Turn an AI citation gap into a content brief your writer can use

To turn an AI citation gap into a content brief, start from one prompt where another source is cited and you are not, then write down five things for your writer: who the page is for, the specific questions it must answer, the evidence only you can supply, the pages you already have, and a checklist for accepting the draft. This post gives you that brief as a copyable template and fills it in once for a fictional company. The brief describes a hypothesis about what the cited pages have that yours lacks. It does not promise a format, length or structure will earn a citation.

In short

• Start from a gap you have already chosen. Choosing it is a separate job, covered in Find the citation gaps that deserve your next content update.
• A useful brief names one prompt, one audience, the questions to answer, the evidence to collect, and the pages that already exist.
• Decide first whether the brief is for an update to a page you own or for a new page. The rest of the brief changes with that answer.
• Write the acceptance checklist before the draft exists, so "done" does not depend on taste.
• Every suggested change is a hypothesis. Record it so you can review it later without pretending it was proven.

What is a content brief in this context?

A content brief is a one-page task description that lets a writer produce the right page without a meeting. In this context, it is built from a citation gap: the difference between what the pages AI engines cite for a prompt contain and what your own page offers, or a situation where you have no page to compare at all. A citation is a page that an AI answer links as a source. A prompt is a question a monitoring tool asks AI engines on your behalf; it is not a keyword and not something a user typed.

The brief exists because a gap on its own is not a task. "A competitor's page is cited and ours is not" tells a writer nothing about what to write, for whom, or what proof to gather. The brief converts the observation into decisions.

Start with a gap you have already chosen

A brief should begin from one gap, one prompt cluster and one decision. If you are still deciding which gap deserves attention, do that first with the prioritisation worksheet in the gap-selection post, then come back.

In DiscoveredBy, a prompt only becomes a candidate for a citation gap when you appear in under 30% of that prompt's own runs and at least one citation in the window points at a domain that is not yours. An opportunity is created only if a difference was found, and it carries that evidence: what the cited pages contain, what your matched page does about it today, and links to the competing URLs. When no page of yours is a close enough match, the recommendation is to create a page rather than fix a weak one. Those two outcomes, update or create, decide which brief you write.

Enlarge image

Accepting an opportunity creates a candidate topic on the Articles screen, carrying the diagnosis with it, but nothing is written until you pick that topic. That is a useful pause. It is the moment to write the brief yourself, or to check the one the tool prepares.

What goes into a writer-ready brief?

A writer-ready brief has eight parts: the gap, the decision, the audience, the questions, the evidence, the existing pages, the constraints, and the acceptance checklist. Each part answers a question a writer would otherwise have to ask.

The sections below explain how to fill each one, then the full template follows.

Enlarge image

The parts that need the most care

Gap record. Copy the prompt wording exactly, the engines and channel, the window, the cited URLs, your visibility and the competitor citation count. These are baseline measurements for a dated window, not live numbers, so record the date. State what you observed, not why the engine did it: "the cited page has a comparison table" is an observation; "it is cited because of the table" is a hypothesis and should be labelled as one.

Decision. Choose the asset before the content. In DiscoveredBy, the writer's brief decides the asset, from updating a page you own to publishing something new to holding off, starting from a rule-based guess that the writer then confirms or overrides. Only a brand new article or a comparison page is actually drafted; for every other outcome the brief is the deliverable. Do the same by hand. If the missing evidence belongs on another publisher's site, this is the wrong document.

Audience and questions. Name the reader in one sentence and list the questions in the order a buyer would ask them, starting from the prompt itself. Six to eight is plenty. Mark each as a product fact (only you can state it), a general explanation, or a judgement needing an owner; the type tells the reviewer what to verify.

Evidence to collect. This is the part briefs most often skip. List every fact, figure, quote or example the writer cannot invent, with an owner and a source. DiscoveredBy's writer treats this the same way: for a new page that needs proof only you can supply, generation can pause at "Needs your input" and, when the brief named what it needed, list it under "Proof only you have". Gathering that proof before the draft starts saves you the stall.

Existing pages and constraints. List the pages that should link to the new one, pages that could contradict it, and the cited pages as reference reading. Tell the writer to learn what the cited pages cover and then use your own words and evidence; DiscoveredBy's drafting step carries the same instruction. Add what must not be claimed, and a line stating that no format is expected to guarantee citations.

Acceptance checklist. Write it before the draft exists. Each line must be checkable by someone who was not in the briefing. "Reads well" settles nothing; "every listed question is answered and every claim names a source" does.

The reusable brief template

Copy this block into your document tool. Fill in every field; write "none" where a field does not apply, so a blank is never ambiguous.

CONTENT BRIEF: [working title]
Owner: [name]        Writer: [name]        Reviewer: [name]
Brief date: [date]   Draft due: [date]     Review due: [date]

1. GAP RECORD
Prompt (exact wording):
Engines and collection channel where the gap appeared:
Window observed (dates):
Our visibility for this prompt in that window:
Competitor citations for this prompt in that window:
Cited pages (URL, and what each one covers):
   a.
   b.
   c.
Recorded as of: [date]. These are baseline figures, not live ones.

2. DECISION
Asset: [update an owned page / create a new page / hold off]
Target page (if updating): [URL]
Reason for this choice (one or two sentences):
What we are NOT doing, and why:

3. AUDIENCE
Reader (one sentence):
What they are trying to decide:
What they will do next after reading:

4. QUESTIONS THE PAGE MUST ANSWER
(Order as a buyer would ask them. Type: F = product fact, E = general
explanation, J = judgement needing an owner.)
   Q1. [question]                                     Type:
   Q2.                                                Type:
   Q3.                                                Type:
   Q4.                                                Type:
   Q5.                                                Type:
   Q6.                                                Type:

5. EVIDENCE TO COLLECT
(Nothing here may be invented by the writer.)
   Item | Needed for | Owner | Source | Due | Received (Y/N)
   1.
   2.
   3.

6. EXISTING PAGES
Pages that should link to the finished page:
Pages that could contradict it, to reconcile first:
Pages to read as reference, and what to learn from each:
Do not copy phrasing or claims from any cited page.

7. CONSTRAINTS
Must not claim:
Must not compare or name:
Must stay private:
Regulated or sensitive topic needing extra review (Y/N):
No structure or format is expected to guarantee a citation.

8. HYPOTHESES TO TEST (label each as a guess)
   H1. The cited pages have [X]; ours does not.
   H2.
How we will review: [prompts, engines, channel, window, dates]

9. ACCEPTANCE CHECKLIST
   [ ] Every question in section 4 is answered.
   [ ] Every evidence item in section 5 is present and sourced.
   [ ] Every factual claim names a source a reviewer can open.
   [ ] No line from section 7 "Must not claim" appears.
   [ ] Links from and to the pages in section 6 are in place.
   [ ] Wording is our own; nothing is lifted from a cited page.
   [ ] Publish details ready: slug, title tag, meta description.
   [ ] Publication date and monitoring review dates recorded.

A worked example: Quillstone

Illustrative example: Quillstone and its competitors are fictional, and the numbers are made up to show the method.

Quillstone sells document-review software to mid-sized legal and compliance teams. Its team has already chosen this gap from a shortlist: the prompt "What should a compliance team check before choosing document review software?".

In the window the team looked at (28 days), Quillstone appeared in 2 of that prompt's 12 runs, which is 17%, under the 30% line. Across those runs, ten citations pointed at other domains: five to a Brieflane buyer's guide, three to a Clausewise checklist, and two to a Docket North blog post. Quillstone has no page that answers this question; its closest page is a general product overview, and it matches poorly. So the decision is to create a page.

Here is the brief filled in.

CONTENT BRIEF: A compliance team's checklist for choosing document review software
Owner: Head of Content   Writer: Freelance writer   Reviewer: Product marketing lead
Brief date: 12 May       Draft due: 26 May           Review due: 30 May

1. GAP RECORD
Prompt (exact wording): What should a compliance team check before choosing
  document review software?
Engines and collection channel: two engines, as recorded in the tool
Window observed: 28 days ending 10 May
Our visibility for this prompt: 2 of 12 runs (17%)
Competitor citations for this prompt: 10 (Brieflane 5, Clausewise 3, Docket North 2)
Cited pages:
   a. Brieflane buyer's guide: a long checklist grouped by risk area
   b. Clausewise checklist: a short list with a comparison table
   c. Docket North blog post: opinion piece, no checklist
Recorded as of: 10 May. Baseline figures, not live ones.

2. DECISION
Asset: create a new page
Target page: none (product overview is a poor match)
Reason: no owned page answers the buyer's question; the overview cannot be
  stretched to do so without losing its own job.
Not doing: a comparison page against named vendors.

3. AUDIENCE
Reader: A compliance manager at a mid-sized firm who has been asked to shortlist
  document review tools.
What they are trying to decide: which requirements to write into a shortlist.
What they will do next: draft an evaluation list and request demos.

4. QUESTIONS THE PAGE MUST ANSWER
   Q1. What should the tool prove about audit trails and access control?   E
   Q2. How does Quillstone handle audit trails and access control?         F
   Q3. Which file types and volumes should a pilot include?                E
   Q4. What are Quillstone's supported file types and limits?              F
   Q5. What should a pilot measure before a full rollout?                  J

5. EVIDENCE TO COLLECT
   1. Supported file types and size limits | Q4 | Product | Docs page | 16 May | N
   2. Audit-trail and permissions description | Q2 | Security lead | Docs page | 16 May | N
   3. A pilot plan the sales team already uses | Q5 | Sales | Internal doc | 19 May | N

6. EXISTING PAGES
Should link to the new page: product overview, security page, pricing page.
Could contradict it: the security page uses an older wording for access roles;
  reconcile before drafting.
Reference reading: Brieflane guide (how it groups requirements by risk area),
  Clausewise checklist (how it uses a table). Do not copy phrasing or claims.

7. CONSTRAINTS
Must not claim: any compliance certification we do not hold; any speed or
  accuracy figure without a source.
Must not compare or name: other vendors.
Must stay private: customer names, pricing beyond the public page.
Regulated topic needing extra review: Y (legal and compliance wording)
No structure or format is expected to guarantee a citation.

8. HYPOTHESES TO TEST
   H1. The cited pages group requirements by risk area; ours has no such structure.
How we will review: same prompt, same two engines and channel, 28-day window
  before and after publication, compared like for like.

9. ACCEPTANCE CHECKLIST
   [ ] Q1 to Q5 are all answered.
   [ ] Evidence items 1 to 3 are present and sourced.
   [ ] Every product fact names a docs page a reviewer can open.
   [ ] No certification or performance claim appears without a source.
   [ ] Links from the overview, security and pricing pages are in place.
   [ ] Wording is Quillstone's own; nothing lifted from the cited pages.
   [ ] Slug, title tag and meta description drafted.
   [ ] Live URL and review dates recorded.

Notice what the brief does not say. It does not say the page will be cited, and it does not name a target citation count. The hypotheses are written as guesses about what the cited pages have, and the review plan says how the team will look at the result later.

How does this map to the DiscoveredBy writer?

Enlarge image

In DiscoveredBy the same flow runs through Articles. Before anything is written, the tool reads the pages cited for the prompt and pages of yours, decides the asset, and, on plans that draft, writes a new article or comparison page from the brief and those pages. Each claim the writer reports gets a label; "verified by source" is awarded only when the quoted evidence is at least four words long and appears in the scraped text of the page it names and shares real terms with the claim. Claims that do not come out verified count against the citation readiness score, which is a checklist score for one draft, not a ranking.

Two limits matter. The check covers only the claims the writer listed, not every sentence, so a human still reads the draft. And DiscoveredBy never publishes for you: you copy the draft out, or, depending on your plan (plans and limits), send it to a connected WordPress site as a new draft post, never a published one. Once you paste the live URL back, an article that came from a gap moves that opportunity into monitoring, with day 7, 14 and 30 checkpoints recorded. For an update to a page you own, the smaller route is Optimizations, whose draft is a starting point to check, not text verified word for word.

Common mistakes and what this cannot tell you

• Writing to a format instead of a reader. "Add a table and an FAQ" is a hypothesis about what cited pages have in common, not a plan. Ask what a buyer needs and let the format follow.
• Copying the cited page. The brief should teach the writer what the cited pages cover, then require your own evidence and wording.
• Leaving evidence to the writer. Anything only you know (limits, pricing, customer proof) must be supplied, or the draft will either invent it or stall.
• Skipping the decision. Writing a new article when an existing page should have been updated splits your own content. Decide the asset first.
• Treating the acceptance checklist as a guarantee. A page that passes every line can still go uncited. The checklist tells you the brief was executed, not that the result will follow.
• Reading the gap as a cause. A citation gap records what cited pages contain compared with yours. It does not reveal why an engine chose them, the impact score is a relative order, and confidence reflects how much evidence was found, not how likely a fix is to work.
• Forgetting to plan the review. If you do not record the prompt, engines, channel and window now, you cannot compare fairly later. See Did your content update help? for the before-and-after review.

Frequently asked questions

Should the brief tell the writer to match the cited pages?

No. Use the cited pages as reference for what the topic covers, then ask for your own structure, wording and evidence. Matching a competitor's page adds nothing that page does not already provide, and DiscoveredBy's own drafting step is instructed not to reuse a competitor's phrasing or claims.

What if no page of ours exists for the prompt?

That is the case where the brief asks for a new page. In DiscoveredBy, a gap with no close-enough page of yours is treated as an opportunity to create one rather than fix a weak match, and accepting it creates a candidate topic on Articles. Write the brief for a new page and list the proof only you can supply.

What if the page is retrieved by an engine but never cited?

That is a different problem from a missing page, and a brief will not fix it. Start with Your page was retrieved but not cited, and read Retrieved vs cited for what your collection channel exposes.

How do I know the brief worked?

Compare the same prompts, engines, channel and window before and after, and accept that "no change" and "not enough evidence" are valid results. Observing a change is not proof the page caused it. For an article that came from a gap, DiscoveredBy records day 7, 14 and 30 checkpoints once you link the live URL, but the judgement about what they mean is still yours.

Next step

Take the gap you have chosen and fill in the template above for it. If you monitor your prompts in DiscoveredBy, the Citation gaps detail page gives you the gap record, and Articles turns an accepted opportunity into a brief and, where the asset is a new page and your plan drafts, a draft you can review. You can see how the workflow fits together on the Optimize feature page, or sign in and open your citation gaps.