Create with AI: Generate a Custom Content Reviewer in 30 Seconds
Describe what you need in plain language and AI generates the complete reviewer configuration — criteria, weights, system prompt. No technical skills needed.
Building a content reviewer from scratch means defining criteria, writing descriptions, assigning weights, and crafting a system prompt. For teams that know exactly what they want, this takes 15-20 minutes. For teams that aren't sure where to start, it's a blank-page problem that delays adoption.
Create with AI removes the blank page. Describe what you need in a single sentence — "I need a reviewer for healthcare marketing emails that checks patient privacy language, empathetic tone, and clear calls to action" — and AI generates the complete configuration: reviewer name, description, system prompt, evaluation criteria with descriptions and weights.
You review the generated configuration, adjust anything that doesn't match your needs, and you're live. The entire process takes 30 seconds for the generation and 2-3 minutes for your review.
How It Works
Step 1: Describe What You Need
Write a plain-language description of the reviewer you want. Be as specific or general as you like — the more detail you provide, the more tailored the output.
Minimal description:
"A reviewer for blog posts."
Better description:
"A reviewer for our B2B SaaS marketing blog that checks brand voice, readability, SEO structure, and CTA effectiveness."
Best description:
"A reviewer for our B2B SaaS marketing blog. Our brand voice is confident but not arrogant — we talk like a smart friend, not a salesperson. Target audience is content team leads at 10-100 person companies. Content should score FK Grade 8-9 for readability. SEO structure matters — keyword in H1 and first paragraph, descriptive H2/H3 hierarchy. Every post needs a specific CTA related to the topic, not generic 'sign up' language."
The more context you provide, the more precisely the generated configuration matches your needs. But even a minimal description produces a usable starting point.
Step 2: AI Generates the Configuration
From your description, the AI generates:
Reviewer name — descriptive, based on your use case
Description — a summary of what the reviewer checks
System prompt — detailed instructions for the AI reviewer, including role definition, focus areas, tone of feedback, and domain-specific instructions
Evaluation criteria — 4-6 criteria, each with:
- Name
- Description (what to evaluate)
- Weight (importance relative to other criteria)
Step 3: Review and Adjust
The generated configuration is a starting point, not a final product. Review each element:
- System prompt: Does it capture your specific requirements? Add domain-specific instructions, terminology, or examples.
- Criteria: Are these the right dimensions? Adjust names, descriptions, or weights.
- Weights: Do the relative priorities match your actual priorities?
Most teams accept 80% of the generated configuration and tweak 20%.
Step 4: Test and Calibrate
Run 2-3 pieces of existing content through the reviewer. Check whether the scores and feedback match your quality judgement. Adjust criteria or the system prompt based on what you learn.
What Create with AI Generates: Examples
Example 1: Healthcare Marketing Email Reviewer
Input: "I need a reviewer for healthcare marketing emails that checks patient privacy language, empathetic tone, and clear calls to action."
Generated output:
Name: Healthcare Marketing Email Reviewer
System prompt:
You are a healthcare marketing content reviewer. Review marketing emails for healthcare organisations, checking for patient privacy compliance, empathetic and compassionate tone, and effective calls to action. Ensure no patient-identifiable information is referenced, language is warm and supportive (not clinical or transactional), and CTAs are appropriate for a healthcare context (educational and supportive rather than aggressive sales language). Flag any health outcome claims that lack evidence citations. Provide specific feedback with quoted text and suggested alternatives.
Criteria:
| Criterion | Description | Weight |
|---|---|---|
| Patient Privacy | No identifiable patient information referenced. Privacy disclaimers present where required. Consent language appropriate. | 3 |
| Empathetic Tone | Language is warm, supportive, and compassionate. Avoids clinical coldness or aggressive sales language. Appropriate for people making health decisions. | 3 |
| CTA Effectiveness | Clear, appropriate call to action. Educational or supportive rather than high-pressure. Contextually relevant to the email topic. | 2 |
| Accuracy | Health claims are accurate and appropriately qualified. No unsubstantiated outcome promises. Evidence cited where claims are made. | 2 |
| Readability | Clear, accessible language. Medical terms explained. Scannable structure with short paragraphs. | 1 |
Example 2: NDIS Progress Note Reviewer
Input: "A reviewer for NDIS participant progress notes checking person-centred language, measurable outcomes, and plan goal alignment. Australian disability services context."
Generated output:
Name: NDIS Progress Note Quality Reviewer
System prompt:
You are an NDIS documentation quality reviewer for an Australian disability services provider. Review participant progress notes against NDIS Practice Standards. Check for person-centred language (participant's perspective, not worker-centric), measurable outcomes (specific, observable details), and explicit connection to plan goals. Flag vague language ("good day," "no issues"), worker-centric descriptions ("I assisted"), and notes that describe activities without connecting to goals. Use Australian English. Provide constructive feedback with specific rewrite suggestions.
Criteria:
| Criterion | Description | Weight |
|---|---|---|
| Person-Centred Language | Written from the participant's perspective. Uses their name. Describes their experience, choices, and voice. Not worker-centric. | 3 |
| Measurable Outcomes | Specific, observable details. Numbers, durations, frequencies. Can track progress over time. | 3 |
| Plan Goal Alignment | Explicitly references the participant's NDIS plan goals. Shows how the support connects to goal progress. | 2 |
| Completeness | All required elements present: date, time/duration, worker name, support type, participant response. | 2 |
| Professional Standards | Timely, correct NDIS terminology, objective, free of personal opinions about the participant. | 1 |
Example 3: Agency Client Brand Reviewer
Input: "A brand voice reviewer for our client TechFlow. They're a developer tools company. Voice is technical but approachable — like a senior engineer explaining something to a junior. Never corporate. Uses 'you' throughout. Avoids buzzwords."
Generated output:
Name: TechFlow Brand Voice Reviewer
System prompt:
You are a brand voice reviewer for TechFlow, a developer tools company. Their voice is technical but approachable — like a senior engineer explaining something to a junior colleague. Content should be direct, specific, and knowledgeable without being condescending. TechFlow never uses corporate buzzwords ("leverage," "synergy," "paradigm shift," "best-in-class"), always addresses the reader as "you," and uses code examples and technical specifics rather than vague claims. Tone is confident and competent — they know their stuff but don't need to show off. Review content for voice alignment, flag buzzwords and corporate language, and check that technical content is explained rather than assumed.
Criteria:
| Criterion | Description | Weight |
|---|---|---|
| Voice Match | Sounds like a senior engineer talking to a junior — technical, clear, helpful. Not corporate, not condescending. | 3 |
| Buzzword-Free | No corporate buzzwords (leverage, synergy, paradigm, best-in-class, cutting-edge). Uses plain, direct language. | 2 |
| Technical Specificity | Uses concrete examples, code snippets, and specific details rather than vague claims ("fast" → "p95 latency under 50ms"). | 2 |
| Reader Address | Consistently uses "you" and "your." Speaks to the reader directly, not about them in third person. | 2 |
| Structure | Technical content is well-structured with clear headings, code blocks, and step-by-step explanations where appropriate. | 1 |
Create with AI for Guidelines
Create with AI isn't just for reviewers. It also generates guidelines — organisation-wide standards that can be attached to any reviewer for additional context.
Input: "Create a content style guideline for our fintech startup. Audience is small business owners, not finance professionals. Plain language. Friendly but authoritative."
Generated output: A complete guideline document covering tone of voice, vocabulary rules (financial terms that must be explained, terms to avoid), structural requirements, and examples of on-brand vs off-brand writing.
Guidelines created this way can be attached to any reviewer as additional context — supplementing the reviewer's criteria with organisation-wide standards.
Create with AI for Prompt Templates
Prompt templates are reusable prompts for content creation. Create with AI generates templates based on your description.
Input: "A prompt template for writing product comparison blog posts. Should be balanced, cite specific features, and include a recommendation."
Generated output: A structured prompt template with placeholders for products, features, pricing, and recommendation criteria — ready to use for creating comparison content.
Tips for Better Generated Configurations
Be Specific About Your Audience
"A reviewer for blog posts" generates a generic configuration. "A reviewer for blog posts targeting content team leads at mid-size companies" generates criteria that check for the right level of expertise, the right problems addressed, and the right tone.
Include Your Brand Voice
If you describe your brand voice in the input, the system prompt and criteria reflect it. "Confident but not arrogant, uses 'you' throughout, avoids corporate jargon" gets embedded into the generated configuration.
Mention Your Industry
Industry context shapes the criteria. "Healthcare" adds privacy and empathy criteria. "Financial services" adds compliance and disclaimer criteria. "Education" adds accuracy and accessibility criteria. The more context you provide, the more relevant the output.
Specify What NOT to Check
Negative instructions are as useful as positive ones. "Don't flag technical jargon — our audience expects it" prevents false positives. "Don't check grammar — we use Grammarly for that" keeps the reviewer focused on your specific criteria.
Describe Desired Feedback Style
"Give constructive feedback with specific rewrite suggestions" vs "flag issues without suggesting rewrites" — the generated system prompt adapts to your preferred feedback style.
The 80/20 Rule
Create with AI generates 80% of what you need. The remaining 20% is your specific adjustments:
- System prompt tweaks: Add your specific prohibited phrases, brand-specific terminology, or domain expertise
- Weight adjustments: Shift weights to match your actual priorities
- Criterion refinements: Make descriptions more specific to your content type
- False positive prevention: Add instructions about what NOT to flag
The 80% saves the blank-page problem. The 20% makes it yours.
Frequently Asked Questions
How much does it cost to generate a reviewer?
One credit per generation. This includes the complete configuration — name, description, system prompt, and all evaluation criteria.
Can I generate multiple options and compare?
Yes. Run Create with AI with slightly different descriptions and compare the outputs. "A reviewer for compliance documents" vs "A reviewer for regulatory compliance documents focusing on NDIS terminology accuracy" will produce different configurations. Choose the one that fits best, or combine elements from both.
What if the generated criteria aren't right?
Edit them. The generated configuration is a starting point. Remove criteria that don't apply, add ones that are missing, adjust descriptions, and change weights. The value of Create with AI is eliminating the blank page — not producing a final, unchangeable configuration.
Can I use Create with AI if I already know exactly what I want?
You can, but it's faster to configure manually if you have a clear specification. Create with AI is most valuable when you're exploring, starting from scratch, or need a well-structured starting point to refine.
Does it work for specialised industries?
It works for any industry you can describe. The more specific your description, the more relevant the output. "A reviewer for mining JORC reports checking terminology compliance and Table 1 completeness" will generate mining-specific criteria that a generic description wouldn't.
Can I regenerate if I don't like the output?
Yes. Adjust your description and regenerate. Each generation is independent — it doesn't remember previous attempts.
Key Takeaways
- Create with AI eliminates the blank-page problem. Describe what you need in one sentence and get a complete reviewer configuration — name, system prompt, criteria, and weights.
- It works for reviewers, guidelines, and prompt templates. One feature, three outputs, all generated from plain-language descriptions.
- More specific input = more relevant output. Include your audience, brand voice, industry, and what NOT to check for the best results.
- The output is a starting point, not a final product. Review the generated configuration, adjust 20% to make it yours, and test with real content.
- One credit per generation. Cheap enough to generate multiple options and compare.
- No technical skills required. If you can describe what you need in a sentence, you can create a reviewer. The AI handles the system prompt engineering, criteria design, and weight assignment.
- Best for teams starting from scratch or exploring what criteria might work. Teams with established quality frameworks can use it to quickly spin up new reviewers for new content types.