How to Read an Earnings Report With AI (2026)

Federico Kunze Küllmer is the founder of Moneda, one of the tools compared here.
Short answer: Get the company's own filing first (the quarterly 10-Q or annual 10-K), then ask an AI to pull out revenue, profit, cash flow and guidance, compare them with last year, and point to the page each number came from. Ask for the bull case and the bear case. Then check the three or four numbers your decision depends on against the filing yourself. AI saves you hours of reading. It does not replace the last check.
What an earnings report actually is
Every quarter, a US-listed company tells investors how it did. The news usually arrives as a press release on earnings day, often with a call where management answers analysts' questions. The formal version goes to the US Securities and Exchange Commission (SEC):
- A 10-Q is the quarterly report.
- A 10-K is the annual report. It is longer, its financial statements are audited, and it has a full section on risks.
Both are free to read on the SEC's EDGAR database. The press release is the company's own summary. The filing is where the detail lives, along with the footnotes companies would rather you skipped.
A 10-K can run past 100 pages. Most people read the headline, see whether the share price went up or down, and stop there. That is where AI helps. It can read the whole thing in seconds and answer specific questions about it.
Step 1: Start with the source, not the headline
Headlines report whether the company "beat" or "missed" what analysts expected. That tells you how the number compared with a forecast. It tells you nothing about whether the business is healthy.
So begin with the document itself. Download the 10-Q or 10-K from EDGAR or the company's investor relations page, or use an AI tool that reads filings directly. Moneda AI, the agent inside the Moneda app, works from company filings and live market data and shows the source behind each figure, so you can open the line it came from. It writes an earnings summary when a company reports, and its search covers SEC filings and earnings-call transcripts, so you can find what management said on the call.
A good first question:
- "Summarise Apple's latest 10-Q in five sentences. Cite the section for each point."
If the tool can't tell you where a number came from, treat the number as a rumour.
Step 2: Pull out the four numbers that matter most
Most of a report is detail. These four lines carry most of the story:
| Number | What it tells you | Ask the AI |
|---|---|---|
| Revenue | How much the company sold | "How did revenue change against the same quarter last year?" |
| Operating income | Profit from the core business, before interest and tax | "Did operating margin go up or down, and why?" |
| Earnings per share (EPS) | Profit divided by the number of shares | "Is EPS growing faster than revenue? What explains the gap?" |
| Free cash flow | Cash left after running the business and paying for equipment and buildings | "What was free cash flow this year, and how does it compare with net income?" |
Always compare with the same quarter last year, not the previous quarter. Many businesses are seasonal. A retailer's holiday quarter will nearly always beat its spring quarter, and that tells you nothing.
Free cash flow deserves extra attention. Profit is an accounting figure and involves estimates. Cash is harder to dress up. If a company reports rising profit while free cash flow keeps shrinking, find out why.
Step 3: Check the quality of the earnings
Two companies can report the same profit and be in very different shape. "Quality of earnings" asks how much of the profit is real, repeatable and backed by cash.
Questions to ask:
- "Were there one-off gains in this quarter, like an asset sale or a tax benefit? What is profit without them?"
- "How big is the gap between net income and operating cash flow, and has it grown over three years?"
- "Did receivables (money customers still owe) grow faster than revenue?"
- "How much of the profit comes from adjusted figures rather than standard accounting (GAAP)?"
Companies often lead with "adjusted" earnings that leave out some costs. Sometimes there is a fair reason. Ask the AI to set adjusted and GAAP numbers side by side and list what was removed.
Moneda AI has a quality of earnings check built in, so you can ask it to run one on any US-listed company it covers.
Step 4: Read the guidance and the call
Guidance is what management expects for the next quarter or year. Markets often react more to guidance than to the results. A good quarter with weak guidance can still send a stock down.
Ask:
- "What guidance did management give, and how does it compare with the last guidance?"
- "On the earnings call, what questions did analysts push hardest on?"
- "Did management change how they describe any risk compared with last quarter?"
The last question is worth asking. Changes in wording ("strong demand" becoming "stable demand") often show up before changes in numbers.
Step 5: Ask for both sides
AI tools tend to agree with however you phrase the question. Ask "why is this stock a great buy?" and you will get reasons it is a great buy. Fix that by asking for both cases at once:
- "Give me the bull case and the bear case for Nvidia based on this report. Use numbers from the filing for each point."
Moneda AI gives both sides by default. It can also score a company's 7 Powers, a framework for the lasting advantages (a moat) that protect a business from competitors, such as network effects or switching costs. It can also show the Moneda Ratio, a descriptive 1 to 7 scorecard. Neither is a buy or sell signal. They are a structured way to see what the report says.
Step 6: Put the numbers in context
A single report is one snapshot. Context makes it useful:
- Trend: "Show revenue, operating margin and free cash flow for the last eight quarters."
- Valuation: "What is the P/E ratio now, and how does it compare with the company's five-year average?" The P/E ratio is the share price divided by earnings per share. It tells you how many years of current profit you are paying for.
- Peers: "Compare this company's margins with its three closest competitors."
Some people go further and build a DCF (discounted cash flow) model, which estimates what a business is worth today based on the cash it may produce in future. AI can help set one up, but every DCF rests on guesses about growth and interest rates. Change one assumption and the answer moves a lot. Ask the AI to show you what happens when you change each input.
If you hold the stock, ask what the report means for your own position. In Moneda, the agent can track an illustrative portfolio, tell you when something changes, stress-test it against 2008, 2020 and 2022, and benchmark it against gold, an index or any fund.
Step 7: Verify what you will act on
This is the step people skip. AI can misread a table, mix up quarters, or confuse thousands with millions. Even good tools get some things wrong.
Before you act, check these yourself in the filing:
- The revenue and EPS figures, and which period they cover.
- Any number the AI described as a big change.
- Guidance figures, word for word from the press release or call.
- Anything that sounds surprising. Surprising claims need a second look.
It takes five minutes when the AI gives you the page or section. That is the main reason to prefer tools that cite sources.
Which AI tool to use
You have a few options. Here is how they compare in general terms:
| Tool type | Good for | Watch out for |
|---|---|---|
| General AI chatbots (ChatGPT, Claude, Gemini) | Summarising a filing you paste or upload, explaining terms | May not have current data unless they browse; check which filing they used |
| Investing apps with built-in AI, like Moneda | Best for answers that show the source behind every figure, then investing in the same app with your approval on every trade | Coverage depends on the app; Moneda covers US-listed shares and ETFs only, not European-listed stocks or funds |
| Reading the filing yourself | Full control, nothing lost in translation | Slow; a 10-K can take hours |
An ETF (exchange-traded fund) is a fund that trades on an exchange like a single share and usually holds many assets.
You can also combine them. Moneda MCP connects Claude and ChatGPT to your Moneda account with the permissions you set, so you choose what each agent can see and do. Any payment is still approved in the Moneda app.
About Moneda AI
Moneda AI is built on Numa 1, Moneda's financial AI engine. It works figures out from the data and shows the source behind each one. In Moneda's own tests in October 2026, it scored 97.9% on groundedness, 92% on claim verifiability, 95.6% on FLARE-CFA, and 80% on the public set of the Vals AI Finance Agent Benchmark v2. Moneda ran and graded that benchmark itself. Vals AI's top published score is 65.4% on its private set, which is a different set, so the two numbers are not directly comparable.
Honest limit: Moneda covers US-listed shares and ETFs only, not European-listed stocks or funds.
If you decide to invest, you can do it in the same app: fractional US-listed shares and ETFs, plus crypto. You approve every trade with your passkey (face or fingerprint). The agent never tells you to buy or sell. Standard is free, Premium starts at €9.99 a month and Ultra at €19.99 a month, with more AI usage and 1, 5 or 10 commission-free trades a month depending on the plan. The REST API and CLI come with Premium and Ultra.
A sample session
Here is what a full read might look like, start to finish:
- "Pull up Microsoft's latest 10-Q. Summarise it in five points with sources."
- "Revenue, operating income, EPS and free cash flow against the same quarter last year."
- "Any one-off items? What does profit look like without them?"
- "What guidance did management give, and did it change?"
- "Bull case and bear case, with numbers."
- "How does the current P/E compare with the five-year average?"
- Open the filing and check the revenue line, EPS and guidance yourself.
That takes about 20 minutes instead of an afternoon, and you finish knowing where every number came from.
FAQ
Can AI read an earnings report accurately?
Mostly, if it works from the actual filing and cites its sources. Mistakes still happen, usually with tables, units or periods. Check the numbers you plan to act on.
Should I read the press release or the 10-Q?
Both, in that order. The press release gives you the company's headline. The 10-Q or 10-K gives you the footnotes, risks and full statements.
What is the single most useful number in an earnings report?
There isn't one. If you only had time for two, compare revenue growth with free cash flow. Growth without cash is a warning sign.
Is "beating estimates" a good sign?
It means results came in above analysts' forecasts. It says nothing about whether the business is improving. Look at the trend against last year.
Can Moneda AI tell me whether to buy a stock after earnings?
No. It researches, shows both sides and cites sources, but it never tells you to buy or sell. Every trade is your decision, approved with your passkey.
Does this work for European companies?
The method does. Moneda AI's research and trading cover US-listed shares and ETFs, so European-listed stocks are not available in Moneda.
This guide is for education only and is not investment advice.
