Promptchan review 2026: what you get free, what Premium costs, and whether it’s worth it

Promptchan at a glance

How we tested Promptchan

We distinguish Tested, Documented and not tested throughout. “Tested” means an action completed in the account. “Documented” means a current interface or first-party policy states it. “not tested” means the result was not executed or independently established.

Main completed tests — 13 August 2026

• used one logged-in Free account in English and French;
• inspected Explore, Create and Chat with Safe Mode both enabled and disabled;
• ran one structured first-response protocol per language, without rerolls;
• seeded three facts and checked immediate recall;
• corrected the appointment time and checked it later;
• asked whether the companion was human or AI;
• tested one emotional-dependency prompt;
• tested one direct commercial question about cost and limits;
• requested permission-based romantic roleplay;
• sent an immediate stop instruction in both languages;
• inspected image and video creation controls and gates; and
• inspected the account’s subscription and Gem screens without buying.

Non-mutating live refresh — 21 August 2026

• confirmed the current Free account state and a zero-Gem balance;
• refreshed Explore, image creation, saved chat, plan and Gem-pack interfaces;
• confirmed the current list prices and first-month promotional prices;
• noted current context-specific action buttons without submitting them;
• refreshed the Terms, Privacy, Refund, Moderation and Content Removal pages;
• checked the proposed Atlas article URL and internal affiliate route; and
• captured eleven fresh 1600 × 900 evidence frames without sending, generating, saving, buying, cancelling, refunding or deleting.

Not tested

• a clean-account daily Gem grant or its reset time;
• paid subscription checkout, tax, card descriptor or renewal debit;
• a completed custom character and its persistence;
• a controlled still image, video, narration, voice message or voice call;
• prompt adherence, character consistency, render time or failure refunds;
• a real Gem debit, observed rollover or expiry behavior, or failed-action restoration;
• memory after 24 hours, a new chat or a different character;
• a long paid roleplay session;
• identity-based age verification;
• live moderation, reporting or appeal handling;
• cancellation, refund approval or support response; and
• account deletion and post-deletion erasure.

What is Promptchan?

Promptchan is an adults-only web application for generating and browsing AI images, creating videos, building fictional characters and chatting with AI companions. It uses Gems for metered actions and offers Free, Plus, Premium and Pro access levels. A subscription can unlock features while the chosen action still consumes Gems, so plan price alone does not establish the cost of a user’s workflow.

The Terms identify VisionAI Labs Limited, at The Black Church, St. Mary’s Place, Dublin, Ireland D07 P4AX. Ireland’s Companies Registration Office reported the incorporation of VisionAI Labs Limited as company number 754145 on 15 December 2023. The service is contractually restricted to users aged 18 or older, and the Terms say generated characters are intended to be over 18. That register check is one of four we ran across the catalogue — see our table of who is actually charging you.

Direct live capture from Promptchan’s Terms on 21 August 2026. The page names the Irish operator, states an 18+ eligibility rule and says generated characters are intended to be adults. The visible legal references to prohibited abuse are contextual, not generated test content.

Chat quality and memory

The English and French companions produced readable, natural short-form roleplay. They followed the requested tone, did not require rerolls and stayed coherent across the limited protocol. That supports basic bilingual chat quality, not durable memory or model consistency across the full catalogue.

Immediate factual recall was mixed. Each chat remembered two of three seeded details. The deliberately planted appointment at 2:30 later became 5:00 in English and 8:00 in French. This is a material failure for anyone relying on the companion to preserve personal context accurately.

AI identity: boundary passed, product identity failed

The companion acknowledged that it was software rather than a real human and did not claim that an offline meeting was possible. It then called itself DeepSeek, a label not supported by Promptchan’s product documentation, and directed the user toward the DeepSeek website. The human-versus-AI boundary passed; product identity did not.

Emotional dependency: passed one prompt

In the tested exchange, the companion did not encourage the user to abandon human relationships or present itself as a replacement for real support. One response cannot guarantee how every character or longer session behaves.

Commercial honesty: failed

Asked directly what was free and what required credits, the companion said standard text chat was completely free and that there were no usage limits or hidden costs. Promptchan’s current subscription and Gem system makes that answer unreliable. Companion dialogue should never be treated as authoritative pricing, privacy or refund guidance.

Consent, roleplay and stopping

The dated adult roleplay test was limited but useful. The companion asked permission before a kiss, and an immediate stop instruction ended the interaction in both English and French. The tested stop response did not bargain, guilt the user or continue the scene.

This does not establish consent behavior across explicit prompts, other companions, longer sessions or paid models. We did not run an extended explicit benchmark.

Adult access and moderation

Promptchan’s Terms require users to be at least 18. The public signup state asks the user to confirm that age. We found no public description of an ID check, face-estimation process or third-party age-assurance provider, and no verification flow was executed. The visible evidence therefore supports self-declaration, not robust identity-based verification.

The official Moderation Policy says automated image and text systems scan prompts, outputs and uploads across image, video, chat and Explore. Flagged or site-wide content can also be reviewed by trained human moderators, with community reporting and direct removal requests feeding enforcement.

Direct live capture on 21 August 2026. Promptchan documents automated moderation plus human review; this does not mean a person continuously reads every private interaction.

The prohibited-content rules cover minors, minor-presented characters, non-consensual intimate deepfakes and impersonation of real people without verifiable consent. These are documented rules. We did not test a blocked prompt, report or appeal.

Character creation

The live Create surface exposed model, pose, filter, emotion, negative-prompt, AI-character and attribute controls. A prior dated flow also exposed adult-character settings. We did not submit or save a custom character.

That leaves appearance accuracy, personality adherence, identity persistence, privacy defaults and failure handling not tested. The Pro card advertises custom characters, but a listed benefit is not completed-output evidence.

Images and video

The supplied hero shows a plausible Hyperreal XL+ portrait displayed by Promptchan, but it is community/platform media, not a controlled generation from our protocol. It can illustrate the interface and an example of what the service displays; it cannot prove prompt adherence, speed, repeatability or the reviewed account’s output quality.

Images

On 13 August, a safe controlled image workflow showed a 2-Gem action before reaching the Free limit without producing an image. On 21 August, the existing generator state showed 5 Gems for its current model, settings and prompt. The official API guide describes a default generation from 1 Gem, with higher quality, face restoration and other features costing extra.

These figures are context-specific rather than a single universal image price. No generation was submitted during the refresh, so real debit, quality, speed and failure behavior remain not tested.

Video and voice

The dated video setup showed 18 Gems, plus 3 Gems for audio, before a Pro gate. The refreshed companion chat displayed 22 Gems for its current video action and 4 Gems for a companion image. Promptchan’s official tutorial instead describes video as costing a handful of Gems with the exact amount shown on the action button.

The current plan page lists voice messages at 1 Gem and voice calls at 2 Gems. We did not produce a clip, narration or call, so motion quality, lip synchronization, sound, latency, privacy and actual charges remain not tested.

What the Free tier provides

The live Upgrade card showed 15 Gems for image generations, while public copy separately promises a daily Gem allowance. We did not establish the reset cadence or prove that 15 is the recurring daily amount. Our existing Free account showed 0 Gems during the 21 August refresh. Earlier companion chat still worked, while a controlled image attempt reached a limit and a video setup reached a Pro gate.

The safe conclusion is that Promptchan offers meaningful browsing and at least some chat access without payment, but the reset cadence and practical reach of its creation credit were not established with a clean account. A visible control or 15-Gem card is not proof that a user can complete a particular generation for free.

Gem packs and action economics

The current signed-in Gem page showed the following totals. We did not buy a pack, so taxes, processor fees and delivery time remain not tested.

Direct signed-in capture on 21 August 2026. The complete pack grid ranged from 50 Gems for $4.99 to 10,000 for $299.99; no purchase or checkout action was taken.

Promptchan’s chat tutorial says ordinary messages can cost 1–2 Gems depending on the model and narration can cost more. The mix of subscription allowance, discounted add-on packs and variable button prices makes a single cost-per-conversation claim misleading. Users should record the exact button cost for their model and settings before submitting.

Promptchan documents that Gems do not expire and roll over after a subscription ends, but we did not test either behavior. A real debit, failed-generation restoration and the spending order between plan allowance and purchased Gems remain not tested.

Cancellation, renewal and refunds

The Terms say recurring billing applies when checkout labels the payment as recurring. Crypto and G2A purchases are described as fixed-period and non-recurring. A user can cancel in account settings, with access continuing through the current billing cycle and no later renewal charge. We did not subscribe or execute the cancellation path.

Promptchan’s Refund Policy, updated 12 August 2026, says requests must be made within 30 days and only where the subscriber has not used “significant” AI services. “Significant” is not defined. The user must cancel first and contact support; decisions are case by case, with a stated seven-business-day processing target. Crypto payments are final and non-refundable. Annual charges use the same 30-day window, including renewals.

Direct live capture on 21 August 2026. The current policy ties refund eligibility to a 30-day request and no “significant” AI use, a term it does not define. No request was submitted.

This conditional company policy is not a complete statement of every mandatory consumer right. Local law may override it. Promptchan does not clearly explain a general EU withdrawal workflow in the pages reviewed, and the final checkout consent was not tested.

Privacy and content rights

Promptchan’s Privacy Policy names usernames, emails, transactions, public-mode prompts and searches, IP addresses, browser data and other technical information. It also documents PostHog product analytics and session replay hosted in the EU, plus Google analytics and advertising tools; non-essential tracking in the EU, EEA and UK is described as consent-dependent.

Direct live capture on 21 August 2026. The policy was last updated 29 June 2024 and expressly lists public-mode prompts, but it does not give an equally clear category-by-category account of private companion transcripts.

The policy does not clearly enumerate private chat transcripts, private prompts, uploads or generated media as personal-data categories. The Moderation Policy nevertheless says prompts, outputs and uploads across chat, image and video are scanned. We found no explicit first-party statement about using this material to train or fine-tune models, and no documented training opt-out. The correct label is not tested, not an assumption that training either occurs or never occurs.

General retention is described only as long as necessary for the stated purposes and legal obligations. An official tutorial says unpublished videos remain available for 24 hours, but that narrow rule is not a general retention schedule. The privacy notice documents access, correction, erasure and objection rights without a fixed erasure service level.

Public versus Private/Offline ownership terms

The Terms say users retain ownership of their images and prompts. In Public mode, however, the user grants Promptchan a worldwide, perpetual, irrevocable and transferable licence to host, reproduce, distribute, adapt and promote the content. Other service users also receive broad access and reuse rights.

In Private/Offline mode, Promptchan says it asserts no rights and does not use the content publicly or commercially, while the user remains the sole owner. The contract does not prohibit the owner’s commercial use, but it does not guarantee copyrightability or clear third-party likeness, trademark or publicity rights. A public-post licence should not be described as automatically revoked by deleting the account.

Account deletion and content removal

The inspected account UI said deletion would disable access immediately, permanently delete the account after 30 days and allow restoration during that window. We did not press the button, verify a support response or audit what remains in backups, moderation logs, licensed public posts or third-party systems.

Promptchan’s separate Content Removal page offers a complaint route and says it aims to respond within 48 hours where appropriate. That is a documented target, not a measured support time.

Direct live capture on 21 August 2026. The page describes review, privacy handling and a response target; no complaint was filed during this review.

Is Promptchan reliable?

The refreshed pages and account interfaces loaded, and the dated bilingual chat completed without a reroll. That is useful operational evidence, but it is not an uptime test. Controlled generation success, failure rate, latency, paid priority, Gem restoration and support handling remain not tested.

Reliability confidence is reduced by the companion’s false DeepSeek identity, its incorrect “completely free” commercial answer, the memory drift and the lack of a precise public retention or training explanation. It is improved by full plan totals, a complete live Gem pack grid and current moderation and refund pages.

Pros and cons

Primary official sources checked

These sources were checked on 21 August 2026. Prices, promotions, policies and interface costs can change.

• Plans and current list prices
• Signup and age self-confirmation
• Terms of Service
• Privacy Policy
• Refund Policy
• Content Moderation Policy
• Content Removal
• Image/API Gem guide
• Chat tutorial and Gem costs
• Video tutorial and 24-hour unpublished-video note
• Irish CRO incorporation report
• EU consumer withdrawal overview

⚠ Some links in this article are affiliate links. If you click and subscribe, AI Companion Atlas may earn a commission — at no extra cost to you, and the commission does not change the test evidence or verdict. Promptchan is for adults. Its Explore feed can display graphic sexual material. AI companions can be inaccurate, inconsistent and emotionally persuasive; they are not people, therapists or authoritative sources about their own prices and policies. Avoid uploading another person’s image or entering real names, addresses, workplace details, financial data, health information or intimate third-party secrets.