Frequently asked questions

PhantomAI is a free AI assistant you use in the browser. You type a question or paste in your material, and it replies — with support for switching between two models, real-time web search, document and image analysis, a live canvas for rendering code, and memory for your custom instructions. Alongside the app, the site publishes practical guides on using AI assistants well.

Two open-weight models, served through Groq: GPT-OSS 120B is the default and handles reasoning, code, and longer answers, while GPT-OSS 20B is the fast option for short questions. You choose per message. PhantomAI does not run GPT-4, Claude, or Gemini, and does not train its own models.

Yes, and there is no paid plan to upgrade to. Every account gets 150 Phantom Credits per day. A message to the fast model costs 5 credits, the default model costs 10, web search adds 4, an image costs 8, and a document costs 15. The full cost table is on the pricing page.

A 100-credit recharge is scheduled automatically two hours after your balance reaches zero, and your full daily allowance returns the next day. You are never asked for money, because there is nothing to buy.

Not to read the site. The homepage, guides, blog, and all informational pages are fully public with no sign-up wall. You only need an account to use the chat itself, so your conversations and settings can be saved against your profile.

Yes, and this matters more than most product pages admit. PhantomAI predicts plausible text rather than looking facts up, so it can produce confident, fluent statements that are false. It is especially unreliable with citations, arithmetic, and recent events. Verify anything that matters — the guide on fact-checking AI output explains how to do it quickly.

Yes. Asking a language model for citations reliably produces correctly formatted references to papers that do not exist. Find sources first using Google Scholar, PubMed, or your library, then paste them in and ask PhantomAI to work with what you provided.

No. There is no execution environment, so any claim that code works is a prediction rather than a test result. The canvas renders HTML, CSS, and interface snippets for preview, but it does not execute or test back-end logic. Always run and review generated code yourself.

A side-by-side panel where generated code, HTML, and interface snippets render as they are produced, so you can preview and iterate without copying anything into a separate editor.

It can run a real-time web search and read specific URLs you give it, then use that material in its answer. Search costs 4 credits on top of the message. Search grounding reduces invented facts but does not eliminate them — it can still misread a page, so open the source for anything important.

You can save custom instructions — your role, preferences, and writing style — and PhantomAI applies them across conversations so you do not have to repeat yourself. It is opt-in, and you can review or clear it in settings.

Your conversations are stored so history persists, and messages are sent to Groq to generate responses. Your conversations are not used to train models. They are not end-to-end encrypted, because a model has to read your text to answer it. Do not paste passwords, API keys, payment details, or other people's personal data.

No. Personal data is not sold. Running costs are covered by advertising on the public guides and blog, not by selling user information. Ads are not shown inside the chat interface or on sign-in and sign-up screens.

No to all three. Accounts are individual, there are no shared workspaces or admin controls, and there is no API to call PhantomAI from your own software. Earlier versions of this site advertised Pro and Team plans; they never existed and have been removed.

Simanta Pratim Das, an independent developer based in Guwahati, India. PhantomAI is a solo project — there is no company, no support team, and no investors behind it. That is a genuine limitation worth knowing: replies to support email come from one person.

Not for decisions. It has no professional credentials and no accountability, and it can be confidently wrong. It is genuinely useful for understanding terminology and preparing better questions for a professional, which is a different thing from getting advice.

Yes, and it works best as something you argue with rather than copy from — have it quiz you, mark your explanation, or generate practice problems. Submitting generated text as your own work is misconduct at most institutions, so check your course policy. The studying guide covers workflows that actually help retention.

Yes. It writes, explains, refactors, and debugs code, and it is strongest when you paste in the real error and the relevant file rather than describing them. It cannot run anything, so review and test everything before it reaches production.

Yes. The site and app are responsive and PhantomAI can be installed as a progressive web app from your browser. There is no native iOS or Android app in an app store.

Yes. Conversations can be deleted from the app, and you can request account deletion by emailing phantomai2009@gmail.com. See the privacy policy for what is retained and for how long.

Email phantomai2009@gmail.com or use the contact page. There is no priority tier — everyone reaches the same inbox, and it is one person reading it, so please allow a few days.

Create a free account and open the chat. If you would rather learn something before signing up, the guides are free to read and require no account.