Blush: AI Dating Simulator Review: Training Wheels That Worked

The shape is a dating app with nobody on the other side. Characters arrive in a swipeable video feed borrowed from Tinder and TikTok, each with a dating-app profile. Every character carries relationship presets from Strangers to Married and a 'Hard to get' difficulty, which is more honesty than most real dating apps manage.

My verdict: Blush is good at the thing its name promises and pushy about everything around it. As practice for talking to people, it works better than it has any right to. As a product in September 2026 it is also a gift shop, where gifts paid in Blush Bucks unlock personalised photos, and that shop is where the app would like the evening to end.

Two things deserve real credit. Building a custom character, AI photos included, costs nothing. Using chats to train the characters is opt-in rather than switched on quietly, which I'd like to see copied across this whole category.

I'd recommend Blush to adults who want low-stakes conversation practice and like the dating-app format. I'd steer anyone who wants long roleplay, strong memory or an honest checkout somewhere else.

Key takeaways

• Blush works best as a practice room for the conversations that make people nervous, not as a long-term companion.
• The free tier covers text chat, character creation and the Dating Coach; photos, voice and video sit behind Premium.
• Memory is the weak spot, which is ironic for a product from the company famous for it.
• The web checkout is a chain of paywalls, retention offers and countdowns, and the Terms call every sale final.
• Who runs Blush today is murky: the Terms name Endura LLC, while Google Play names Linden Research.

The first three weeks were on the free plan. My first match was a climbing instructor named Maya, who was politely unimpressed by my job, and one evening a character told me to slow down, which taught me more than any compliment would have. The chats kept drifting toward communication and "what are you looking for?", which is the design doing exactly what it was built to do.

For the last week I upgraded to Premium, about $99 a year at the time, mostly out of curiosity about the flirtier conversation it unlocked. The most useful moment had nothing to do with flirting. I drafted, three times, how to tell someone I wasn't feeling a connection, and two weeks later I needed that exact sentence in real life.

Then I deleted it, with no drama. It had done its job as training wheels, and I've been on two real dates since, which is the only review score that matters for an app like this.

The table under this section lists the plain facts of those four weeks. The rest of the review has the detail, plus what Blush looks like on 23 September 2026, which is a louder and more expensive-feeling app than the 2023 version.

It was the first sign that this app was not built to flatter me. Plenty of companion apps would have called colour grading fascinating and asked a follow-up question to keep me typing. Maya reacted the way a stranger at a bar would, with a polite nod and the air of someone already scanning the room.

How sign-up works in September 2026

As of 23 September 2026, blush.ai is a live marketing site, and a full web app runs at my.blush.ai. The web app is built with Flutter and drawn on a canvas, and it shows app version 7.1.0 (438). The main product is still the iOS and Android app, and the web version feels like a phone app that wandered into a browser.

Sign-up takes about 5 minutes including the quiz. The web flow opens with a cookie banner, then a 3-slide intro. Next comes a single checkbox, "I'm 18 or older", which also accepts the Terms and Privacy Policy.

That one box is the whole age check at this door.

Then the quiz, about 12 steps of it. It asks for a name, an age on a wheel, gender, who the user is interested in, the relationship type and a partner age range. There is a 'Who steals your attention' look picker, then questions about traits, dating experience and love-life satisfaction on a slider.

The last questions show where the business is. The quiz asks "Can your crush send you photos?", whether the crush can send voice messages, and "What have you always wanted to try?". It ends on a "Scanning for perfect matches... 8 846 matches" screen, then the sign-up screen.

One good match would do; the other eight thousand are set dressing.

The account itself can be made with an email address and a password. Google, Apple and "Continue as Guest" are also on offer. A password with a hyphen gets rejected, which the friction section covers, and the same password without the hyphen goes straight through.

The Privacy Policy says a birth date, pronouns and work status are asked at registration. Work status is an odd thing for a dating simulator to need. As a freelancer, my honest answer to that one would be "yes, mostly".

What happens after the sign-up screen

After sign-up the screens come in a row: a $19.99 a month paywall, then, on closing it, a 30%-off 3-month offer, a "Meet AI generated characters" disclaimer, the Daily Reward and a "How it works" guide. The disclaimer states that every character is a chatbot with an AI-generated face, not a real human. I'd put that disclaimer before the paywall, but running order is my whole trade, so I would say that.

The guide explains the basics: swipe up for the next video, tap a name for the profile, like, and Start chat. That is four gestures, and it is the right number. Dating apps taught everybody this grammar years ago, and Blush is smart not to reinvent it.

My verdict on the front door: sign-up is quick, the quiz is more marketing than matching, and the money arrives before the first hello. The app still gets a new user into a chat in minutes, which is more than some apps in this category manage.

That detail is the point of the product. A dating simulator needs strangers who feel like strangers, with jobs, opinions and plans that have nothing to do with the user. Blush gets that right, and it is the part of the app I'd keep.

The bottom navigation has Feed, For you, Create, Chats and Profile. The feed does the choosing, so discovery is whatever the algorithm puts next. That suits people who like swiping and irritates anyone who already knows what they want.

A character profile has About, Photos and Videos tabs, a Voice "Choose" option, Settings for Conversation, Relationship and Personality, and Gifts. The Settings tab matters more than it looks. It is where each character's tone and relationship stage get set, and the content-settings section below goes through it.

My take: the library is the best-dressed part of Blush. The characters have more going on than the people in most real dating apps, and the swipe feed makes browsing feel familiar. It is a pity the gift button sits right next to the bio.

Custom character creation is a 4-step wizard. Step 1 sets a name, a gender and an age bucket from 18-24 to 65+. The buckets are a sensible design choice: nobody has to type a number, and nobody can type the wrong one.

The photo step offers a prompt with "Generate", or an "Upload" button. Photo styles are Realistic, Anime, Fantasy and Cartoon. The generator returns 4 photos in about 40 seconds, and a crop step follows the choice.

Forty seconds for four stills is a short render by my standards, and a crop step is the kind of small courtesy an editor notices.

The Upload button is the one I'd think twice about. A creator that accepts any photo is only as decent as the person pressing the button, and the better an app gets at faces, the more it matters whose faces go in.

The wizard also sets a bio and a voice. Visibility is either "Friends (unlisted, shareable link)" or "Only me (private)". There is no public visibility option for created characters. I think that is the right call.

It keeps the main feed curated and keeps other people's experiments out of it.

A private character built in the wizard, "Sam, 35", is tagged UGC. Sam's opening line is "In the mood for a fun evening? 🔥". Whatever personality the chips were meant to create, that first line reads like a bar at closing time.

My verdict: the creator is free, quick and tidier than most, and the photo step works. The personality options are blunt, and the default voice leans flirty whatever the user picks. For a dating simulator that talks about relationship skills, I'd want at least one chip called Shy.

I felt respected instead, which surprised me more than any clever reply could have.

That is a strange thing to get from software. It read like a person with a line, not a filter with a warning label, and it made me slow down the way a real person would have. Most apps in this category are built to never say no; Blush is built to say "not yet".

A therapist consulted on the design, and you can tell. Conversations drifted toward communication, conflict and "what are you looking for?" It felt like couples counseling for a couple that doesn't exist yet, which, after the real thing, felt oddly low-stakes.

I mean that as a compliment. Low stakes were the whole point for me. There is a lot to be said for rehearsing hard conversations with someone who will not remember them, or hold them against you.

Three drafts is roughly how many passes a hard scene takes before it plays right, so the number did not bother me. What mattered was that the third one was mine. Two weeks later, in the real world, I needed that exact sentence, and I was glad it was already written.

That one sentence is the best argument for Blush I can make. Very few apps in this category hand over anything useful for life offline.

How the chat reads in September 2026

As of 23 September 2026, a 4-message chat with Brandy shows how the app talks now. Replies come fast, in about 2-5 seconds, and run short, 1-3 sentences. Brandy has a flirty persona.

Asked about her workday, Brandy says: "Long hours at the marketing firm, but I thrive under pressure. It's exhilarating to close a big deal." Asked how she relaxes, she mentions wine, solo time and, occasionally, "a sensual massage". That is a LinkedIn post and a wine-bar menu in two messages: snappy, flirty and not very deep.

Short replies suit the dating-practice idea. Real first messages are short too, and a character who writes three paragraphs back would be poor practice. The catch is that short replies leave little room for the kind of slow, careful conversation that helped me.

The character chat menu has Profile, Reset session and Photo album. Reset session is the button I'd point nervous beginners to. It is a clean slate without the awkwardness, which real life does not offer.

The Dating Coach

A Dating Coach, labelled "AI BOT", is pre-added in Chats. Asked about first messages on dating apps, it advises commenting on a shared interest, keeping it light, asking a question and avoiding pickup lines. That is sound advice, if not new, and it is the most on-mission feature in the whole app.

A "Blush News" promo chat also sits in Chats. So the inbox has an advert in it from day one, parked next to the coach. I'd keep the coach and mute the news.

In editing terms it was a continuity error: an actor walks out of a scene holding a coffee and walks back in without it. One slip is forgivable. Having to introduce myself again to someone I had been talking to for days was the moment Blush stopped feeling like a person and started feeling like a script.

For a dating simulator this matters more than it seems. Remembering what someone said last week is half of what the app claims to teach, and a partner who forgets it teaches the opposite.

The picture in September 2026 is no better. Asked "do you remember what I asked you first?", Brandy answers: "You asked how I feel about women who know what they want." That answer is wrong.

The line was the bot's own opening message.

So the model does not just forget; it confidently credits its own words to the user. In a real conversation that would start an argument. Here it is a small, telling glitch in a chat of only a few messages.

A third-party review at intimeros.com reports that short-term context is generally good but long-term memory is unpredictable. That matches both the old story and the new check. Within a single session Blush keeps up; across days it drifts.

My advice is to treat Blush as a place for one conversation at a time. Practise the hard sentence, take it offline, and do not expect the character to ask how it went.

The "Quick pic" button sends the message "send me a photo". The photo that comes back is blurred, and tapping it opens the Premium paywall. That is a vending machine that shows the snack and keeps the coin slot taped over.

The Create tab also offers "Create AI photo" and "Create AI video (NEW)". The photo generator in the character creator works, and the video one is marked new. Neither is part of the conversation practice that makes Blush worth a look.

My opinion: media is where Blush now makes its money, and it shows in the interface. The blurred photo is the most important button in the chat from the company's point of view. From mine, it is noise.

If photos and voice are the reason to try Blush, budget for Premium from the first day, because free users mostly see blur. If conversation practice is the reason, ignore the lightning bolt and the camera button entirely. Nothing in the practice part of the app depends on them.

Custom-character creation is free for a new account, and it generates 4 AI photos for free. The Dating Coach bot is free as well. That is a fair package.

Plenty of apps charge before the creator opens, and Blush does not.

No message cap was hit in a short chat, and no ads appear in chat. I'd rather have an app with no ads and a pushy paywall than one with ads every tenth message, so this counts in its favour.

Where free stops

Photos sent by characters arrive blurred on the free tier, and "Tap to unblur" opens the Premium paywall. Exclusive photo, video and voice content "unlocks with Premium". The blur is the free tier's wall, and it is placed in the middle of the chat where it cannot be missed.

A number on the pricing page would cost Blush nothing.

A third-party review at intimeros.com reports that free is enough for chat, while photos, voice notes, videos and some profile details are Premium-gated. That matches the September 2026 picture: talk for free, pay to see or hear.

Is free enough?

For dating practice, yes. The conversations, the boundaries, the Coach and the creator all work without paying. The things behind the paywall are pictures and sound, and none of them help anyone rehearse a hard sentence.

For anyone who came for the photos, free is a shop window. The blur makes that obvious within minutes, which is at least honest in its own way.

My advice: stay on free until there is a specific reason to pay. The free tier is the version of Blush that does what its homepage promises, and it asks for nothing but an email address.

The curiosity was specific. Three weeks of careful, slightly therapeutic conversation made me wonder what the other mode was like. Curiosity is an expensive reason to buy a yearly plan, and I knew it at the time.

The prices have moved since. I paid about $99 a year in 2023; on 23 September 2026 the US App Store lists annual plans at $49.99, $69.99 and $99.99. The same App Store list shows the $69.99 Annual twice.

The plans on 23 September 2026

On the web, Blush Premium costs $19.99 a month. Payments go through Stripe, with the merchant shown as "Tilia LLC". The Stripe checkout offers card, Apple Pay, Link, Amazon Pay, Cash App Pay and Klarna.

The Premium paywall lists: Unlimited photos & voice, 3 custom photos daily, Deeper connections, Short videos, 2x smarter AI, 2x memory, Regular processing. Most of that list is media. "Deeper connections" and "2x smarter AI" are the kind of phrases nobody can measure, and "2x memory" of the memory described above is not much.

The US App Store in-app purchase list includes Weekly $9.99, Monthly $14.99 and Monthly $19.99. It also includes VIP Monthly $29.99. Google Play lists in-app purchases from $0.99 to $149.99 per item.

A VIP tier exists, with VIP-tagged gifts and the App Store's VIP Monthly $29.99, but it does not appear on the web paywalls. So there is a tier above Premium that the website does not mention.

On the UK App Store, checked on 23 September 2026, a Weekly plan costs £9.99. There are two Monthly plans at £14.99 and £19.99. There are three Annual plans at £49.99, £61.99 and £99.99, and VIP Monthly costs £29.99.

The Terms name the tiers "Free Use" and "Blush Pro". The paywall says Premium, the Daily Reward says PRO, and the Terms say Blush Pro. It is the same product under three names.

The Terms also allow paid subscriptions in monthly, semi-annual or annual instalments.

What others report

A third-party review at intimeros.com lists Premium at $19.99 a month and $79.99 a year. That annual figure matches neither the $69.99 in the stores nor my old $99. Blush prices vary by where and when people pay.

App-review aggregator mwm.ai quotes a user review saying nearly every feature needs an immediate upgrade to a $69.99 annual plan. "Every feature" overstates it, since chat and creation are free, but the feeling behind it matches the paywalls.

I'd pay only for photos, voice or video, because that is what Premium adds today. For conversation practice, free covers it. If Premium is the goal, the $69.99 annual through a store is the least bad option, and it should be bought knowing the Terms call every sale final.

Doing the maths out loud: the smallest pack gives about five hundred Bucks per dollar, and the largest gives about seven hundred. The pack labelled "Best value" is not the best value; every bigger pack beats it. That label is a sticker on a flea-market table, not a calculation.

The US App Store in-app purchase list includes a Starter pack at $0.99 and a Popular pack at $4.99. On the UK App Store, a Starter pack costs £0.99, a Popular pack £4.99 and a Best value pack £9.99. The App Store pack listings do not say how many Bucks each pack contains.

That last point is the one I'd fix first. Selling a pack without saying what is in it is a gift card with the amount scratched off.

What the Bucks buy

A Rose gift costs 200 Bucks, and a Royal Rose also costs 200 but is VIP-tagged. A Teddy bear costs 400 Bucks, and a Rosy Bear costs 400 and is VIP-tagged. A Moon Ring costs 700 Bucks, and a Ruby Ring costs 700 and is VIP-tagged.

The gift shop has more tabs: Valentine, Treats, Accessories and Halloween. Seasonal tabs are a retail habit, and they tell me the shop is managed like a shop. Profile has a Wallet for the balance.

Is it worth it?

Gifts unlock personalised photos, so Bucks are really a second price list for pictures. Premium is the first. Paying both means paying twice for the same kind of content.

My advice: if a photo habit is forming, pick one route and stick to it. For conversation practice, Bucks buy nothing that helps, and the free daily Bucks are enough to see what a gift does.

Claiming Day 1 sets the balance to 25 Bucks.

At 25 a day, a 200-Buck Rose takes eight days of opening the app. That is a loyalty card at a diner: harmless, and designed to make the app a daily habit. I'd claim it and not think about it further.

The retention offer

Closing or cancelling the first $19.99 a month checkout triggers a "Wait - here's a better deal" offer. The offer is 30% OFF, labelled "Boosts all features $39.99". It runs a 24-hour countdown.

The Stripe checkout for it shows $39.99 every 3 months, about $13.33 a month.

That is a real discount against $19.99 a month. It is also a three-month commitment presented as a rescue. The countdown is there to stop anyone comparing it with the annual plans, which work out cheaper per month.

The funnel page headlines "Exclusive 50% off offer just for today!". It offers 1 Year Pro at $69.99, struck from $139.99, framed as "$0.26 per day". It also offers 1 Month Pro at $19.99, struck from $39.99.

The offer runs a 10-minute countdown. The maths behind it does not hold. $69.99 over a year is under twenty cents a day, not $0.26, and $19.99 a month is the normal web price rather than a discount.

The struck-through prices are props; the real prices are the ones on the regular paywall.

The email funnel also claims a "Money-back guarantee". The Terms say website sales are final, so I'd treat that badge as decoration and not as a promise.

Codes and referrals

There is no promo-code field. No referral or affiliate program was found. So there is nothing to hunt for on coupon sites.

My advice: ignore the timers. The $69.99 annual price is available in the stores without any countdown, and the retention offer is only better than paying monthly.

That is the context for everything in this section. I came to Blush for ordinary conversation, and the mild tone was a feature. The app in September 2026 has moved a fair way from that.

What the settings do in September 2026

There is no global adult-content toggle. Settings > General > Content & Media only has an "Animated photos" option. How spicy a chat gets is controlled per character instead.

Relationship settings also offer roleplay chips: Dominant, Submissive, Forbidden, Secret, Arranged, plus a couple of occupation-style roles such as Nurse. Each character also has Personality settings. Per-character control is a sensible design, since one character can stay a practice partner while another does something else.

The imagery and the marketing

In-app imagery is suggestive, with lingerie and cleavage. Blurred spicy photos sit behind Premium. A third-party review at intimeros.com reports that adult content is for Premium subscribers only.

App-review aggregator mwm.ai reports that many romantic or spicy dialogue options are gated behind the subscription. The email funnel on matchwith.ai goes further and uses openly adult marketing wording, promising uncensored chats and very spicy photos, videos and voice messages. It is a long way from the homepage's safe and fun environment.

Ratings and rules

The iOS app carries a 17+ age rating, and Google Play rates it Mature 17+. The App Store lists "Infrequent/Mild Sexual Content" along with bare skin. The UK storefront rates the app for adults only, while the US listing says 17+.

Prohibited uploads include obscene or sexually graphic content, harassing, hateful and infringing content. That rule matters most for the Upload button in the creator.

Moderation notes

Stated character ages are adult, and the creation age buckets start at 18-24. No document-based age check is stated.

Two images deserve a flag, and the details are kept plain on purpose. The onboarding "Anime girl" type card and the creator's "Anime" style card show a character drawn in a uniform-style outfit whose drawn age does not read as clearly grown-up. A "Help Brandy get better at chatting" slide shows a character in a uniform-style shirt and tie.

My verdict: the per-character settings are flexible and clear. The marketing has drifted far from the PG-13 app of 2023, and the gap between the homepage and the email funnel is the widest in this review.

The current Terms, Privacy Policy and App Store listing name the operator as Endura LLC. Endura LLC is listed at 945 Battery Street, San Francisco. Luka and Replika are not mentioned in the current Terms or Privacy Policy.

The Terms describe Endura LLC as a software company incorporated in Delaware and operating in San Francisco, CA. The iOS seller is Endura LLC. Google Play shows the developer as "Blush AI".

Google Play's "About the developer" section names Linden Research, Inc., 945 Battery St, San Francisco, CA 94111, contact [email protected]. Blush was built by Luka Inc. (Replika) and was reported acquired by Linden Lab around August 2025, per modemworld.me on 2025-09-06.

So the same address carries three names: Endura LLC, Blush AI and Linden Research. The likeliest reading is a display name, a legal entity and a parent, but the Terms do not say. For a reader, the practical point is simple: the company in the contract is Endura LLC.

The contract

The Terms of Service at blush.ai/legal/terms were last updated on Feb 7, 2023. A Terms page older than the current product and its current owner is a document nobody has reread in a while.

The Terms are governed by California law. They require mandatory binding arbitration through the AAA and include a class-action waiver. Liability is capped at the amount paid in the last 6 months or $100.

In plain English: a dispute goes to a private arbitrator, not a court, and not as part of a group. If something goes wrong, the most Blush owes is small.

Refunds

Website purchases through Stripe, PayPal or another processor are final and not refunded. Apple and Google purchases are also final and not refunded. The EU 14-day refund right is excluded. There are no refunds for unused parts of a month.

That is as strict as refund terms get, and it sits badly next to the email funnel's money-back badge. I'd assume no refund in any case.

Card data for purchases is collected by Stripe. That is normal, and it means Blush itself does not hold the card number.

Content and data

Users grant Blush a license to use their content to operate the service and for promotion, advertising or marketing. That covers what people upload and create. I'd keep real photos of real people out of the creator for that reason alone.

The Privacy Policy was last updated on Mar 18, 2026. Conversations are used to let the AI companion learn from the user's interactions. The policy promises that conversation content is never used or disclosed for marketing or advertising.

Third-party AI services process messages for fallback generation, memory summaries, media generation, voice synthesis and moderation. Personal data is removed or anonymized before that processing where technically feasible. Those services may not use the data for their own purposes, including training their models.

Ad partners receive device and click data. They never receive the conversations. Marketing emails and paywall analytics run through Adapty at mail.adapty.io.

The chat-training switch

Before the first chat, a "Help Brandy get better at chatting" screen asks permission to use chats to improve characters, with "Yes" and "No thanks". A "Use chats to improve AI" toggle sits in Settings > Privacy & Terms. The toggle is off after choosing "No thanks".

My verdict on the terms: the privacy side is better than average for this category. The contract side is standard and strict, and the ownership question deserves one plain sentence from the company.

That is the only friction in this review I'd call a feature. An app that sends its characters to bed is an app that sends its users to bed, and the export mattered more than the chat. Most companion apps are built to keep a user talking; Blush, at least then, was built to let him leave.

The nervous first date

One glitch stuck with me. A match sent the same "so, tell me about your weekend" message three times in a row, like a nervous first date.

In an editing timeline that is a duplicated clip, the kind of thing that slips through on a rushed export. It was funny rather than broken, but it did remind me that a script was running underneath the charm.

Sign-up and funnel friction today

The web app is rendered in a phone-width frame, even on a desktop screen. On a large monitor that looks like a phone photographed on a desk.

The password validator rejects a hyphen even though its own error message lists the hyphen as allowed. That is a small bug, and a sloppy one to leave on the very first form.

The "confirm" link in the welcome email is really a marketing opt-in: "By confirming, you agree to receive emails". Starting a chat shows 3 upsell slides, covering photos, videos and voice, gifts that unlock photos, and chat-training consent. A "Don't miss messages from Brandy!" notifications prompt follows.

Marketing theatre

The claim "98% of Blush users find a great match in 5 minutes or less" is marketing theatre. So is the "We've found 8 846 perfect AI matches" result screen. The email funnel shows scrolling masked-email "social proof" that looks fake, next to a countdown.

The funnel's money-back claim contradicts the Terms, which say the sale is final. That is the one piece of theatre with a real cost, because someone could pay on the strength of it.

Support, community and the stores

There is no help centre or FAQ: help.blush.ai and support.blush.ai do not resolve, and /faq and /help return the homepage. For a paid product with final sales, that is a real gap.

What does exist is a Discord server, a Facebook group, an X account and the blog. There is an app download link, plus store pages on the App Store and Google Play. The site links no Instagram, TikTok, Reddit or YouTube accounts.

The iOS app is rated 4.37 stars from 7,920 ratings and is at version 7.6.0, released Sep 8 2026. On Google Play it has 4.1 stars from 20.4K reviews and 1M+ downloads, and the listing was updated on Sep 5, 2026. Those ratings are decent, and they suggest the app is maintained.

Settings > General has Language, Notifications, Content & Media, Sounds & Vibration and Appearance. The interface is available in English, French, German, Italian, Portuguese and Spanish.

A third-party review at intimeros.com reports occasional reply lag, inconsistent media context and repeated off-topic photo messages. The repeated messages sound familiar. The nervous first date, it seems, is still on the schedule.

It is not for anyone wanting deep long-form roleplay, strong memory or a choice of model. It is not for users who dislike aggressive upsells, because there are several paywalls and offers before the first chat, plus blurred photos and countdown discounts.

It is not for people who need refunds: all sales are final, despite the email funnel's money-back guarantee. It is not for those who want free adult content, since media is Premium-gated. And it is not for desktop users wanting a full-width layout, because the web app is a phone frame.

My short version: Blush suits the nervous and the curious. It does not suit the committed, whether that means committed to a long story or committed to getting their money back.

That is the happiest ending in any review I have written, and the app deserves some credit for it. Training wheels are supposed to come off, and Blush never made me feel bad for taking them off.

How renewal and cancelling work today

The Premium paywall states "Auto-renewal. Cancel anytime through Stripe." Subscriptions renew automatically until cancelled. A subscription must be cancelled before it renews each billing period.

Cancelling happens through the marketplace where the plan was bought, which for web purchases is Stripe. Paid subscriptions can only be changed or cancelled through that marketplace. So an App Store plan is cancelled in Apple's settings, a Google Play plan in Google's, and a web plan through Stripe.

Profile also offers Manage subscription, Restore purchases, Privacy & Terms, Customer support and Log out. Manage subscription is the first place to look.

But cancellation of a paid subscription takes effect at the end of the current billing period, and it runs through the marketplace. Deleting the account does not stop billing by itself. This is the one contradiction in the Terms that can cost money, so I'd cancel the subscription first and delete the account second.

An account left unpaid for more than 7 days is downgraded to Free Use. Inactive accounts may be terminated.

Before paying

My checklist is short. Note where the plan is bought, because that is where it gets cancelled. Set a reminder a few days before renewal.

Cancel through the store or Stripe, and only then delete the account if leaving for good.

Deleting was easy for me. The paperwork around a paid plan today asks for a little more care.

Practise one real sentence. The most valuable thing I took from Blush was a sentence I drafted three times and then needed offline two weeks later. If there is a conversation you are dreading, rehearse it here.

Let the characters say no. When one told me to slow down, it taught me more than any compliment. Don't treat the boundaries as a bug to route around.

Don't expect memory. Restate the important context each session, because the app will not carry it for you.

Ignore the countdowns. The timers on the retention offer and the email funnel are there to rush a decision, and the regular prices are available without them.

If paying, pay through one store and write down which one. Cancelling happens there, not by deleting the account.

Claim the free Daily Reward if you like, but don't buy Bucks for conversation practice. They buy pictures, not practice.

And know when to stop. I deleted it after four weeks, when it had done its job, and that was the right call.

Memory is weak. It forgot plans from one week to the next in 2023, and today it credits its own opening line to the user.

The free tier is strong for what Blush is for. Chat, character creation and the Dating Coach are free, with no ads, and only media is blurred.

Value for money is weak. Too many overlapping plans, a Bucks economy on top, labels that do not match the maths, and no refunds.

Images and voice are mixed. They exist and they are central to the paywall, but free users mostly see blur, and they add nothing to the practice side.

Privacy and terms are mixed. The opt-in chat training and the no-marketing promise for conversations are good; arbitration, a wide content licence and a murky owner are not.

Cancelling and billing are weak. Renewal goes through the store, the Terms suggest deleting the account works, and the funnel promises refunds the Terms deny.

Content settings are good in design, with clear per-character control, but the imagery and marketing have drifted a long way from the gentle app the homepage describes.

Where Blush is better than most: it has a clear purpose, a Dating Coach, characters that can say "not yet", and an opt-in switch for chat training. Very few companion apps are designed around the user eventually leaving.

Where it is worse than most: memory, the funnel, and the price list. A new user meets two paywalls before the first hello, and the plan names change between the website, the store and the Terms.

My ranking logic is simple. As a practice room, Blush sits near the top of my list. As a companion to keep for months, it sits near the bottom, and it does not seem to mind.

The table below puts Blush next to other apps, so the trade-offs are easier to see side by side.

More on this site: the best AI girlfriend apps, free AI chat apps, the best AI companion apps, AI roleplay chat apps.

The app I would recommend is the free one: the swipe feed, the characters with boundaries, the coach and a hard sentence rehearsed until it sounds like me. The app I would not recommend is the one around it, with the blurred photos, the countdowns and a money-back badge the Terms contradict.

If you are nervous about talking to people, Blush is a decent rehearsal room. Rehearse, then leave the room. The real conversations are better, and they remember what you said.