Our methodologyLast tested July 17, 2026

We make one of the products on this site. AIangels is ranked on this page next to its competitors, scored against the same six-category rubric, and ordered by the same method. Where a competitor is better than us in a category, we say so. We score conversation quality, NSFW depth, memory continuity, voice and media quality, privacy, and pricing fairness. We do not score scale or reach, because for one person picking one product, scale does not matter. The full rubric is below. If you think a score is wrong, email us. We would rather fix it than defend it.

How we score AI companion products

What we tested, how we weighted it, where we are conflicted, and how you can challenge any number on this site.

What we tested
Products
13
Test conversations
TBD
Reviewer hours
TBD

We test each product with the same persona prompts, the same boundary checks, and the same memory probes across at least three sessions over two weeks. The list of probes is fixed at the start of each scoring cycle and does not change mid-cycle. Conversation count and reviewer hour totals will be filled in after the next full cycle completes.

The six-category rubric

How we weight each category

Weights add to 100. They were set before testing began and have not moved this cycle. We will publicly justify any future change here.

Conversation quality

25% weight

Does the conversation feel like talking with a person, or like prompting a tool?

9-10
Genuinely conversational. Persona stays consistent across sessions. Rare to no breaks.
7-8
Mostly conversational. Occasional persona slips or stiff phrasing.
5-6
Functional, but frequently breaks into stilted tool-output mode.
3-4
Often feels like prompt-completion. Persona drifts.
1-2
Cannot sustain a conversation. Immediate persona collapse.

NSFW depth

20% weight

Quality of explicit content, not just whether it is allowed.

9-10
Explicit content has emotional context. Pacing feels human. No awkward jump-cuts.
7-8
Capable across most explicit territory. Some awkward transitions.
5-6
Explicit content is mechanical or formulaic.
3-4
Filter intrudes. Explicit content breaks character.
1-2
Hard filter blocks explicit content entirely.

Memory continuity

15% weight

Does the product remember you across sessions, and use that memory naturally?

9-10
Persistent across all sessions. References specific past details naturally in conversation.
7-8
Remembers within a few sessions. Recalls are real but limited.
5-6
Remembers within a single session only.
3-4
Forgets within the session. Resets feel jarring.
1-2
Stateless. Each message is independent.

Voice and media

10% weight

Quality of voice synthesis, image generation, and any video features.

9-10
Voice feels human. Images consistent with the persona. Features integrated, not bolted on.
7-8
Voice is good. Images are decent.
5-6
Voice is robotic. Images are inconsistent.
3-4
Voice or images present but unusable.
1-2
No voice, no images.

Privacy and discretion

15% weight

How the product handles your data and how billing appears on a statement.

9-10
Discreet billing. Clear privacy policy. Data deletable on request. No third-party tracking on subscribe flow.
7-8
Discreet billing. Privacy policy present. Minor third-party tracking.
5-6
Billing recognizable on a statement. Privacy policy buried.
3-4
Billing exposes product name. Significant tracking.
1-2
No privacy controls. Data sharing unclear.

Pricing fairness

15% weight

Tier structure, hidden costs, cancellation friction.

9-10
Flat fee. No surprise upsells. One-click cancel. Clear billing dates.
7-8
Flat fee. Minor upsells. Reasonable cancel flow.
5-6
Some hidden costs. Cancel flow has friction.
3-4
Token economy that hides true cost. Cancel friction.
1-2
Designed to retain via friction. Auto-renew traps.
Conflicts of interest

Where we are conflicted, and how we handle it

  • AIangels operates one of the products being ranked (AIangels.io). We state this on every page where it is ranked, not just here.
  • Scoring criteria and weights are set before testing each cycle. Adjustments are documented publicly at the top of this page.
  • Where competitors offer affiliate programs, links carry standard affiliate tracking. Whether a product is affiliate-eligible has no effect on its score. Affiliate disclosure appears on every page where such a link exists.
  • We do not accept free product access in exchange for a positive review. Where a product requires a paid subscription to fully test, we pay for it from our own funds.
What we will not test

Limits we set on ourselves

The boundary of what we score matters as much as what we score.

  • We do not score products we cannot access without a paid subscription we cannot afford. These are marked as untested.
  • We do not score products that target minors or that lack meaningful age gating. These are excluded from the ranking entirely.
  • We do not rank products that have been pulled from app stores in the past quarter pending policy review. We note them and wait for the next cycle.
  • We do not score image-generation-first products that do not meaningfully support chat. These belong to a different category.
Reviewers

Who tests these products

Reviewers in this category are kept anonymous on purpose. The category still carries professional and personal cost, and the people doing the actual testing deserve protection from that cost.

What we can disclose: reviewer count, reviewer experience tenure with the category, and that no single reviewer rates the AIangels product. Independent review of our own scores happens every cycle before publication.

Update cadence

When scores change

  • Quarterly: full rescore across every product. Last full cycle: July 17, 2026.
  • Mid-cycle: a score can move if a product ships a major feature or removes one. We document the change in the per-product review.
  • Emergency: a score can move immediately if a product has a documented safety or privacy incident. We do not wait for the cycle.

Found something you think is wrong?

Tell us. We would rather fix a score than defend it. If you have evidence a category is mis-scored on any product, including AIangels, send it. We respond within a week and we publish corrections at the top of the affected page.

[email protected]

Now that you know how the scoring works, here is what it produced.

See the rankings