How to Read a 10-K With AI (2026)

Federico Kunze Küllmer is the founder of Moneda, one of the tools compared here.

Short answer: Download the 10-K for free from the SEC's EDGAR site, then read it in four passes with an AI: the business (Item 1), the risks (Item 1A), management's own explanation of the year (Item 7) and the audited financial statements (Item 8). Ask the AI to quote the page or table behind every number it gives you. Then check the three or four figures your decision rests on against the filing yourself. The AI saves you hours of reading. It does not replace the last check.

A 10-K is the annual report that every US-listed company files with the Securities and Exchange Commission (SEC). It is long. A large company's 10-K often runs past 100 pages, and most of it is legal boilerplate. The useful parts are buried in the middle.

That is exactly the kind of document AI is good at. It can read the whole thing in seconds, find the sections you care about and pull numbers out of tables. It can also misread a table or invent a figure that sounds right. So the method below uses AI for speed and keeps you in charge of the facts.

What is in a 10-K

The SEC's own investor guide splits the 10-K into numbered items. You only need a handful of them:

A few terms you will meet:

• Income statement: revenue, costs and profit for the year.
• Balance sheet: what the company owns and owes on the last day of the year.
• Free cash flow: cash from operations minus spending on equipment and buildings. It is the cash left over that the company could use to pay debt, buy back shares or pay dividends.
• P/E ratio (price to earnings): the share price divided by profit per share. It tells you how much you pay for each unit of profit. It is not in the 10-K itself, because the price changes daily.

Step 1: Get the right filing

Go to the SEC's EDGAR database and search the company name or ticker. Filter by form type "10-K" and pick the most recent one. Check the fiscal year on the cover page. Many companies end their year in a month other than December, and an AI will happily mix up two years if you hand it the wrong file.

If you want to compare years, download the last two or three 10-Ks. Each one also contains the prior year's figures, so two filings give you three years of numbers.

Then either upload the file to your AI tool or paste the link. General chat assistants like ChatGPT, Claude and Gemini accept uploaded documents. A research tool like Moneda AI pulls data straight from SEC filings and earnings-call transcripts, which you can search, along with live market data, so you just name the company. It also writes earnings summaries when a company reports.

Step 2: Read the business section first

Start with Item 1. You want to come out of it able to explain the company in two sentences.

Questions to ask:

• "In plain English, what does this company sell and who pays for it? Quote the sentences you used."
• "Break revenue down by segment and by region for the last three years. Give the table and page number."
• "Who are the main customers? Does any single customer account for more than 10% of revenue?"
• "What does the company say about its competitors?"

That last question leads into the idea of a moat: a lasting advantage that keeps competitors from taking the company's profits, such as a strong brand, a network that gets better as more people use it, or costs nobody else can match. The 10-K will not use that word. Ask the AI to argue both ways: "What in this filing suggests the company has a lasting advantage, and what suggests it does not?"

Moneda AI does a structured version of this. It scores a company's 7 Powers, a framework of seven kinds of lasting advantage, and shows the evidence for each score.

Step 3: Read the risks, but skim the boilerplate

Item 1A is often the longest section. Much of it is generic ("our stock price may be volatile"). The useful risks are the specific ones.

Questions to ask:

• "List the risk factors that are specific to this company, not ones every company would write. Skip the generic ones."
• "Which risk factors are new or changed compared with last year's 10-K?" (You need both filings for this.)
• "Does the company depend on a single supplier, country or product? Quote the relevant passage."
• "Is there any ongoing lawsuit or government investigation mentioned? Where?"

The year on year comparison is the most valuable question here. Lawyers add a risk factor when something has actually changed.

Step 4: Let management explain the year, then check them

Item 7, the MD&A, is management telling its own story. It is useful and it is also the most flattering part of the report.

Questions to ask:

• "Why did revenue and operating profit change this year, according to management? Quote them."
• "What reasons does management give for any margin changes?"
• "Does management use any adjusted or non-standard profit measures? How do they differ from the reported figures?"
• "What does the company say about spending plans, debt and cash for next year?"

Pay attention to adjusted figures. Companies often present "adjusted EBITDA" or "adjusted earnings" that leave out certain costs. That is allowed, but ask the AI to put the adjusted and the reported numbers side by side, so you can see what was left out.

Step 5: Go through the financial statements

Item 8 has the audited numbers. This is where AI saves the most time, and where it makes the most mistakes, because tables in filings are dense and often split across pages.

Questions to ask:

• "Build a three year table of revenue, gross profit, operating income, net income and diluted earnings per share. Cite the page for each figure."
• "Calculate free cash flow for each year: cash from operations minus capital expenditures. Show both inputs."
• "How much debt does the company have, and when does it come due?"
• "Compare net income with cash from operations over three years. If profit is growing faster than cash, what explains the gap?"

That last question is about quality of earnings: whether reported profit turns into real cash. A company whose profit keeps rising while its cash flow does not is worth a closer look. Moneda AI runs a quality of earnings check as part of its research and works out each figure from the data, with the source shown next to it.

Some readers also want a DCF (discounted cash flow), a valuation that estimates future cash flows and converts them into a value today. An AI can build one from 10-K numbers, but every DCF rests on guesses about growth and interest rates. Ask the AI to list its assumptions, and change them to see how much the answer moves.

Step 6: Ask for both sides

By now you will have a view. The risk is that you only notice the evidence that agrees with it.

Questions to ask:

• "Give me the strongest bull case and the strongest bear case based only on this 10-K."
• "What would have to be true for the bear case to be right?"
• "What is the one thing in this filing that a skeptic would focus on?"

Moneda AI gives both sides of the story by default, a bull and a bear case for each company it researches.

Step 7: Verify the numbers you will rely on

This step matters most. AI tools can get a figure wrong while sounding certain. Decide which three or four numbers your view really rests on, and check each one in the filing yourself.

What to verify yourself:

1. Revenue and net income for the latest year, from the income statement in Item 8.
2. Cash from operations and capital expenditures, from the cash flow statement. Redo the free cash flow subtraction.
3. Total debt, from the balance sheet and the debt note.
4. The fiscal year, from the cover page, so you know you are comparing the right periods.
5. Any quote the AI attributes to management. Search the filing for the exact words.

If a number does not match, ask the AI where it got it. Often it took a figure from the wrong year's column or from a segment table instead of the total.

A tool that shows its source for each figure makes this faster. For what it is worth, Moneda's own tests in October 2026 measured groundedness (whether answers stick to the sources) at 97.9% and claim verifiability at 92%. Those are Moneda's own tests, not independent ones, and even 97.9% is not 100%. Check the numbers that matter.

Choosing a tool

You can do all of this with any capable AI. The difference is how much setup you do yourself, and whether research and investing happen in the same place. Most tools do one or the other. Moneda is built for the case where you want answers that show the source behind every figure, then invest in the same app.

If you already use Claude or ChatGPT, Moneda MCP connects them to your Moneda account with the permissions you choose. Moneda has a free Standard plan. Premium starts at €9.99 a month and Ultra at €19.99 a month, with more AI usage.

FAQ

Where do I find a company's 10-K?

On the SEC's EDGAR database, free. Search the company, filter by form type 10-K. Most companies also post it on their investor relations page.

Can AI read the whole 10-K at once?

Most current AI tools can take a full filing. Answers are better when you ask about one section at a time and ask for the page behind each figure.

Which parts of a 10-K matter most?

Item 1 (business), Item 1A (risk factors), Item 7 (management's discussion) and Item 8 (financial statements). The rest is mostly legal and governance detail.

Is a 10-K the same as the annual report to shareholders?

Not always. The SEC notes that some companies send a separate, glossier annual report. The 10-K is the official filing with the full required detail.

What about foreign companies or ETFs?

Foreign companies listed in the US usually file other forms, such as the 20-F. An ETF (exchange-traded fund, a basket of investments that trades like a single share) publishes a prospectus and holdings reports rather than a 10-K.

Should I trust an AI's summary of a 10-K?

Use it to find things fast, and check the figures your decision depends on. Ask for sources every time.

This guide is for information only and is not investment advice.