dating apps that read like a spreadsheet do not.
The honest answer is yes, and not really. A two-minute manuscript read is roughly forty times longer than a two-second swipe, which sounds inefficient until you count the second-order cost of two-second swipes. The bio-grid economy produces a high volume of low-quality first messages, ghosting, and matches that go nowhere. The hour spent reading five manuscripts produces fewer matches but fewer of those matches collapse before the second message. The slow-dating literature has been making this case for two years; the Paris Review made it implicitly for seven decades.
There is also a selection effect. People who are happy to write a manuscript and read a few in return are filtering themselves on the front end. The user who walks away because the format is too slow is, in a useful sense, already filtered out. The remaining users are the ones who would read the magazine if you handed it to them — and those are the users any reading-first matchmaking platform was always going to be for.
The 48-hour structure makes the time investment finite. Once two people match, there is a defined window during which the early chat happens; after that window, the match either becomes permanent or it does not. The slow-reading premise is not "spend years on the app". It is "read carefully for an hour the first time, then trust the filter".
What is "prose as character signal"?
The idea, formalised in psychology by James Pennebaker and practised editorially by The Paris Review since 1953, that a person's vocabulary, sentence rhythm, and topic selection reveal stable features of their personality more reliably than self-report. In dating terms: a paragraph someone wrote is a more honest signal than a checkbox they ticked.
How is Anketta's manuscript different from a long bio?
A long bio is still a bio — written for the reader to evaluate quickly against a set of yes/no criteria. An Anketta manuscript is a free-form document with no questions and no length target. The intended reading-method is closer to reading a magazine column than to scanning a CV. The matching engine learns from what readers highlight, not from what writers list as interests.
Do I need to be a "writer" to use a manuscript-first dating app?
No. The corpus voice does not require professional polish. A person writing honestly, at enough length that a reader has something to read, is the entire bar — the same bar a first-time letter to a literary magazine would meet. The moderation layer rejects spam and impersonation, not unpolished prose.
How long should a manuscript actually be?
There is no enforced floor or ceiling. Longer manuscripts surface more potential topics of contact for highlight-driven matching, so they tend to do better at producing conversations. Shorter manuscripts narrow the writer's own match surface as a natural consequence. Most users settle around the length of a long magazine column once they get used to the format.
What does the highlight-driven preference model actually do?
When a user highlights a phrase as love this, the system stores that phrase as a positive preference signal. Future manuscripts the algorithm surfaces will skew toward those containing the phrase or its semantic neighbours. Cross-outs work the same way with negative polarity. The model becomes more accurate the longer a user swipes — without anyone ever filling out a personality test.
Is this the same as "AI matchmaking"?
It is closer to editorial matchmaking with an AI assistant than to the algorithmic-from-the-start framing of most modern apps. The reading-method is human; the storage and surfacing of preferences is machine. The Paris Review parallel is not "an AI publishes the magazine" — it is "an editorial filter that has always existed gets a much larger reading queue".
Unsure about writing? Try reading first.When the first manuscripts in the queue are read the way a Plimpton editor would read them — slowly, attentively, with the assumption that the prose is the person — the rest of the queue starts to feel different too.