Free, and we mean it
No card, no trial that quietly ends, no counter ticking down in the corner.
How it compares to ChatGPTAI chat that searches the web
Ask it something that happened this morning and it goes and looks, then shows you where the answer came from. Free, no account, the box below is live.
The problem with an AI that never looks anything up
A language model is a very well-read assistant who stopped reading on a particular day. Everything it knows, it knows from training data with a cutoff date — and the failure mode isn't that it says "I don't know." It's that it answers anyway, fluently, in the same confident register it uses for things it actually knows.
That's fine for "explain recursion" and useless for "is that shop open today." The gap isn't intelligence, it's recency, and no amount of better prompting closes it.
Search closes it. When a question needs current information, this chat runs a web search while it's composing the answer, reads what comes back, and answers from that — then lists the pages it drew on as links underneath the reply. That last part is doing more work than it looks like: it turns "trust me" into "here's where this came from," and checking takes one click.
It decides when to search, so you don't have to
There's no toggle to flip and no special command to remember. Ask a question that needs current information and it searches; ask it to rewrite your email, explain a concept, or brainstorm names and it just answers, because a search there would add latency and nothing else.
The tell is the sources list. If links appear under an answer, it looked things up and those are the pages it used. If none appear, the answer came from what the model already knew — which is the right call most of the time, and useful to know when you're deciding how much to trust it.
You can also push it. Asking "check what the current version is" or "look up what happened this week" is usually enough to tip a borderline question toward searching.
What you actually get
It makes images too
Describe what you want and it draws it, in the same chat, at the same price.
Try the image generatorA tutor that goes slow
It lays out a path, teaches one piece, checks you got it, then keeps going.
Meet the tutorNothing to sign up for
No email, no password, no setup screen. The chat box is the front door.
Why we skipped the signupIt looks things up
Ask about the news, the weather, a price — it searches the web and links what it used.
Ask it something currentFine on your phone
Works on whatever you've got open — phone, laptop, the tab you forgot to close.
Open the chatWhere this actually changes the answer
Anything with a date attached. Today's news, this week's weather, whether a film is out, what a company announced yesterday. This is where an unaided model is confidently wrong most often.
Numbers that move. Prices, exchange rates, scores, current release versions. Worth opening the source on these — a search result can be stale too, and the link is right there.
Things too new or too small to have been memorised. A library released last month, a local business, a niche timetable. Search reaches these; training data generally didn't.
Anything you plan to rely on. Even when the model knows the answer, cited sources give you something to verify against rather than a claim to take on faith. Our guide to fact-checking AI answers covers what's worth checking and what isn't.
Where it won't save you: a search that returns a wrong page produces a wrong answer with a citation attached, which is more convincing rather than less. Sources make checking possible, not unnecessary.
Related reading
- How-to
How to Fact-Check AI Answers and Catch Hallucinations
AI chatbots sound just as confident when they are wrong as when they are right. Here is the verification habit I actually use to catch hallucinations before they bite me.
- How-to
How to Write Better AI Prompts: 9 Techniques That Actually Work
I spent a week paying attention to which prompts got me useful answers and which ones wasted my time. Here are 9 prompt-writing techniques that held up, with copyable examples and before/after comparisons.
- Comparisons
ChatGPT Alternatives: 7 Free AI Chatbots You Can Use Without Signup in 2026
I tested seven free AI chatbots that work without an account - no signup, no credit card. Here is what each one is actually good at, and where they fall short.
Frequently asked questions
How do I make it search the web?
You don't have to do anything — ask your question normally. When it needs current information the assistant searches on its own, and the sources appear under the answer. If you want to nudge it, phrasing like "look up" or "what's the current…" usually does it.
Can I see where an answer came from?
Yes — that's the point of it. Any answer that used a search lists the pages underneath as links you can open. If an answer has no sources listed, it wasn't based on a search.
Is the web search free?
Yes. It's part of the ordinary chat, with no separate tier, credit balance, or upgrade prompt, and no account needed. The site is ad-supported, which is what pays for it.
How current is the information?
As current as the pages it finds, which is normally today. Search results can themselves be out of date, which is exactly why the sources are listed rather than hidden — for anything that matters, open the link and look.
How is this different from Perplexity?
Perplexity is built specifically around search and does that job well. This is a general AI chat that searches when it needs to — so you also get image generation and a step-by-step tutor in the same box, without an account. If deep sourced research is the bulk of what you do, a dedicated research tool is still the better fit.
Does it search every message?
No, and you wouldn't want it to. Most questions — writing help, explanations, coding, brainstorming — are answered better and faster without a lookup. It searches when recency actually matters.