How to Brief an AI Writer for Better Research Output

Most people who order a research brief — from an AI writer or a human one — write the topic in about eight words and stop there. "The history of remote work." "How blockchain works." "AI and jobs." Then they get back something generic, because that's exactly what an eight-word prompt deserves: a generic answer to a generic question.

The fix isn't writing more. It's writing four specific things. Here's what actually moves the needle, based on the difference between briefs that come back useful and briefs that come back as filler.

1. Name the angle, not just the topic

"The history of remote work" could become a dozen different briefs. Is it about: - the technology that made it possible (email, VPNs, video calls)? - the labor economics (why companies resisted it, then embraced it)? - the productivity research specifically (does it help or hurt output)? - the cultural shift (what changed about how people think about offices)?

Each of those is a legitimate, useful brief. None of them is "the history of remote work" in general, because that phrase doesn't tell a writer — human or AI — which of the four you actually want. If you don't pick one, the writer has to guess, and the safest guess is always the blandest one: a surface-level tour that touches all four angles briefly and goes deep on none.

The fix: add one sentence that names the lens. "Focus specifically on the productivity research — what studies have found about output, not the technology or culture side." That one sentence cuts the possible-briefs space by 75% and the result gets noticeably sharper.

2. Say who it's for

A brief written for "anyone curious" and a brief written for "a hiring manager deciding whether to offer remote positions" are different documents even if the topic is identical. The second one needs different framing: what does a hiring manager actually need to decide? Probably not the full technological history — probably the productivity findings, the retention effects, and maybe a caveat section on what the research doesn't settle.

You don't need a paragraph of audience description. One phrase does it: "written for a hiring manager who's deciding whether to offer remote roles" or "for a student who needs background before a seminar discussion" or "for someone explaining this to a non-technical board." That phrase tells the writer what to include, what to assume as known, and what tone to use.

3. List your actual sub-questions

If you have specific questions, write them down as questions, not folded into the topic line. "The history of remote work, and I'm curious whether it actually started before the internet" is a topic plus a buried question. Pull it out: "Sub-questions I want addressed: (1) did meaningful remote work exist before the internet, (2) what changed in the 2000s vs. what changed in 2020, (3) is there a consensus on whether it increases or decreases output."

A brief that answers three questions you actually had feels completely different from a brief that covers "the topic" in general, even if the word count and overall quality are similar. Specificity in the request becomes specificity in the answer — almost mechanically.

4. Tell the writer what you'll do with it

This is the one people skip most, and it's often the highest-leverage sentence in the whole order. "I need this to open a presentation to my team" is different from "I need this as personal background reading before a job interview" is different from "I'm writing a paper and need a starting outline of the key debates."

Each of those implies a different structure. A presentation opener wants a hook and a short, quotable framing. Interview background wants the kind of thing you could paraphrase out loud without notes. A paper outline wants the debates laid out as debates — positions, not just facts. None of this needs to be stated at length; one clause at the end of your request usually does it: "...since I'll be using this to open a team presentation."

Putting it together

Compare these two order-form inputs for the same underlying topic:

Weak: "How blockchain supply-chain tracking works."

Strong: "How blockchain-based supply-chain tracking works, focused on the practical mechanics (what data actually gets put on-chain and why) rather than the general concept of blockchain. Written for a small-business owner evaluating whether it's worth adopting for their own supply chain — skip the crypto-investment angle entirely. Sub-questions: what's the actual cost/complexity to a business that adopts this, and what problem does it solve that a regular database doesn't? I'll be using this to decide whether to bring it up in a vendor meeting."

The second one isn't longer because it's padded — it's longer because it's four extra facts, each one cutting down the space of possible answers. A writer (human or AI) working from the second prompt doesn't have to guess what you meant by "how it works." It already knows.

The general version of this rule

This isn't specific to ordering from us. It applies to prompting ChatGPT, briefing a freelance writer, or even asking a colleague for a summary: vague input produces safe, generic output, because the writer has to hedge against all the things you might have meant. Specific input — angle, audience, sub-questions, end use — lets the writer commit to one version of the answer instead of averaging across all possible versions.

If you want a sharper starting point the next time you need a brief on anything: don't write more about the topic. Write one sentence each about angle, audience, open questions, and end use. That's the whole method.

---

Need a research brief built this way? Our Custom Research Brief service turns a well-specified request into an 800-1200 word organized brief — the more specific your order-form input, the sharper the result, since there's no back-and-forth to fill in gaps afterward.