AI prompt templates for content ops (copy-ready)

Content ops fails when every writer invents a new chat ritual. Templates fix that: same inputs, same gates, same quality bar. Below are copy-ready prompts you can paste into any LLM chat or API workflow. Pair them with the AI content ops workflow and the prompt cleaner.

How to use these templates

1. Keep a short style sheet (tone, banned phrases, product names) in a shared doc.
2. Fill every {{variable}} before sending. Incomplete variables produce vague drafts.
3. Run one stage at a time. Do not ask for outline + full draft + SEO meta in one shot.
4. Store the filled prompt next to the brief in git or Notion — chat history is not the archive.

Estimate token spend with the token estimator before you batch a week of drafts.

Template A — Intake brief expander

Use when a stakeholder dumps a half-sentence idea and you need a real brief.

You are a content ops editor. Expand the rough idea into a structured brief.
Do not write the article. Output Markdown with these exact headings:

## Audience
## Job to be done
## Primary search intent
## Must-include facts (with source placeholders)
## Must-avoid claims
## Format and length budget
## CTA
## Open questions for the requester

Rough idea:
{{idea}}

Product / brand context:
{{context}}

Template B — Outline from frozen brief

You are a senior SEO editor. Produce an H2/H3 outline only — no prose paragraphs.
Align structure to search intent and the brief. Flag any missing facts as [NEED FACT].

Brief:
{{brief}}

Primary keyword: {{keyword}}
Secondary topics: {{topics}}

Human gate: edit structure, freeze outline, then move on.

Template C — Section draft

Long pieces drift when you generate everything at once. Draft one section at a time.

Write the section titled "{{h2}}" for a practical guide.
Audience: {{audience}}
Tone: {{tone}}
Constraints:
- No fake personal testing claims ("I tested", "in our lab")
- Cite only facts provided below; otherwise write [NEED SOURCE]
- Target {{word_count}} words for this section
- End with a short checklist or actionable next step if natural

Frozen outline (for context):
{{outline}}

Facts for this section:
{{facts}}

Style rules:
{{style_sheet}}

Template D — Fact & claim audit

You are a skeptical fact-checker. Review the draft. List every claim that needs a source,
every number that looks estimated, and every absolute ("always", "never", "best").
Do not rewrite the article. Output a table:

| Claim | Risk | Suggested fix |

Draft:
{{draft}}

Template E — SEO pack (after human edit)

Given the final edited article and primary keyword, produce:
1. Title tag options (3), ≤60 characters
2. Meta description (2), ≤155 characters
3. Slug suggestion
4. H1 confirmation (or one alternative)
5. Internal link suggestions using ONLY this allow-list: {{internal_urls}}
6. FAQ of 3 questions answered in 2 sentences each from the article

Do not invent product features. Keyword: {{keyword}}

Article:
{{final_article}}

Template F — Social / newsletter cutdown

Turn the article into:
- 1 LinkedIn post (≤1,200 characters), practical tone, no hype
- 1 X/Twitter thread outline (5 tweets, bullets only)
- 1 email subject + 80-word blurb

Article summary:
{{summary}}
CTA URL: {{url}}

Template G — Style normalization pass

Rewrite for clarity only. Preserve meaning and claims. Apply style sheet.
Remove filler openers ("In today's world…"). Keep technical terms.
Return the full revised draft.

Style sheet:
{{style_sheet}}

Draft:
{{draft}}

Operating rules for teams

• Version the templates. Put them in a repo (prompts/content-ops/v3.md). Bumping a version is cheaper than arguing in Slack.
• Separate system vs user. If you use an API, put role and hard constraints in the system message; put brief and facts in the user message. See production system prompts.
• Budget tokens. Section drafts + audits use more input tokens than people expect. Track weekly with a shared rate table.
• Never skip the human gate on claims, legal, pricing, or medical/financial advice.

Common failure modes

Failure	Fix
One mega-prompt for everything	Split stages; freeze outline
Style rules buried in chat	Dedicated style sheet doc
Model invents case studies	Ban “I tested” language; require sources
SEO meta written before edit	SEO pack only on final copy
Writers paste competitor dumps	Clean with prompt cleaner first

Minimal weekly cadence

Monday: fill briefs. Tuesday: outlines + freeze. Wednesday–Thursday: section drafts + audits. Friday: SEO pack + publish. Measure cycle time, not “AI vibes.”

If you need a full pipeline diagram, read building an AI content pipeline. For comparing writing models, see ChatGPT vs Claude vs Gemini for writing.