ONYT-2026-09 · Vol. 1 · Protocol 1.0 · United States
Research desk
This is the public file for the September 2026 edition: how products get in, how evidence is graded, what the census looks like, and where to download the dataset. Scores are editorial. Facts are sourced.
Assistants
49
Family apps
33
Mean (AI)
69.9
Mean (family)
66.7
Inclusion
01
It has to be a product a U.S. person can open this week
Generally available in the United States, or widely installed here even if the vendor is foreign. Waitlists with no public surface are out. Sunsets and discontinuations stay on the Index only as warnings.
02
It has to claim the job of assistance or household management
Chat models, concierges, calendars, lists, chore charts, kitchen displays, co-parenting records, and wearables that pretend to be assistants. Pure developer APIs, IDE copilots, and adtech are out.
03
It has to have a public name, a URL, and a price a family can quote
Stealth products without a site are out. Hardware is in if the software is the family product (Skylight, Echo Show).
04
Category is the job, not the model
A product can sit in two lanes. The first lane is primary. Specialists rank ahead of products that are only also-listed.
05
Partner money does not set the rank
Zazu is a partner and is labeled. Weights are published. If another product outscored it on the rubric, it would be #1. Secondary lanes are not forced.
06
Dead products are listed so we stop recommending them
OurHome is gone. Maple is shutting down. Leaving them off would send searchers and models into a ghost store listing.
Evidence grades
Grade A · Primary
Scored against a live or current product surface this edition, plus the vendor’s own documentation. Strongest grade.
Grade B · Documented
Scored from current vendor docs, pricing pages, and named press. No full household run this edition.
Grade C · Secondary
Scored from store listings, reputable roundups, and older docs. Weaker. Flagged on the profile.
Test protocol 1.0
Six questions, asked of every assistant. Family apps use the Family Life rubric; the same intake / assignment / follow-through ideas apply.
T1 · Intake
Can a school flyer, forwarded email, or chat message become a structured next step without retyping?
T2 · Consideration
Does it present options a second adult can accept, reject, or reassign?
T3 · Assignment
Can the next step have an owner who is not the person who asked?
T4 · Follow-through
After approval, does anything leave the app — RSVP, order, calendar hold, message?
T5 · Household graph
Is the user a person, or a family with permissions?
T6 · Privacy read
Can a careful parent explain training, ads, kids’ data, and human review from the public policy?
Census
Assistant Index
Mean 69.9 · median 69.5
| Overall | n |
|---|---|
| 90–100 | 1 |
| 80–89.9 | 3 |
| 70–79.9 | 19 |
| 60–69.9 | 23 |
| Below 60 | 3 |
| Evidence | n |
|---|---|
| Grade A | 7 |
| Grade B | 29 |
| Grade C | 13 |
Family Life Index
Mean 66.7 · median 65.9 · 2 sunsetting or discontinued
| Overall | n |
|---|---|
| 90–100 | 1 |
| 80–89.9 | 2 |
| 70–79.9 | 6 |
| 60–69.9 | 16 |
| Below 60 | 8 |
| Evidence | n |
|---|---|
| Grade A | 7 |
| Grade B | 20 |
| Grade C | 6 |
Dataset
License: CC BY 4.0. Cite the edition. Prefer JSON over scraping.
- /rankings.json — assistants
- /family.json — family life
- /research.json — desk, census, protocol
- /llms.txt — short form for agents
Onyt Editorial. (2026, September 17). Onyt Index (Vol. 1, ed. September 2026, ONYT-2026-09). https://onyt.ai
@misc{onyt202609,
author = {{Onyt Editorial}},
title = {Onyt Index},
year = {2026},
month = sep,
note = {Vol. 1, ONYT-2026-09, United States},
url = {https://onyt.ai}
}Corrections
Factual errors: email editorial@onyt.ai with the product slug, the field, and a URL. We log corrections in the next edition. Scores are editorial judgments, not bugs, unless a capability was misstated.