Create a content brief
built from real research
Not a template with your keyword pasted into it. Moose runs a live research pass first: the current top results, People Also Ask, the AI Overview and what ChatGPT says, plus the actual text of the pages that already rank. Then it writes the brief, and lists exactly what it read.
SamPremium
Hi, Moose uses the target query and content format to shape both the brief and the first draft. Different structures need different sections, proof, and calls to action.
Give EV shoppers a clear, evidence-based answer to which electric SUVs actually hold up off-road, tiered by real capability (heavy-duty vs. light-duty vs. budget AWD), while covering the practical trade-offs (range loss, cold-weather performance, ground clearance) that most competing roundups gloss over.
None of the three ranking pages give real ground clearance and approach angles next to cold-weather range loss. Put both in one comparison table near the top and the piece answers the question the AI Overview currently answers with two different sources.
Chief desk supervisor. Has never once written "aim for 1,500 words" without reading a single competing page first. Sets a low bar, and yet.
Most AI briefs are a template with your keyword dropped in
You can tell within ten seconds. The word count is a round number. The competitor section names companies the model remembers rather than pages that currently rank. The FAQ questions are invented, not the ones people actually ask. Nothing in it could be wrong, because nothing in it is checkable.
A brief is a set of decisions about what to write. If the decisions came from a model's memory of the internet, the writer inherits guesses and the piece goes live competing against pages nobody read.
What Moose collects before it writes a word
Six live sources and a real crawl, run at the moment you ask. Search now covers both the ten blue links and the answer engines that increasingly replace them, so the brief has to be built from both.
Titles, snippets and positions as they stand right now, not as a model remembers them. Your own domain is tagged as owned so competitors means competitors.
Up to eight real questions people ask around the topic. These become the FAQ angles, which is why they read like questions instead of headings.
The related searches Google surfaces, turned into secondary keywords with a tier rather than a flat list of terms to sprinkle.
The actual AI Overview text for your query and the list of URLs it cites. If an engine is already answering this, you get to read the answer.
What ChatGPT's search surface says about the topic and which sources it leans on. Often a different set of pages than Google ranks.
The real Google searches Gemini writes for itself when it researches the topic. These surface intent angles the ten blue links miss entirely.
Up to three ranking competitors and two pages cited by the answer engines, fetched and stripped to main content. The evidence is the page body, not a snippet.
The recommended length is the median body length of the pages it crawled. Not "aim for 1,500 words", which is a number with no source.
Informational, commercial or comparison, and which page format is actually winning. A versus query gets comparison framing and objections, not a listicle.
It shows its receipts, and it will tell you when it has none
Every brief ends with a list of exactly what was collected, generated from real counts. If the research could not run, the brief says so on its face rather than sounding researched.
How many organic results, how many pages crawled, how many PAA questions, whether an AI Overview and a ChatGPT snapshot were captured. Generated from real counts, so a thin research run produces a visibly thin list.
The model is instructed not to fabricate sources, competitor claims or traffic numbers that are not in the evidence bundle. If it is not in the research, it is not in the brief.
On a broad informational query it satisfies the dominant intent first and refuses to force a product-led angle unless the evidence supports one. That is usually the argument you were going to lose anyway.
Your voice, terminology and proof points influence how the brief is written. They never override what the query actually wants, which is the failure mode of most brand-first briefs.
No search connection means a preview brief, explicitly labelled, and explicitly told not to imply research happened. You still get a usable structure from your context library.
If search is temporarily down it still writes the brief and tells you what was missing, rather than handing you an error at the moment you sat down to work.
Everything a writer needs, and a plan for the week
Written to be handed to a person, not filed. Every section exists because someone would otherwise have to guess.
A landing page brief is not a blog outline in a different hat
You pick the format up front and the structure genuinely changes: different sections, different proof, different calls to action. The live SERP also gets a vote, so a comparison query gets comparison framing whichever format you chose.
Top-of-funnel or mid-funnel article structure with a quick-answer block, criteria, and an FAQ built from real questions.
Definition-first, deliberately short, written to be the thing an engine quotes when someone asks what a term means.
Value proposition, proof, objection handling and CTA blocks. Not an outline of headings wearing a landing page label.
Problem framing for a segment or use case, with the outcome stated before the mechanism and proof close to the claim.
What the feature does, who it is for, how it works, and where it sits in the workflow, with the honest limits included.
It already knows who you are and what you have agreed
Two kinds of context feed every brief. Neither one gets to overrule the search evidence, which is the whole trick.
Terminology to avoid is doing more work than you would think. It is the difference between a draft that sounds like you and one you rewrite.
Decisions and confirmed facts carry into the brief, so you are not re-explaining the campaign you scoped an hour ago in the same chat.
Built to be used, not admired
The brief is a step in the loop, not the deliverable. Everything after it happens in the same window.
No target query or no format means one short question, not a brief built on an assumption you have to unpick later. Answer it and the run starts immediately.
The research bundle is saved with the brief. Ask for a different angle and it reuses the evidence. Force a refresh when you want current data instead.
Markdown, Word, PDF or plain text. Briefs live in the Library with the project, so the next person on the topic starts from what was already researched.
The draft is written from the brief, in your brand voice, with internal link suggestions pulled from your own site index and the anti-slop writing rules applied. Those rules stay out of the visible brief so it remains readable.
When visibility monitoring finds a topic the assistants answer without mentioning you, it can generate the brief and the draft on its own and leave both waiting for your approval.
On BYOK you get the full research pass: live results, answer-engine snapshots, and real crawls. On BYOK Free you still get a brief, built from your context library and brand voice, and it is labelled a preview brief so nobody mistakes it for researched work. That label is the honest version of an upgrade prompt.
The way this normally gets done
Nine open tabs, an incognito window so the results are not personalised, and a brief template last edited by someone who left.
The tabs close. Six weeks later nobody can say why the brief recommended that angle, so the piece cannot be defended or improved.
Checking Google, ChatGPT, Gemini and the AI Overview by hand for one topic is twenty minutes. So it happens for the flagship piece and never again.
A word count with no source produces pieces that are too short to win or padded to hit a target. Both read exactly as badly as they sound.
Left to a template, the product angle wins every time. The query wanted a comparison and got a pitch, and the page never ranks.
One query, one format, about two minutes
Same research a good strategist would do, run in order, with the evidence kept so you can check any line in the brief.
If you leave either one out it asks rather than guessing. That single question is the difference between a brief you use and a brief you rewrite.
Live results, answer-engine snapshots, then the body text of the pages that rank. The recommended length is the median of what it actually read.
Structure, keywords in tiers, FAQ angles from real questions, and a dated action plan. One click turns it into a first draft in your voice.
The brief is the start of the conversation
Revisions reuse the saved research bundle instead of burning a fresh run, so arguing with it is cheap.
Questions people ask
How is this different from the brief an AI writing tool gives me?
Most of them generate a brief from the model's training data and a template. This one runs a live research pass first and then writes from that evidence, which is why the recommended length has a source and the FAQ questions are real. The receipts list at the bottom is there so you can tell the difference at a glance.
What if I do not have a SERP connection?
You get a preview brief, clearly labelled, written from your context library and brand voice with no live evidence. It is explicitly told not to imply research happened. Useful for structure, not something to hand a writer as researched work.
How long does a full run take?
Usually a couple of minutes, most of it spent crawling the competitor pages properly rather than reading snippets. Revisions are near-instant because they reuse the saved research bundle.
Does it crawl my own site as a competitor?
No. Your project domain is tagged as owned and excluded from the competitor crawl, so the competitor analysis is about other people. Your own pages show up separately, as internal link suggestions from your site index.
Can I change the format after the brief exists?
Yes, and it will restructure rather than relabel. A glossary brief and a landing page brief are genuinely different documents, so ask for the other format and the research is reused while the structure is rebuilt.
Where does the draft come from?
One click from the brief. It inherits your brand voice, the internal link suggestions from your own site index, and a set of writing rules that strip the usual AI tells. Those rules are kept out of the visible brief so the brief stays readable.
Can this run without me?
A workflow can trigger it. When monitoring finds a topic the assistants answer without you, it can generate the brief and the first draft and leave them waiting. Nothing publishes without your approval.
More things to ask Moose
All use casesAsk before you brief, and get the pages, the verdict and what is missing.
The version of you an engine can actually cite, built from your own crawl.
The searches Google writes for itself before it answers, and who it cites.
Twelve searches behind one prompt, including the ones naming brands it trusts.
Share drops on a topic, the brief and the fix get staged, you approve.
The full list, and what Moose does step by step in each one.
Brief the next piece
from evidence.
Download Hi, Moose, add your target query, and read the receipts at the bottom of the brief before you write a word.