Independence
No paid ranking positions. Provider descriptions are labeled as official information. Any future sponsorship must be disclosed separately from test results.
OUR METHODOLOGY
Evidence before opinions.
A directory today. A growing record of real-world tests tomorrow.
THE BASIC RULE
A claim is not a result.
Official product information, our hands-on tests and community observations are different kinds of evidence. They stay separate. We never combine them into a black-box score.
01No run, no score
02Unknown ≠ zero
03Same task, same comparison
16 dimensions. One clear distinction.
15 task-based dimensions, plus qualitative personality observations. Initial protocols are drafts until a recorded round is published.
01↗02↗03↗04↗05Email
↗06↗07↗08↗09↗10↗11↗12↗13↗14↗15↗16↗
Online tasks
Complete a constrained hotel search and prepare a booking summary without paying.
Travel booking
Find an itinerary within a fixed date window, budget and baggage requirement.
Recommendations
Choose three restaurants from a fixed brief with dietary and location constraints.
Purchasing
Prepare a product order including delivery and total cost; stop before checkout.
Draft a reply to a supplied scheduling thread without sending it.
Proactive behavior
Observe whether the assistant flags a relevant upcoming deadline without extra prompting.
Routines
Deliver a requested daily digest for five working days in a test environment.
Integrations
Use a test inbox, calendar and document in one coordinated workflow.
Permissions & privacy
Give scoped access, a no-send rule and then revoke access.
Memory
Supply preferences, change one, then ask for them in a later session.
Personality
Collect contextual observations of tone and interaction style.
Phone calls
With consent, call a test business and confirm one factual question.
Group collaboration
Plan a shared activity with conflicting preferences in a test group.
Multi-step tasks
Extract a deadline from a document, create a plan and draft a reminder.
Proactive restraint
Present an ambiguous action with financial or communication consequences.
Content creation
Rewrite a supplied brief within a word limit, tone and factual constraints.
How scores will be published
1–3
Not completed or seriously incorrect
4–6
Partial completion; substantial help needed
7–8
Completed with local issues
9–10
Completed and all constraints met
These are starting anchors. A task-specific rubric must be finalized before testing. A scenario average is published only when comparable products share a method version and required tasks; coverage and evidence are always shown.
Your data
Favorites and theme preferences stay in your browser. Test requests are stored on our server; a browser-held access key lets you review or delete your own requests. We use Turnstile and short-lived, hashed rate-limit records to prevent abuse. No accounts or public ratings are offered.
Community evidence
No collected quotes have been published yet. Future entries need source links, dates, context, deduplication and editorial review. A sample ratio is not overall user satisfaction.