A popular tool is not automatically the right tool for your task. Compare candidates on work you understand, using the same inputs and the same acceptance criteria.
Build a small comparison set
Select three representative tasks: one routine, one difficult and one with missing information. For example, compare a product-description draft, a policy summary and an enquiry with an incomplete brief. Remove confidential details that are unnecessary for the comparison.
Score results, not fluent writing
Use a simple scorecard: factual errors, important omissions, usable format, appropriate tone and minutes spent fixing the output. Ask a reviewer who knows the subject to evaluate results. Record failures even when a different example was impressive.
Include access and operating costs
Check the current official plan, actual limits, export options, team access and data settings. A tool that looks inexpensive may require additional editing or a separate integration. Record the date of your comparison because features and terms change.
ChatGPT, Claude and other candidates
Use the same task brief when comparing assistants. Test the relevant language and document types yourself. Avoid treating one good answer as a general ranking. Choose the smallest tool set that meets your checks; keep the sources and test inputs so you can repeat the comparison later.
Keep this checklist
- Same tasks for each candidate
- Subject review completed
- Total editing time counted
- Plan and data settings checked
This is an original practical guide. Examples are illustrative; tool interfaces and available features may change.