Quick Answer
Solo creators are publishing 5x more with this stack; the bottleneck moved from writing to research quality.
Step-by-Step Workflow
7 steps · 2 hours to set up · 1 hr per post ongoing
- 1
Pick the topic from a research-backed list
Open your idea backlog (we keep ours in Notion). Pick a topic with: (a) confirmed search demand (Ahrefs or Google Trends), (b) a personal angle or original take, (c) a definable reader outcome. If you can't write one sentence describing what the reader will know or do after reading, don't write the post.
5 minOutput: Topic + reader outcome sentence - 2
Run a Perplexity deep research pass
Use Perplexity Pro to gather 8-12 sources on the topic. Prompt: 'Research [topic] for [audience]. Surface contrarian takes, recent data within last 12 months, and citations from at least 3 different domain types (academic, practitioner blogs, news).'
10 minOutput: Research doc with sources and citationsTools: PerplexityTip: Always include 'recent data within 12 months' in the prompt - otherwise Perplexity defaults to older general knowledge.
- 3
Generate an outline with Claude
Paste research into Claude and ask for a 5-7 section outline. Constraint Claude: 'Each section header must be a specific reader question, not a generic topic. Sections must build toward the reader outcome of [outcome from step 1].'
5 minOutput: Approved outline with section H2sTools: Claude - 4
Draft section by section
Don't ask Claude to write the whole post. Ask for one section at a time, with the constraint: '300 words max, conversational, include one specific example from the research doc.' Read each section as it generates and fix the voice as you go.
25-30 minOutput: Full first draft, 1,200-1,800 wordsTools: ClaudeTip: If a section sounds AI-generated, paste it back with: 'Rewrite this in a more conversational voice, cut transition phrases like therefore and however, and add one specific number or example.'
- 5
Personal edit pass
Add your POV. Cut filler. Replace generic claims with specific stories or numbers. This is the step that most AI content skips - and it's why AI content gets flagged. Aim for 5-10 small edits per 500 words of draft.
10-15 minOutput: Final draft with human voice - 6
Run Hemingway + Frase final checks
Paste into Hemingway to catch dense sentences. Paste into Frase to verify you've covered the topics other top-ranking pages cover. Add anything missing. Don't chase a perfect Frase score - 70-80% coverage is the sweet spot.
5 minTools: Frase - 7
Publish + log
Publish in your CMS. Log the topic, time spent, and final word count in Notion. This data tells you which topic types are fastest to write and which deserve more research time per post.
5 min
Common Pitfalls
- Asking Claude for a 'complete blog post in one shot' - the output reads generic and you'll spend longer rewriting it than if you'd drafted section by section.
- Skipping research and going straight to Claude. The draft will be confident-sounding but factually thin; LLMs hallucinate specific stats unless you feed them sourced research.
- Editing as you draft. Doubles your time. Draft first, edit second.
- Letting AI write your intro. Intros are the most-read part of a post and the most-recognized as AI-generated. Always write the first 100 words yourself.
Frequently Asked Questions
Why use both Perplexity and Claude instead of just one?
Won't Google penalize AI-written content?
How much does this stack cost?
Can I use ChatGPT instead of Claude for the drafting step?
How we built this workflow
ToolJunction's editorial team tests each workflow with real accounts and real budgets before publishing. Cost figures reflect public pricing pages as of May 2026. Reply rates, time estimates, and outcome metrics come from our own runs or vetted operator interviews. We update this page when a tool's pricing changes or a step stops working.