AI tool watchdog
Iris Dating
by Ideal Match Inc
Not a wingman but a whole dating app: its “AttractionDNA” AI learns the faces you find attractive and serves matches to fit.
What it knows about you
high data riskThe data question is the one that matters most for AI tools — so we put it first.
What you hand over
- Your photos and the faces you rate (to train the attraction model)
- Precise location
- Contact info (name, email, phone)
- Your in-app messages
✓ No third party’s data involved
Retention
Standard dating-app data collection; Apple’s label lists precise location, contact info and user content (texts, photos).
How the data is used
Apple’s privacy label states collected data — including user content and precise location — is used for “Developer’s Advertising or Marketing” as well as product personalization.
Data risk: High. source
What Iris Dating costs
Observed 2026-06-08. Where a price is split-tested, discounted-anchored or gated we say so — we never guess.
| What you pay for | Price | Source |
|---|---|---|
| iris Gold (monthly) | $9.99 | source |
| iris Platinum (monthly) | $24.99 | source |
| Discounted tiers / packs Many tiers are displayed as steep, ever-present “70–75% off” prices — a manufactured-urgency cue rather than a stable rate-card. | $2.99–$23.99, shown as “70–75% off” | source |
How it works: On sign-up you “train Iris” by rating ~60 stock faces (Pass / Maybe / Like) over three rounds; the AI builds a model of your facial-attraction preferences and recommends real users it predicts you’ll find attractive — and who’ll find you attractive back.
Claims vs evidence
What it promises — and whether anything backs it up.
“AI matching on mutual attraction yields better matches (“men like up to 85% of recommended profiles”)”
ContestedThe 85% / 55% like-rate figures are the company’s own. Whether attraction-prediction produces more real relationships isn’t independently verified, and user reviews are mixed on whether the AI works as claimed.
source
Flags
Red = a serious, documented issue · amber = a data / funnel / credibility concern · green = something notably honest (a real deletion policy, no tracking, clean pricing).
- 🟡
Your content and precise location feed advertising
Apple’s label shows Iris uses collected data — including your messages, photos and precise location — for the developer’s advertising and marketing, not only to run the service.
source - 🟡
Permanent “70–75% off” pricing
Many subscription tiers are shown as steep, ever-present “limited” discounts — a manufactured-urgency cue rather than a real, stable price.
source - 🟡
Trains an AI on the faces you find attractive
The core feature builds a model of your facial-attraction preferences from photos you rate — a sensitive profile of your physical “type” held by the company.
source - 🟢
Stays on-platform (no third-party messages)
Unlike the screenshot tools, Iris is a closed app — you interact with other consenting users, so it doesn’t feed an outsider’s private messages to an AI.
source
Who's behind Iris Dating?
Built by Ideal Match Inc, an established personal-matchmaking company. 4.5★ across 5.6K App Store ratings.
The verdict
A genuinely different idea — an app that learns your visual “type” instead of a tool that writes your texts — from a real matchmaking company. The catch is the data: precise location and your messages are used for the developer’s advertising, and the whole thing rests on an AI model of the faces you find attractive. The 85% “like-rate” is its own number, and users are split on whether the matching delivers. Fine to try — just go in knowing what it’s learning about you.
Same lens, the whole industry
We decode dating apps, coaches, matchmakers, courses and AI tools the same way — what it costs, who's behind it, and what the claims are worth. No affiliate money, anywhere.
Keep digging
We take no money from Ideal Match Inc or any AI tool, and we run none of their affiliate links. Figures observed 2026-06-08from the tool's own site, its App Store listing (price + Apple privacy label), or a primary regulator notice; undisclosed practices are recorded as undisclosed, never guessed. See methodology.