Undark Submission Guidelines: Pitching Science Policy Stories

Key Takeaways

• Undark's own submission page is the authority on how to pitch it. Any editor address, word count, or section list circulating elsewhere should be checked against the current page before you send anything.
• Science policy pitches work when they name a specific decision, its administrative status, the evidence behind it, and who disagrees.
• AI disclosure rules differ by outlet and by date. Record prohibited uses, permitted uses, disclosure location, required detail, and the policy date before you draft.
• A reporting plan and a verified source list carry more weight with an editor than a polished subject line.
• Match the venue to the material. A three-week-old regulatory decision and a records-heavy investigation do not belong in the same format.

1. Introduction

A science policy pitch has two audiences. The editor deciding whether to assign it, and the readers who may include the researchers, regulators, and legislative staff whose work the story describes. Undark occupies that space, publishing reported journalism and commentary about science as it meets public life, and its readers check claims as a habit.

Most pitches that fail do so for reasons the writer could have fixed in twenty minutes. The story restates a paper the journal already publicized. The policy angle is asserted instead of documented. The pitch arrives three weeks after the decision it describes, or it argues a position while presenting itself as reporting.

This article covers the preparation that comes before you contact an editor: how to shape a science policy story for Undark's readership, what a reporting plan needs to contain, how to handle AI-assisted work under outlet-specific rules, and how Undark compares with other venues that publish policy writing. Undark's own guidelines remain the authority on contact details and current requirements. Treat everything here as preparation, and confirm the specifics at the source.

2. What Undark publishes, and where policy stories fit

Undark is a non-profit science magazine associated with the Knight Science Journalism program at MIT. Its editorial range covers research and the public decisions surrounding it: public health, climate, biotechnology, technology regulation, and the institutions that fund or constrain them. That range is why science policy pitches fit there, and also why they are judged more strictly than a pitch about a single study.

A policy story needs a decision at its center. Pending rules, appropriations votes, agency guidance, court rulings, and international standards all qualify. So does the absence of a decision, when delay has measurable consequences for identifiable people.

Once the decision is fixed, the reporting burden becomes concrete:

• What is the legal or administrative status of the action, and which date matters next?
• Which studies, datasets, or cost estimates did the deciding body cite?
• Who disputes that evidence, and on what specific grounds?
• Whose work, health, or budget changes if the decision stands?

Undark separates reported journalism from opinion and commentary. Decide which one you are pitching and say so in the first paragraph. An essay arguing for a policy position is a reasonable pitch, provided it arrives labeled as commentary with the argument's evidence visible. Reporters and editors there will want to know which parts of your story are documented and which are inference. Have that distinction ready before you write the email.

3. A pitch structure that gives an editor what they need

Treat the pitch as a short evidence memo. Editors read it to answer three questions: is the story real, is it timely, and can this writer deliver it. Structure the email so those answers appear early.

1. Subject line naming the decision and the stake, in plain words.
2. One-paragraph summary: what happened, who is affected, why the moment is now.
3. The gap: what existing coverage missed or what is not yet public.
4. Reporting plan: documents, datasets, records requests, and named interviewees you have approached.
5. Evidence map: claims you can support today and claims that depend on reporting you have not finished.
6. Author note: relevant expertise, access, and two or three clips.
7. Logistics: proposed length range, deadline flexibility, and any date the piece depends on.

Leave out rhetorical questions and claims of firstness you cannot document. The table below lists the problems editors encounter most often and the fix for each.

A pitch does not need to prove the story is finished. It needs to show that you know what finished looks like, that you have started, and that the remaining work is feasible on a realistic schedule.

4. AI disclosure rules are outlet-specific

Publication policies on AI assistance differ, and they change without much notice. Before drafting, record five items for the outlet you are pitching: prohibited uses, permitted uses, where a disclosure must appear, what detail it must contain, and the date of the policy you read. The date matters because a rule written eighteen months ago may have been replaced, and an editor will apply the version in force when you submit.

Read the policy by task, not by tool. Translating a source document, checking grammar, and generating analysis are different activities, and an editor may classify them differently. A product's marketing label does not decide the category. If a policy bans "AI-generated content" and you used a model to translate a German-language regulatory filing, ask the editor for an interpretation instead of assuming the answer from another outlet's more permissive wording.

Keep a private process log while you work. Record the tool, its version if you know it, the date, the task, the material you supplied, and how you checked the result. The log makes an accurate declaration possible when one is required, and it helps you answer the harder question of who stands behind each factual claim and citation. That responsibility stays with the author. A disclosure does not repair an invented quotation or a reference to a study that does not exist.

Science policy raises the stakes on this point. A fabricated citation in a policy story is easy to check, and the readers most likely to check it are the researchers, regulators, and advocacy staff whose work the piece describes. Your first byline at a magazine of Undark's kind is judged on accuracy before voice.

5. Comparing venues for science policy writing

Undark is one option among several, and the right venue depends on how much reporting you have done and how fast the topic moves. Each outlet below has a different bar, audience, and tolerance for uncertainty.

The choice usually comes down to two variables. How much reporting exists already, and how quickly the topic expires. A three-week-old draft rule suits a newsbrief. A story that needs records requests and six interviews belongs in a magazine that can wait for them. Format your material to the outlet rather than sending the same pitch everywhere with the name changed.

Before submitting to any venue, confirm these five items on its current pages:

• The live guidelines document, not an archived copy or a third-party summary
• Payment terms and the rights the outlet requests
• The AI policy, including its effective date
• Whether simultaneous submissions are permitted
• The editor or desk that currently handles pitches

6. FAQ

Q1. Does Undark accept pitches from writers without a science journalism background?

Editorial decisions usually rest on the reporting plan and the evidence, so a byline history in another field is not automatically disqualifying. What matters is access and judgement: whether you can reach the relevant documents and people, and whether you can tell which claims are settled and which are contested. If you come from law, medicine, engineering, or public administration, say so in the author note and connect that background to the specific reporting the story requires.

Q2. Can I use AI tools while preparing a pitch?

Check the outlet's rules first, because policies often address submitted manuscripts and say nothing explicit about pitch emails. When the pitch stage is unstated, ask the editor for an interpretation rather than importing a permissive rule from somewhere else. Keep a process log regardless. The log costs a few minutes and settles questions about what you did and when you did it. Verify every reference yourself, since a tool cannot take responsibility for a claim you publish.

Q3. How long should a science policy pitch be?

Long enough to demonstrate that the reporting is under way, short enough to read in two minutes. One page of prose plus a short source list is a workable target for most magazines. Long attachments describing the topic are less persuasive than four concrete sentences about documents you have obtained and people you have contacted.

Q4. What should I do if the guidelines page does not mention AI at all?

Ask a specific question of a specific person. Describe the task, the tool category, and what you would disclose, then request the editor's classification. This produces a written answer you can rely on, and it demonstrates that you read the guidelines closely. Never assume that silence means permission, and never carry over an allowance from a different publication.

7. Conclusion

Undark's submission page and the equivalent pages at every other outlet you are considering decide whether your pitch gets read. Read the live version, note its date, and write to match it. Science policy pitches that succeed share a recognizable shape: a named decision, a documented evidence base, a reporting plan with real names and records, and an honest account of what remains unproven.

Two habits protect you across outlets. Keep a dated record of the tools and sources involved in producing the manuscript, so a declaration is a copy of your notes rather than a reconstruction. Verify each citation against the primary document, because in science policy the readers checking your work are often the people it describes. Everything else, from the subject line to the length range, follows from those two.