VirtualGF Review

VirtualGF's privacy policy tells you which company generates its vector embeddings. It tells you the country, the transfer mechanism, and links the data processing agreement. It does the same for its language model vendor and its database host. It cites GDPR legal bases by article number.

It never tells you who VirtualGF is.

That is the whole review in two paragraphs, and the rest is working out how much it should cost them.

What is VirtualGF?

VirtualGF is an AI companion platform at virtualgf.app, pitched squarely on memory: "The first AI companion who truly remembers you. Every word. Every day. Every emotion." You chat with a companion, she messages first, she sends photos, and you can send her one back. There is no character builder marketplace, no video generator, no gallery of other people's output. It is one companion and a conversation, which is a narrower product than most of this category ships and is clearly a deliberate choice.

It runs in six languages and the interface strings are written Italian-first, which is a small clue about where the people behind it sit. It is the only clue you get.

Because here is the thing the documents do not do. The terms of service define the Owner as "the operator of the VirtualGF service". The privacy policy states that "the data controller is the operator of the VirtualGF service". The governing law clause says disputes go to "the courts of the Owner's jurisdiction" without naming one. Every one of those definitions is circular. There is no company name, no registration number, no country and no postal address anywhere on the site.

That is not merely unhelpful. A privacy policy that cites Article 6.1.b by number is a policy that knows what the GDPR requires, and Article 13 requires the identity and contact details of the controller. Doing the hard parts of data protection law while omitting the first and easiest one is a strange way to run a service.

Policy check, and VirtualGF passes it cleanly on mechanism. Across its bundles there is no image-to-image, no initial image, no reference image, no identity adapter, no face swap, no nudify and no inpainting. There is no image generation pipeline here at all: the companion's photos are served assets, not generations conditioned on anything you provide.

But there is an upload path, and it is an advertised feature rather than a hidden one. "Send a photo to your companion" appears on both tiers, with Premium allowing more of them, and the chat API carries a photo field alongside the message. This is the same shape as the vision feature we documented on AIMOUR, where your companion looks at a picture you send. It is not a route to generating anything.

What it is, though, is a route for personal photographs to leave your device. And the privacy policy never mentions photos or images once. It enumerates account data, messages, extracted memories, payment identifiers, IP addresses, browser language and interaction timestamps. Not photographs. A service that invites you to send pictures should say what happens to them, how long they are kept, and which of its named American sub-processors sees them. This one lists everything except the category of data most likely to identify you.

The features that actually matter

Memory as the product, not a feature. Deep memory on Premium, basic on free, with the free tier holding only the last 50 messages. The entire pitch rests on continuity, and the architecture backs it: a dedicated embeddings vendor exists purely to power the memory system.

She messages first. The companion initiates rather than waiting, which is the mechanic that makes these products sticky and the one worth knowing about before you install anything.

A disclosed model vendor. The privacy policy names Anthropic as the language model provider, states the integration uses the zero-retention API, and says conversations are not used for training. We have reviewed dozens of these platforms and can count on one hand the ones that tell you whose model is answering you.

No currency. No tokens, no gems, no credits, no coins. One subscription, one price. After a month of reviewing platforms where the subscription is merely the entry fee to a metered economy, this is worth saying out loud.

Pricing, without the bullshit

That is the entire rate card, taken from the application itself. Cryptocurrency is accepted, and the interface shows a crypto credit balance filling up toward the $9.99, which is an unusually honest way to present a partial payment.

Refunds are handled by the merchant of record rather than by VirtualGF, and the terms say so plainly while declining to name which merchant of record that is. There is no refund policy page. There is no 2257 statement page either, though with no user-directed image generation the exemption argument is straightforward enough that its absence matters less here than it would elsewhere.

One clause is worth quoting because it is the kind of thing people find out too late: if your account is closed for a terms violation, "you are not entitled to any refund". Reasonable, common, and better stated than hidden.

On the credit side, the operator reserves the right to change prices with at least 30 days' notice by email. Thirty days is a real commitment and most rivals promise nothing at all.

What users say

There is modest independent coverage and it is mildly positive. One reviewer reports testing a paid subscription over several weeks and finding no hidden charges, billing problems or deceptive practices, which given how many platforms in this category fail exactly that test is worth reporting. The same coverage flags two limitations we can corroborate from the product: the free tier's context window is short, and there is no readily available support channel.

What does not exist is a large reputation footprint. No Trustpilot, no scam-scanner consensus, no volume of user reports. For a platform that is asking for a modest $9.99 that matters less than it would at $239, but it still means you are extending trust rather than verifying it.

The privacy documentation cuts both ways here. Naming DigitalOcean with EU primary storage, Anthropic under zero retention with standard contractual clauses and the EU-US Data Privacy Framework, and Voyage AI for embeddings with its own DPA, is real work by someone who understands the regime. Committing to delete everything within 30 days of account deletion, suspending unverified accounts after 7 days and purging them after 16, is specific in a way that boilerplate never is. Nobody writes that by accident.

Which makes the anonymity read less like carelessness and more like a choice. Somebody here knows exactly what the law asks for. They have answered every question except who they are.

Final verdict: who is VirtualGF for?

Get VirtualGF if: you want one companion, one price and no currency to manage, memory and continuity matter more to you than image generation, or you specifically want a platform that discloses its AI vendor and its retention posture.

Skip it if: you need to know who is processing your data before you hand it over, you were planning to use the send-a-photo feature (nothing published covers what happens to it), or you want a support channel that exists.

The formula we'd pick: stay on the free tier while you judge whether the memory actually delivers, since 20 messages a day is enough to test the one thing this product sells. If you upgrade, $9.99 monthly is a low-risk commitment and the crypto route is there if the anonymity of the operator makes you want some of your own. Do not send it photographs of yourself. Not because we think something bad happens to them, but because nobody has told us what does.

Rating: 6.8/10. The most carefully documented data processing in this category, performed by nobody in particular.