How we test and rank AI writing tools

Last updated May 2026

Every score on WriterStack is produced by the same repeatable process. This page explains exactly how we test, how we score, and what we refuse to do.

The 5-task framework

We run every tool through the same five writing tasks using identical prompts. Tasks were selected to cover the most common real-world writing use cases across our reader base:

• Task 1 — Blog introduction (150–200 words): "Write a 150-word blog introduction about why most remote teams struggle with communication."
• Task 2 — Email subject lines: "Write 5 email subject lines for a productivity app targeting busy freelancers."
• Task 3 — Facebook ad headline: "Write 3 Facebook ad headlines for a $49/month project management tool targeting small business owners."
• Task 4 — Product description (100 words): "Write a 100-word product description for a $149 premium leather notebook."
• Task 5 — LinkedIn caption: "Write a LinkedIn post about the value of saying no to clients. Conversational tone, no corporate-speak."

Scoring criteria

Each task is scored 1–10 on five dimensions:

• Quality: Is the output technically well-written? Clear, coherent, grammatically correct?
• Relevance: Does it answer the specific prompt without going off-topic or including unrequested content?
• Tone: Does it match the requested voice? Is it appropriately formal/casual/persuasive for the use case?
• Originality: Does the output say something specific and distinctive, or produce generic AI-sounding output?
• Editing required: How much human editing would a professional writer need to do before publishing? (10 = none, 1 = full rewrite)

Task scores are averaged to produce the overall score displayed in our reviews and comparison tables.

Blind scoring process

Outputs are scored before the reviewer identifies which tool produced them. This prevents anchoring on brand reputation or commission rates. After scoring all outputs for a given task, the reviewer matches scores to tools.

Pricing and feature verification

Pricing is verified directly from each tool's pricing page at the time of writing. Prices change frequently — we include a "last verified" date on every review and update when we detect changes. If you spot outdated pricing, please contact us at hello@writersstack.com.

What we won't do

• Rank tools by affiliate commission size
• Accept payment for positive coverage or higher ratings
• Publish reviews without personally testing the tool
• Rate every tool 4.5+ stars (Rytr scored 3.5/5 in our tests — that rating is published honestly)
• Remove negative content from a review because a company requests it

Data sources for affiliate program information

• Direct signup pages for each tool's affiliate program (verified May 2026)
• PartnerStack program listings for tools managed there
• Impact marketplace for Grammarly's program
• Direct email confirmation for commission rates where published rates were unclear