← tickerscout.ai

About Ticker Scout

Updated August 26, 2026

Ticker Scout publishes finished financial statements for public companies, assembled from the reports those companies file with the SEC. Income statement, balance sheet and cash flow, on each company's own fiscal calendar, in actual dollars, with every figure carrying the accession number of the filing it came from. It is free to read, by a person or by a machine, and there is no signup and no API key.

Why this exists

I started out using AI to write my own analysis reports, one company at a time. I wanted to learn the businesses I was looking at properly, and I wanted more out of the public record than a summary. I wanted a real opinion, argued from the filings.

What I got was skimming. Ask a model to look into a company and it reaches for headlines, because headlines are what is easy to reach. So I started pulling the filings myself and cutting them down to what a model could actually digest: the financial figures, and the passages of the 10-K that carry the substance.

That was much harder than I expected. You cannot template the text, because filings genuinely do not follow one shape, and a rule that finds the right section at one company quietly finds the wrong one at the next. The figures are worse. The same line is tagged differently from one filing to the next, a bank and a retailer do not have the same statements to begin with, scale is stated in one place and left implicit in another, currency is not always what you assume, and figures that exist in a company's XBRL are missing from the convenient machine-readable summary the SEC publishes alongside it. Going to the XBRL instead means a slower parse and a different structure again.

I got there eventually. The scripts worked and the reports coming out of them were good. But there were always thousands of edge cases left, and the ones that bothered me most were in the text, where a rule I had written would cut out the one paragraph that mattered and I would never know, because what came back still read fine.

So I changed the approach. Instead of scripts deciding what matters, the cleaning and the preparation are done by AI and then checked back against the filing, so what comes out is what is actually needed, whether that is an analysis report or a fundamentals layer for an agent to work from.

The catch is that preparing context this way is slow and very expensive in tokens. It is not something most people are going to do for every company they are curious about. I had a method that worked for me, so it seemed worth giving everyone the same level of access and sharing the context I was already generating.

None of this is because the data is hard to come by. The SEC publishes all of it, free, and I know that. Availability was never the problem. The problem is what a model does with it once you hand it over.

You can give a model an entire filing and a JSON file full of figures. Unless you are driving a coding agent that can write scripts and take that apart, it will not hold it. Details get dropped, the context bloats, and the answer still comes back sounding confident and authoritative. That is the failure worth worrying about, because it does not look like a failure.

So this is for the investor who wants to go deeper than a summary, who wants to be able to check that what the model said is what the company actually reported, and for anyone handing real decisions to an agent.

How the data is made

Every company on this site is built from its own filings and from nothing else. No data vendor, no pre-tagged dataset, no other website. For a US filer that means the annual report on Form 10-K, the quarterly reports on Form 10-Q, and the current reports on Form 8-K, and it is normally several filings rather than one, because no single filing holds a company's whole record. Each page states how many filings went into it and which forms they were.

The reading and the assembly are done by AI, for the reason above: rules break on the variance between filings and a model does not. What keeps it honest is what happens next. Every pass is audited back against the filings it came from before anything is published, and each figure carries the accession number of the filing it came from, so any line on this site can be checked against the original at sec.gov in about a minute.

The published record is also audited by hand. I periodically spot check figures against the filings they cite and read the synthesis for inaccuracies. Automated checks catch the mechanical problems, the wrong scale, the missing period, the figure that does not reconcile. I audit the AI's work so the workflow itself improves: what I find becomes a new check or a new guardrail, so a problem caught once stops coming back.

Where the filings do not support a figure, the record says so and leaves it out. It does not report a zero and it does not estimate. Each company's file lists those gaps by name, along with anything unusual about how a number was arrived at, such as a quarterly cash flow derived by subtraction because the company never published a standalone one. AI can still make mistakes, and so can I. The filing is the authority. This site is a route to it.

What this is not

Expanding

Coverage is large US filers today and it is growing quickly. Bigger companies and the ones that have filed most recently are prioritized, so the record stays current where it is read most. See every covered company for what is actually here right now, rather than what this page claims.

Coverage will keep expanding, the process behind it is refined continuously, and what each company publishes here will grow as I find things worth adding. I am an active investor using this data myself, for real research and for agentic trading, so what the site provides is driven by what actually turns out to be needed rather than by what is easy to generate. I am committed to publishing the best quality data I can, and to fixing it in public when it is wrong.

Using it

The data is free to use and free to cite, with attribution to Ticker Scout (tickerscout.ai). If you are pointing an AI agent at it, the files it wants are described in llms.txt, and the same data is available as tools through the MCP server. The Terms of Use cover the rest.

Who publishes it

Ticker Scout is published by Ticker Scout, LLC, established 2025. It is built and operated independently.

Questions, corrections and anything else go to info@tickerscout.ai, which reaches me. If you find a figure on this site that does not match the filing it cites, I would genuinely like to know, and that is the message that gets answered first.

Ticker Scout