Why Is This Stock Up Today? Ask a Free AI Chatbot

You know the moment. You glance at your phone, a stock you own — or wish you owned — is down 7%, and you have no idea why. So you type "why is this stock down today" into a search engine and get the worst page on the internet: five sites with the same auto-generated paragraph ("shares fell amid market volatility"), a forum thread from 2019, and a "news" article that turns out to be a press release about a different company.

I've spent the last few weeks doing something else instead, and it's stuck. I ask an AI chatbot that can actually search. Not the version that answers from memory — a model that runs a real web search, reads what's out there right now, and hands back a short explanation with the sources it read attached. It's the difference between asking a friend who follows the market and asking a friend who last read the news in 2024.

This post is about how I use a free AI chatbot for stock market questions: the one-click version, the prompts that work, and — because this is money — the exact spots where the AI will confidently tell you something false.

The one-click version: trending stocks on the home page

We just added something to the Smillee AI home page that turns the whole thing into a single tap. Under the message box, next to the row of trending topics, there's now a row of trending stocks — the tickers people are actually watching right now. Each chip shows the ticker, the company name, the percent move for the session, and a green or red arrow so you can read the row at a glance.

Click one, and it asks the question for you:

Why is Nvidia (NVDA) stock moving today?

The answer comes back streamed, in plain English, with numbered citations you can click through to the actual articles it read. Under it you'll get three suggested follow-ups — usually something like "what did analysts say about the guidance" or "how does this compare to last quarter" — so you can keep pulling the thread without typing.

Two honest notes about what's behind that row, because I'd want to know:

• It's a fresh search, not a live quote feed. The list and the percent moves come from a web lookup that refreshes every few minutes. That's plenty for "what's everyone talking about," but it's not a trading terminal. If you're about to act on a price, check your broker.
• It's the most-talked-about stocks, not the biggest movers. A boring mega-cap that's down 0.4% can be on the list because it reported earnings that morning. A tiny stock up 40% might not be, because nobody's writing about it. That's a feature — it's the stuff there are actually good explanations for.

The row is free and needs no signup, same as the rest of the chat. Even if you never type a stock question yourself, it's a nice two-second read on what the market is chewing on today.

Why an AI chatbot with search beats googling stock market news

A plain chatbot — no search, no tools — is a very good student who stopped reading the news months ago. Ask it why a stock moved today and it has two bad options: admit it doesn't know, or invent a plausible reason. Models are famously bad at picking the first one. (I wrote about that whole failure mode in how to fact-check AI answers.)

Grounding changes the job. The model runs a search first, reads a handful of current pages, and writes its answer from those — and the citations under the answer are anchored to the specific sentences they support, not sprinkled on afterwards to look credible. That gives you something you can't get from a search results page or from a memory-only chatbot: a one-paragraph explanation and a receipt for every claim in it.

Which is exactly what you want for a "why is this stock up today" question. The answer is almost always some combination of a small number of things — earnings, guidance, an analyst note, a product event, a lawsuit, a macro headline, or a big move in a related stock — and a search-grounded model is very good at finding out which one it was today.

Seven stock market prompts I actually use

The chip asks the first one for you. The rest are for when you want to go deeper. Copy them, swap in the ticker. (If you want to tune any of them, the habits in how to write better AI prompts all apply.)

1. The plain "why"

Why is [company] ([TICKER]) stock moving today? Give me the main reason in two sentences, then any secondary factors. Cite where each claim comes from.

What a good answer looks like: a specific catalyst ("cut full-year guidance on the earnings call last night") with a source next to it. What a bad answer looks like: "shares fell amid broader market weakness and investor concerns." If you get the second kind, the model didn't find anything real — either there's genuinely no story, or the move is too fresh for the news to have caught up. Ask again in an hour.

2. Explain the earnings report like I'm new

[Company] just reported earnings. Explain what happened like I've never read an earnings report: what did they make, what did Wall Street expect, what did the company say about the future, and why did the stock react the way it did?

This is the one I'd hand to anyone who finds finance news impenetrable. The "what did Wall Street expect" part matters — a company can grow revenue 30% and drop 10%, and that's the sentence that explains the paradox.

3. Decode the jargon

In the news about [TICKER] today, people keep saying "[guidance / beat / miss / short interest / ex-dividend / dilution]." Explain what that means in plain English and why it matters for this specific stock right now.

The "for this specific stock right now" clause is what turns a dictionary definition into something useful. Any chatbot can define short interest. A grounded one can tell you it's unusually high on the stock you're asking about.

4. Is this a one-day story or a trend?

Put today's move in [TICKER] in context: how has the stock done over the last month and year, and is today's news a new development or a continuation of something already going on?

Single-day moves lie. A 5% drop on a stock that's up 80% this year is a very different story from a 5% drop on one that's been sliding since spring. Asking for the context up front keeps me from over-reacting to a headline.

5. Compare two names in the same news cycle

[TICKER A] and [TICKER B] are both moving today. Are they moving for the same reason? Compare the two in a short table: today's move, the stated reason, and whether it's company-specific or sector-wide.

Great on days when a whole sector lurches at once. It'll usually tell you that one company actually reported news and the others are moving in sympathy — which is a real, useful distinction.

6. Steelman both sides

Based on today's news about [TICKER], give me the strongest honest case a bull would make and the strongest honest case a bear would make. No hedging, no "it depends" — commit to each side in three bullets, then tell me what a long-term holder would actually need to watch.

I like this because it forces the model out of its natural mush. You're not asking it to predict anything — just to lay out the arguments that are already out there so you can think.

7. What's on the calendar

What scheduled events could move [TICKER] in the next two weeks — earnings date, dividend dates, product events, court dates, or macro announcements that hit its sector? List each with the date and the source.

Extremely useful. Also the single most dangerous prompt on this list, which brings me to the part that matters.

Where an AI chatbot will lie to you about money

Everything I said in the fact-checking guide applies double here, because the wrong answers in finance cost actual dollars. These are the specific failure modes I've hit:

Stale prices dressed up as current. The model finds an article from Tuesday, and it's now Thursday, and it tells you the stock is "trading around $142" with total confidence. The price is a fact about a moment, and the moment has passed. Never take a price from a chatbot. Take a reason from it, and get the price from your broker.

Invented specifics. A fabricated analyst quote, a price target nobody set, an "EPS of $2.14" when the real figure was $1.94. These are the classic hallucination: the shape of a fact without the fact. The tell is a precise number or a named person with no citation next to it. If it isn't sourced, it didn't happen until you've checked.

Wrong quarter, wrong year. Earnings coverage is repetitive, so the model will occasionally blend last quarter's story into this quarter's. If the explanation seems to describe news you remember from three months ago, it might be exactly that.

Confusing similar companies. Tickers that look alike, companies with the same first word, a parent and a spun-off subsidiary. Read the company name in the answer, not just the ticker, and make sure they match.

Dates on the calendar prompt. Prompt #7 is a magnet for this. Earnings dates get "estimated" by data sites and then reported as confirmed. Verify any date you'd plan around against the company's own investor-relations page — the citation should take you there, and if it doesn't, that's your answer.

The rule I've landed on is simple: the AI is for the story, not the numbers. Let it tell you what happened and why. Any figure you'd actually act on — a price, a date, a percentage — you verify at the source before it touches a decision.

What this is not

It's not investment advice, and it's not a good idea to ask for any. "Should I buy [TICKER]?" gets you one of two answers: a responsible refusal, or a confident opinion from something that doesn't know your income, your timeline, your tax situation, or what else you own. Neither is worth much.

What it is good for is closing the gap between "a number on my screen changed" and "I understand what happened." That gap used to cost twenty minutes of reading bad websites. Now it costs one click and thirty seconds of reading — plus the sixty seconds of checking that I hope this post has made a habit.

Try it

Open Smillee AI, look under the message box for the trending stocks row, and click whichever ticker you're curious about. Read the answer, click a citation or two, and hit one of the follow-up chips. If you want to go further, grab a prompt from the list above and swap in the name. And when it hands you a number, do the thing: check it.

— Maya