Content Creation

Content Marketing Workflow

Research, draft, optimise and publish a marketing article end to end

Featured Verified Trending Beginner Multi Tool Semi-automated Freemium
Use this workflow
Est. time
~1h 35m
Est. cost
$2.4
Steps
8
Automation
Semi-automated
Trigger
Manual

Overview

Turn a topic into a published, optimised article without the usual week of back-and-forth. The workflow covers the full path — research, outline, draft, optimise, review, publish, measure — and names exactly which tool does what at each stage.

It is deliberately semi-automated: the AI does the research synthesis and the first draft, a human owns the edit and the final approval. That split is what keeps the output publishable.

Business value

Cuts the cost of a researched article from roughly a day of writer time to about ninety minutes of review, while keeping a human accountable for every published claim.

Expected outcome

One publish-ready article of 1,200–2,000 words with a target keyword, internal links, a hero image and a measurement plan attached.

Problems it solves

  • Blank-page paralysis at the start of every brief
  • Inconsistent tone across writers and freelancers
  • SEO treated as an afterthought instead of an input
  • No repeatable process, so quality swings article to article

Who it is for

  • Content marketers running a regular publishing cadence
  • Solo founders producing their own content
  • Agencies standardising output across accounts
  • SEO teams that need volume without dropping quality

Benefits

  • Repeatable quality — the same steps every time
  • SEO built into the outline rather than retrofitted
  • The human review point is explicit, not assumed
  • Every step names its tool, so the process survives staff turnover

Limitations

  • Factual claims still need a human check — the draft will state things confidently that are wrong
  • Competitive keyword research needs a paid SEO tool to be worth much
  • Does not replace subject-matter expertise on technical topics
  • Image generation rights vary by provider — check before publishing

Success metrics

  • Time from brief to publish-ready draft
  • Organic sessions per article at 90 days
  • Editor changes required per draft
  • Cost per published article

Implementation guide

Run steps 1–3 manually for two articles first, to calibrate the prompts against your brand voice. Once the outline step produces something your editor accepts without restructuring, wire steps 1, 2 and 4 into your automation platform and keep step 6 manual.

Flow diagram

Every step in order, colour-coded by how much of it runs without a human.

Scroll to pan · use the controls to zoom, expand or export.

Steps

8 steps from trigger to result.

1

Topic and audience research

Perplexity GPT-4o 15 min Semi-automated
Establish what the audience already knows, what the top-ranking pages cover, and where the gap is. This is the step people skip, and it is why most AI content reads generic.
InputTopic, target audience, 3 competitor URLs
OutputResearch brief with angle and gap analysis
Prompt
You are a content strategist. Research the topic "{{topic}}" for an audience of {{audience}}.

Review these competitor pages: {{competitor_urls}}

Produce:
1. What every competitor covers (the table stakes)
2. What NONE of them cover (the gap)
3. Three angles that would differentiate our piece
4. The five questions a reader most wants answered

Be specific. Do not pad. If the gap is small, say so.
Expected result: A one-page brief naming the angle no competitor has covered.
  • Feed it real competitor URLs, not just the topic — the gap analysis is only as good as the input
  • Ask for the questions readers ask, not the questions competitors answer
  • Verify any statistic it surfaces against the original source before it reaches the draft
2

Keyword and SERP analysis

Surfer SEO 10 min Semi-automated
Pick one primary keyword and a handful of secondaries with real volume, and note the search intent the SERP is actually rewarding.
InputTopic, target market
OutputPrimary keyword, 5-8 secondaries, intent classification
Expected result: A keyword set with volume, difficulty and the intent the current top three satisfy.
  • Match the intent the SERP shows, not the intent you wish it had
  • If the top results are all product pages, an article will not rank — change the format
3

Outline generation

Claude Sonnet 8 min Automated
Turn the brief and keyword set into a section-by-section outline, with the argument each section makes.
InputResearch brief + keyword set
OutputH2/H3 outline with per-section intent
Prompt
Using this research brief and keyword set, write a section outline for an article of about {{word_count}} words.

Brief: {{brief}}
Primary keyword: {{primary_keyword}}
Secondary keywords: {{secondary_keywords}}

For each section give:
- The H2
- The single point that section makes
- Which keyword it naturally serves
- Roughly how many words

The outline must build an argument, not list topics. No section should be skippable.
Expected result: An outline your editor would approve without restructuring.
4

Draft generation

Claude Sonnet 12 min Automated
Generate the draft section by section against the outline. Section-by-section beats one long generation: tone holds, and you can regenerate one part without losing the rest.
InputApproved outline + brand voice guide
OutputFull draft
Prompt
Write the section "{{section_heading}}" of an article titled "{{title}}".

The point this section makes: {{section_point}}
Brand voice: {{voice_guide}}
Target length: {{words}} words

Rules:
- Lead with the point, then support it
- Concrete examples over abstract claims
- No filler phrases ("in today's fast-paced world", "it's important to note")
- Do not repeat what earlier sections covered: {{previous_sections}}
Expected result: A complete draft in your brand voice, ready for a human edit.
  • Generate one section per call and pass the previous headings in — it stops the model repeating itself
  • Keep the voice guide short and specific; a long one dilutes
5

SEO optimisation pass

Surfer SEO 10 min Semi-automated
Check keyword placement, heading structure, internal links, meta title and description, and image alt text.
InputDraft + keyword set
OutputOptimised draft + meta fields
Expected result: A draft scoring in the target range, with meta fields written.
  • Do not stuff to hit a content score — a readable article that scores 78 beats an unreadable one that scores 95
6

Human review and fact check

30 min Manual Human review
A person reads the whole piece, verifies every factual claim and statistic against its source, and makes the edit. This step is not optional.
InputOptimised draft
OutputApproved article
Expected result: An article a named person is willing to put their name on.
  • Check every number and every attributed quote
  • Read it aloud — that catches the AI cadence faster than reading silently
  • Skipping this step is how organisations publish confident falsehoods under their own brand
7

Publish

WordPress 5 min Automated
Push to the CMS with meta fields, hero image, canonical URL, internal links and schema.
InputApproved article + assets
OutputLive URL
Expected result: Article live with correct metadata and internal links in place.
8

Measure and iterate

Google Search Console 5 min Automated
Schedule a 30/60/90-day check on impressions, position and conversions, and feed what you learn back into the outline prompt.
InputLive URL
OutputPerformance report
Expected result: A scheduled report and a decision: leave, refresh or expand.

AI tools

Each tool records the role it plays and whether it is required.

Integrations

Services this workflow connects to, and how each authenticates.

Prompt library

Every prompt in the workflow, in chain order. Swap the highlighted placeholders for your own values.

1. Topic and audience research · User prompt · GPT-4o
You are a content strategist. Research the topic "{{topic}}" for an audience of {{audience}}.

Review these competitor pages: {{competitor_urls}}

Produce:
1. What every competitor covers (the table stakes)
2. What NONE of them cover (the gap)
3. Three angles that would differentiate our piece
4. The five questions a reader most wants answered

Be specific. Do not pad. If the gap is small, say so.
topic audience competitor_urls
3. Outline generation · User prompt · Claude Sonnet
Using this research brief and keyword set, write a section outline for an article of about {{word_count}} words.

Brief: {{brief}}
Primary keyword: {{primary_keyword}}
Secondary keywords: {{secondary_keywords}}

For each section give:
- The H2
- The single point that section makes
- Which keyword it naturally serves
- Roughly how many words

The outline must build an argument, not list topics. No section should be skippable.
word_count brief primary_keyword secondary_keywords
4. Draft generation · User prompt · Claude Sonnet
Write the section "{{section_heading}}" of an article titled "{{title}}".

The point this section makes: {{section_point}}
Brand voice: {{voice_guide}}
Target length: {{words}} words

Rules:
- Lead with the point, then support it
- Concrete examples over abstract claims
- No filler phrases ("in today's fast-paced world", "it's important to note")
- Do not repeat what earlier sections covered: {{previous_sections}}
section_heading title section_point voice_guide words previous_sections

Prerequisites

Have these ready before you start.

OpenAI or Anthropic API key
Any current frontier model works; the prompts are not provider-specific.
SEO research tool Optional
Surfer, Semrush or Ahrefs — needed for real keyword volume.
CMS with an API
WordPress, Webflow or similar for the publish step.
Image generation credits Optional
Only if you generate the hero image rather than sourcing it.

Use cases

Weekly blog cadence

Publish two researched articles a week with one part-time editor instead of a full content team.

Consistent cadence without new headcount

Agency client delivery

Standardise output quality across accounts so every client gets the same process.

Predictable quality across writers

Product launch content

Spin up a cluster of supporting articles around a launch in days rather than weeks.

Launch coverage ready on time

Content refresh

Run steps 1, 2 and 5 against existing articles that have slipped in the rankings.

Recovered organic traffic

Metrics

Author estimates — treat them as a starting point and measure your own runs.

Execution time
~1h 35m
Estimated AI cost
$2.4
Token usage
48K
Success rate
92%
Automated
70%
Human review
30%

Accepts

TextWebsite URL

Produces

BlogImageSummary

FAQ

How long does this actually take end to end?
About 95 minutes of wall-clock time per article once the prompts are tuned, of which roughly 30 is human review. The first two or three take longer while you calibrate the voice guide.
Was this helpful?
Can I automate the human review step?
You can automate the checks — readability, keyword placement, broken links — but not the judgement. Someone has to verify the factual claims, because the draft will state wrong things fluently.
Was this helpful?
Which model should I use for drafting?
Any current frontier model produces a usable draft. The bigger lever is outline quality and the voice guide, not the model choice.
Was this helpful?
Do I need a paid SEO tool?
For step 2, effectively yes — free keyword data is too coarse to pick between close options. The rest of the workflow runs without one.
Was this helpful?

Reviews

No reviews yet — be the first to share how this worked for you.

Write a review

You will be asked to sign in.