In July 2026, Vox added one sentence to its pitching guidelines: “We’d like to hear your voice, so don’t use an AI chatbot to write your pitch.”
That is a revealing line to put in a public document, and it is the most useful starting point for anyone working out how to pitch Vox right now. A publication does not add that instruction because of a theoretical concern. It adds it because the inbox filled up with competent, well-structured, entirely interchangeable pitches, and the editors needed a way to say that the thing they are selecting for is not competence.
Everything else in the guidelines points the same direction, and the rest of this is about what the direction implies.
Why Did Vox Add a Line About Chatbots?
Because the pitch is not a document. It is an audition for a working relationship, and the guidelines say so in a way that is easy to skim past.
Vox states that it rarely accepts already-written pieces and prefers to shape stories with writers from start to finish. That sentence describes a collaborative editing process, which means the editor reading your email is assessing whether a month of back-and-forth with you would produce something good. A pitch with no detectable author tells them nothing about that, however clean it reads.
There is a second reason the line matters, and it is about what gets rejected rather than what gets accepted. A pitch assembled from general knowledge about a subject will be fluent and will contain nothing an editor cannot already assume. Fluency used to be a filter because it was scarce. It is not scarce now, which means it has stopped being evidence of anything and the burden has shifted entirely onto specificity.
The uncomfortable implication for writers is that a pitch now has to contain information, not just framing. A well-argued case for why a subject deserves coverage is no longer a pitch. It is a suggestion, and suggestions are free.
There is a practical reading of this that is more useful than any advice about authenticity. Your pitch should contain at least one sentence that only you would have written. A judgement you hold, an observation from the reporting you have already done, a reason you find the subject strange. Not a flourish. A specific thought.
The structural instructions support this. Everything goes in the body of the email, with no attachments, which keeps the whole pitch in a single readable voice rather than splitting it between a cover note and a document. The bio is a line or so, not a biography. The samples should be links to previously published pieces similar in subject or form, which is a narrower ask than a portfolio: they want to see you doing this specific thing, not everything you have done.
The Three-Why Frame
Vox asks for a concise description, around a couple of paragraphs, of a piece with a strong unique angle, covering the story itself, why it matters now, and why you should write it.
That is three questions, and treating them as three explicit beats rather than as a vibe is the single most effective way to structure the email. Call it the Three-Why Frame.
Why this. Not the topic, the mechanism. Vox’s house form is explanation, which means a pitch succeeds when it identifies something readers think they understand and do not. “Rents are rising” is a topic. “The thing driving rents in these four cities is a change in how insurers price flood risk, which nobody has connected to housing cost” is a mechanism. The second has a shape an editor can see.
Why now. This is the beat writers treat as decoration and editors treat as load-bearing. A real answer names something that changed: a ruling, a dataset released, a threshold crossed, a policy taking effect. A fake answer says the subject is increasingly important, which is true of everything and therefore says nothing. If you cannot name what changed, the piece is not timely and the pitch should wait for the thing that makes it so.
Why you. Two legitimate answers exist. Either you have access that others do not, meaning named sources, a dataset, a location, or you have standing in the subject, meaning a background that makes your judgement worth more than a generalist’s. Enthusiasm is not an answer. Having pitched the idea first is not an answer either.
Write the three as three short paragraphs in that order. It will feel mechanical and it will outperform the version that tries to be graceful, because an editor scanning a queue can find all three answers in six seconds.
What Will Vox Not Run?
The guidelines are unusually specific about exclusions, and reading them saves you a wasted send.
Traditional reviews are out. Books, television, films, albums and other cultural products are named directly. This is the exclusion that catches the most writers, because Vox publishes a great deal about culture and the inference that it must therefore want criticism is reasonable and wrong. Vox writes about why a cultural thing is happening, not whether it is good.
Shopping roundups are out, which is worth noting mainly because a lot of freelance culture work in the current market is exactly that.
Q&As are out as a default, with a conditional opening: acceptable when you have access to a hard-to-get subject and a compelling reason the format serves the story. Both conditions, not either. A Q&A with someone available to any reporter fails the first. A Q&A with a genuinely inaccessible subject still needs an argument for why a conversation beats a written piece drawing on it, and the honest answer is usually that it does not.
It is worth understanding what this exclusion list has in common, because the pattern predicts the next thing that will be rejected. Reviews, Q&As and shopping roundups are all formats where the value sits in the thing being covered rather than in the explanation. A review’s worth depends on the book. A Q&A’s worth depends on the interviewee. A roundup’s worth depends on the products. None of them require the writer to have worked out how something functions.
Vox’s form inverts that. The value is in the explanation, which means the writer has to supply the thing that makes the piece worth reading. Any pitch where the subject is doing the work will read as one of the excluded formats even if it is technically none of them.
What is left, by elimination, is the explanatory reported piece and the essay that carries an argument about a mechanism. That is a narrower target than the site’s breadth suggests, and aiming at it directly is most of the work in learning how to pitch Vox.
Pre-Reporting Is the Part You Cannot Fake
The guidelines ask you to include any pre-reporting and source access, and state that diverse sourcing is of paramount importance.
Take the first part literally. Pre-reporting means work already done before the pitch, and saying so specifically changes the email. “I have spoken with two of the four insurers involved and have the 2025 rate filings for three states” is pre-reporting. “I would speak with insurers and review filings” is a plan, and a plan is what every other pitch in the queue also has.
This is the beat that separates pitches from ideas, and it is why pitching into a subject you already work in is so much more efficient than pitching into one you have only read about. The first comes with pre-reporting attached as a byproduct. The second requires a week of unpaid work before the email is worth sending, which is sometimes worth doing and should be a deliberate choice.
One caution on how to present pre-reporting without overclaiming. Say what you have and what you do not, in the same sentence. “Two of the four insurers have spoken to me on background, the other two have not responded, and I have the filings for three of the five states” is a stronger sentence than a version that implies full access, because an editor who commissions on the basis of access you do not have will find out in week three and that is a worse outcome than a pass.
Editors are also reading for whether you understand the difficulty of your own story. A writer who says a source will be hard to reach and explains their second route to the same information is demonstrating judgement. A writer who treats every access question as solved is demonstrating that they have not tried yet.
On diverse sourcing, the practical test is whether your source list contains more than one vantage point. A piece about a housing policy sourced entirely from economists has one vantage point, however varied the economists. Add a tenant, a small landlord, a municipal official and the piece can say something the economists cannot. Name the sources in the pitch rather than describing them as categories, because the editor is assessing whether your list can actually carry the piece, and a category is not a person who will take your call.
The Headline and Word Count Are a Test
Vox asks for a proposed headline and a proposed word count, and both requests are doing more work than they appear to.
The headline forces you to state the piece in one line. Writers resist this because a headline feels like the editor’s job, which it eventually is, but a proposed headline you cannot write is a strong signal that the piece has no single claim yet. Write it as a sentence that makes an assertion rather than as a label. “The flood-insurance change behind four cities’ rent spikes” beats “Rents, risk and insurance” because the first promises a finding.
The word count forces you to state the form. A proposal at 900 words and one at 3,000 words on the same subject are different pieces with different reporting loads and different slots in the publication. Getting the number approximately right tells the editor you have thought about what kind of thing this is. Getting it wildly wrong, in either direction, tells them you have not.
A rough calibration: an explanatory piece that turns on one mechanism and three or four sources tends to sit between 1,200 and 1,800 words. A reported feature that needs to establish a pattern across places or cases runs longer. An argument essay drawing on existing material rather than new reporting runs shorter. Pick the number that matches the reporting you have described, and if those two do not match, the problem is upstream of the pitch.
Put all five elements together and the email is short: three why paragraphs, a line on pre-reporting and sources, a proposed headline and word count, a sentence about you and two links. That fits on a screen, which is where it should fit.
So before you send it, read it back and ask the question the guidelines are really asking. Is there a single sentence in here that nobody but you would have written?