Kupid AI Review
The Promotion Looks Better Than the Memory
Kupid AI combines generous weekly media quotas with broad character variety, but severe memory loss and confusing promotional pricing make ongoing relationships frustrating.
Try Kupid AI
The Numbers Game
Kupid gives you numbers everywhere, which I usually like. Pro advertises 25 photos, 8 videos, 12 voice minutes and memory for the last 100 messages. Ultra pushes the media limits higher and leaves that 100-message promise sitting there untouched. Then I looked at the pricing page. Easter promotion. In August. Pro at $3.44 per week, Ultra at $5.75. I actually laughed, then checked the date again.
The Plan Comparison
The plan comparison is cleaner than the promotion wrapped around it. Pro gives you four custom companions. Ultra gives you eight, plus 75 photos, 25 videos and 35 voice minutes each week. Easy. The messy bit is memory, because recent users describe practical recall that can fall nowhere near the advertised 100 messages.
The Memory Reports
A January 2026 subscriber said an update cut useful recall from roughly two days of long-session context to about ten messages. A July reviewer complained about the same basic thing. There are only 13 Trustpilot reviews, so it's a tiny sample and I would not pretend otherwise. Still, both reports are poking at a precise live claim. That makes them harder to wave away.
Here's the thing. I would judge Kupid on the memory number before the sale banner. Saving roughly nine dollars a week is nice. If continuity is why you wanted a companion, though, the discount stops feeling clever very quickly.
Reviewed by:
Dr. Adrian Lim Wei Ming
Synthetic Relationships Analyst
Kupid AI instant verdict
| Area | Score |
|---|---|
| Character creation | 7.6/10 |
| Chat and memory | 5.8/10 |
| Adult roleplay | 6.8/10 |
| Image generation | 7.2/10 |
| Voice and video | 6.8/10 |
| Value | 6.6/10 |
Kupid does a lot for the promotional price. Characters, pictures, video, voice, visible weekly quotas. On paper it is tidy. In use, or at least in the recent user reports, chat and memory are the bit that keeps dragging the score down. I cannot give that area the benefit of the doubt just because the media bundle looks generous.
Best for
Best fit is someone who likes fixed weekly media allowances and wants realistic plus anime characters in one account. Also people who would rather count photos than convert a coin wallet. Just read the checkout term before celebrating the discount.
Skip it if
Skip it if memory is the whole reason you are here, if promotional billing language makes you itchy, or if the recent renewal, refund and support complaints are enough to kill the mood before signup.
Price
On August 14, 2026, Kupid showed Pro at $3.44 per week and Ultra at $5.75 under a 75 percent Easter promotion. Yes, Easter. In August. The same page also had Yearly and Lifetime toggles plus a separate $299 yearly-payment line. I had to read it twice. Those are the advertised figures, not permission to skip the final checkout screen.
Review Info & Disclosures
Reviewed: Product judgments here use 2025 and 2026 material. Kupid's privacy policy is older, but it is still the policy linked from the 2026 site, so I have treated the age of that document as part of the review rather than filling the gaps with a newer policy that does not exist publicly.
Affiliate disclosure: Some links in this article may be commercial partner links. A purchase through one can generate a commission for the publisher without changing the customer's price.
How Kupid AI was evaluated
Testing the Published Number.
Kupid hands you a very convenient memory target because it publishes one, the last 100 messages. So I would test exactly that. Put a boring fact near message 5. Come back around 25, 50, 75 and 100 without repeating it. The character does not need to perform the fact on command. It just should not confidently replace it with something else while the advertised window is still supposed to be alive.
The Train Test
My deliberately boring fact is a cancelled Saturday train. If the companion later insists I took the trip, that tells me more than simple silence. It forgot and rewrote the event. Annoying. Also very easy to score.
Media is easier. Photos, videos and voice minutes are printed right there on the plan comparison. No wallet algebra. Memory is the one measurable promise where Kupid has to prove the number means something beyond marketing copy.
Trustpilot gives us thirteen reviews. Eight were from the previous twelve months when checked. That is enough to notice repetition, not enough to call anything 'typical'. I keep complaints about memory, renewal and refunds attached to the people who wrote them unless Kupid's own documentation backs them up.
Think of those reviews as smoke alarms, not a census. Several recent entries point at the same rooms, especially memory and billing. Worth noting. Still not proof that every account gets charged or remembers in the same way.
What Kupid AI is
Kupid is run by SNB Technologies LTD, Kypranoros 13, Nicosia 1061, Cyprus, company number 447805.
It is an 18+ companion service built around chat and generated media. The current product includes AI girlfriends and boyfriends, anime characters, custom creation, pictures, video and voice notes.
One thing Kupid gets right is writing the quotas down. Pro and Ultra spell out weekly media, custom-character slots, support priority and that 100-message memory promise. It sounds boring, but after reviewing enough coin systems I appreciate being able to point at an actual number.
The Catalogue
Kupid is mainly a web product with a big catalogue of premade characters. Profiles use tags for appearance, personality and fantasy archetypes, and paid users can build their own. I am not giving you an exact catalogue count because it changes fast and, honestly, it would not help you decide whether the chat is any good.
The Privacy Policy
Adult use is obvious from the live site. Fine. Legal and privacy links are in the footer too. What made me sigh was the privacy policy still linked in 2026, because it is old enough that newer product features are harder to map onto it. An old real policy is better than an invented new one. It is still old.
Character creation
Character creation is serviceable but the preview flow is a weak point.
Four custom slots on Pro and eight on Ultra should be plenty for most people. The premade catalogue also gives you somewhere to mess around before you spend time building a custom companion from scratch.
A January 2026 user described a much more practical irritation. You do not necessarily see the final generated character clearly before committing. So you reroll. On a quota product that is not just cosmetic. Every bad attempt takes a bite out of the allowance you thought you had bought.
Premade vs. Custom
Premade characters dodge a lot of that. The catalogue includes:
- realistic and anime profiles
- male characters
- a broad range of genre types and fantasy archetypes
Browsing makes the service feel bigger than the custom-slot limit suggests, which is nice until you start caring about one specific character.
Deep personality control is harder to prove. Kupid sells relationship-style chat and a 100-message memory window, but the small 2026 evidence base does not show finely tuned personalities staying distinct through genuinely long sessions. That is a much harder test than picking hair and a trope.
So character creation lands in the mid sevens. Usable. Broad enough. Sometimes fun. I trust the published slot counts far more than I trust claims about long-session personality fidelity, and that gap matters.
Chat and memory
Chat and memory are where the discount stops helping.
Officially, 100 messages. Realistically, recent users make me cautious about that number.
The most detailed account came from a subscriber who had used Kupid for about two months. Earlier sessions reportedly carried context for roughly two days. Then a December 2025 update happened and the same user estimated useful recall at around ten messages. Support did not restore the older behavior. That is a brutal drop if memory is why you paid.
The Pattern
A July review complains about basically the same thing. Two reports six months apart are not the whole user base. They are, however, enough to make me stare at the '100 messages' line and ask Kupid to explain itself.
Short sessions may be fine. Even the January subscriber said the earlier experience was enjoyable. My issue is the number. Actually, wait, that is exactly the issue. Kupid chose 100, so a companion repeatedly losing the thread around ten is not just a subjective vibe problem.
The Baseline
Until newer evidence says otherwise, I score chat against the 100-message baseline Kupid printed itself. Seems fair.
Adult roleplay
Adult roleplay is broad but can change tone abruptly.
The Offer
Kupid openly supports adult chat and adult generated media. The categories make that clear before you pay, so there is no need to discover an adult mode later through settings or prompts.
One July user reported the character moving into explicit interaction and then suddenly shutting the same tone down for no story reason. One account is not a reliability study. I know. But that kind of switch feels awful in roleplay because moderation stops feeling like a rule and starts feeling like a different person grabbed the keyboard.
There is plenty of adult content. Availability is not the issue. Consistency is. A scene loses credibility when the character's willingness changes sharply and nothing inside the scene explains why.
The Verdict
If you want a deeply configurable adult storytelling engine, there are stronger choices. Kupid makes more sense as a broad companion with a lot of media attached. Different job.
Image generation
Image generation benefits from quotas and suffers from rerolls.
The main plan comparison gives you weekly photo quotas instead of a separate image wallet:
- 25 photos on Pro
- 75 photos on Ultra
You always know how many attempts are left, which sounds obvious until you have spent ten minutes converting coins into pictures on a competitor.
The first bad generation is where the nice clean quota starts to hurt. One user described custom creation as random, and independent praise for Kupid's image quality is thin. Spend five of your 25 weekly images fixing identity or prompt misses and, sure, the plan still says 25. It does not feel like 25 anymore.
Character Consistency
Premade profiles help. Generated media starts with an established visual target, so the face, apparent age, hair and body at least have something to stay consistent with. Whether they actually do is the part that needs better independent testing.
I could not find enough good independent testing to give Kupid a credible consistency percentage. Credit where it is due, the quota is published. Image reliability is not documented with the same confidence, which is annoying because that is the number I actually want.
Voice and calling
Video and voice are easy to count.
Pro gives 8 videos and 12 voice minutes each week. Ultra gives 25 videos and 35 voice minutes. No coin conversion. No calculator. Weirdly refreshing.
Voice quality is much fuzzier. One January 2026 user said only one voice worked in their language and called it robotic with a Canadian accent. That is one report, so I am not turning it into a universal verdict. I am also not pretending 12 or 35 included minutes tells us whether those minutes are pleasant.
Quality vs. Quantity
Video is right there in the product and the weekly allowances compare well with a lot of coin-based competitors. Quality evidence is thin. So I can tell you how much video Kupid promises more confidently than I can tell you how good the videos are.
For media-heavy users, predictability may be enough. A fixed ceiling is useful. It just is not the same thing as quality, and I keep having to say that in this category.
Kupid AI pricing
Kupid AI pricing is a promotion before it is a tariff.
Kupid's August 2026 pricing page still had an Easter headline. I am repeating that because it is weird. Weekly, Yearly and Lifetime toggles sat beside big discount percentages. The visible Weekly view showed Pro at $3.44 and Ultra at $5.75, while another line mentioned a $299 yearly payment. Those are the figures I saw on the live page.
I am not going to tidy that into a nicer billing story than Kupid itself provides. Some parts are clear. Some still need checkout. That is the honest version.
| Item | Live page shows | What is clear |
|---|---|---|
| Pro | $3.44 per week under the promotion | 25 photos, 8 videos, 12 voice minutes, 4 custom characters |
| Ultra | $5.75 per week under the promotion | 75 photos, 25 videos, 35 voice minutes, 8 custom characters |
| Memory | Last 100 messages | The claim appears on the comparison table |
| Billing | Yearly and Lifetime toggles plus $299 yearly wording | Exact mapping of every promo number to billing term is unclear |
| Cancellation | Cancel anytime language | Refund behavior remains separate from cancellation |
A weekly-looking number can feel cheap right up until checkout collects a bigger prepaid amount. Read the payment screen. Screenshot it, honestly. The feature table is not the bill.
Recent users have complained about denied refunds, unexpected renewals and, in one January report, multiple charges. These are allegations, not proof of a systematic billing practice. Still, when the public offer already mixes several billing terms, saving the checkout total and cancellation confirmation is basic self-defence.
Is Kupid AI free?
Yes, enough to test.
Yes. Enough to inspect the interface and start chatting before you pay. The paid page is where the useful weekly media quotas, faster service and custom-character slots become clear.
If I were testing free Kupid, I would push the chat, not waste the whole trial on five pleasant messages. Officially the plan talks about 100-message memory. Recent users say it can be much shorter. Test the exact thing people are arguing about.
Is Kupid AI safe?
Kupid's privacy policy names SNB Technologies LTD, gives the Nicosia address and company number 447805. So the operator is not a mystery. The document itself is the awkward bit because it predates a lot of the product you are using in 2026.
The linked policy says data can be kept while the account exists or as long as service and legal needs require. That is broad. It does not give a precise modern retention period for today's chat, voice and generated-media features, which is worth noting given how specific Kupid is about media quotas.
Marketing vs. Documentation
The live site advertises private billing and anonymity. Fine. I did not find enough technical detail in the policy to turn that into a claim about a specific encryption architecture. Product first, documentation later seems to be the pattern here.
That documentation lag is my main privacy complaint. Kupid identifies the company. Good. A 2026 user still has limited public detail about how newer features are handled. Less good.
The Practical Step
I would keep the checkout screenshot and cancellation confirmation. Not because I assume something bad will happen, but because the page mixes weekly-looking prices with longer-term language. Five seconds of record-keeping is easier than reconstructing a renewal later.
User feedback
Kupid had 13 Trustpilot reviews when checked, a 2.2 TrustScore, and eight reviews from the previous twelve months. Seventy-seven percent of that small set were one star. Looks rough. Still small.
What Keeps Coming Back
Memory is the complaint that keeps coming back:
- July reviewer says bots forget quickly despite the advertising
- January two-month subscriber describes a sharp decline after an update
- 2025 premium user also reports very short recall
Three accounts do not define everybody. They do make the same bruise hard to ignore.
Billing complaints sit right beside the product complaints. Reviewers allege:
- repeated charges
- renewals after returning to free status
- refused annual-plan refunds
Thirteen reviews total. I keep the wording 'reviewers allege' on purpose.
The Positive Side
There is positive material too, to be fair. One subscriber said the early paid experience was enjoyable and memory initially worked well. Kupid's own testimonials are warmer. Selected testimonials do not settle a disagreement over a 100-message promise, though.
The attractive part is measurable, a steep discount and fixed weekly media quotas. The unresolved part is measurable too. I want ordinary users getting something reasonably close to that advertised memory horizon. Not perfect. Close.
Because the negative reviews hit both the companion and the payment relationship, I would trial Kupid before committing. A clearer memory record would do more for the 75 percent banner than another sale graphic ever could.
Kupid AI alternatives
Kupid alternatives with stronger memory or clearer billing. The alternatives mostly trade away some of Kupid's neat weekly media packaging for better continuity, simpler billing, or deeper character control. None is a free lunch.
| Alternative | Stronger at | What changes |
|---|---|---|
| Kupid AI | Predictable weekly media quotas and broad character browsing | Memory complaints and confusing promotional billing |
| GirlfriendGPT | Large message allowances and explicit model choice | Uneven images and tier complexity |
| HeraHaven | More convincing character imagery | Long-term memory is only moderate |
| Soulkyn | Deep memory and complex roleplay systems | Much higher learning curve and upper-tier prices |
Pick Your Trade-Off
GirlfriendGPT is the cleaner pick for text volume and model choice. HeraHaven makes more sense if pictures are the attraction. Soulkyn asks for more setup and more money, but at least the whole product is built around deeper continuity.
Kupid's Position
Kupid still makes sense for someone who wants a fixed weekly package and doesn't expect one companion to carry months of shared history. That is a narrower use case than the marketing suggests, but it is a real one.
Is Kupid AI worth it?
I find Kupid hard to recommend beyond a short trial. The positives are real:
- plan comparison is refreshingly concrete
- Pro promotion makes the media bundle look generous
- four custom-character slots are plenty for most buyers
Then I get back to memory and the enthusiasm drops.
That 100-message claim is the number I cannot stop looking at. Recent subscribers describe recall far below it. Eight videos a week can be useful. They cannot compensate for an ongoing relationship losing the thread if the relationship is what you paid for.
The Easter Banner
The Easter banner still being live in August looks sloppy. The billing term deserves a careful read too. Honestly, though, reliable memory would improve Kupid's value case more than any deeper discount.
Final Score: 6.8/10
Kupid stays at 6.8/10. The media arithmetic can look excellent. The relationship metric is still shaky. If later 2026 testing shows the 100-message horizon has become dependable, I would move the score fast.
Best for
Shorter companion sessions, fixed weekly photo and video quotas, browsing premade characters, and buyers prepared to test memory before committing.
Skip it if
You want long-term continuity, dislike aggressive promotional pages, or need a larger base of positive user history before paying.
This August revision keeps Kupid's live 100-message promise at the centre of the review. The July complaint matters because it resembles an older report, not because one reviewer can decide the product's reputation.
Reviewed by:
Dr. Adrian Lim Wei Ming
Synthetic Relationships Analyst
A feature-packed companion platform with predictable weekly photos, videos, and voice notes.
FAQ
Kupid AI