FOR INVESTORS WHO CHOOSE THEIR OWN STOCKS

Research you can check.

AI-powered research on SEC filings. Screen over 5,000 U.S.-listed companies, let AI read the filings of the top-ranked ones that pass, and check what it cites against the filings yourself.

Sign up See how it works

FREE TO BROWSE · PLANS FROM $29/MO

Read the text in the video

Everything is in the SEC filings. Nobody has time to read them all.

Describe what you're looking for. Or start from a built-in framework.

Screen over 5,000 U.S.-listed companies. You confirm the estimated cost before it runs.

AI reads the filings of the top-ranked companies that pass. Citations name the filing section, so you can check them.

It names what the filings don't say.

Less noise. More evidence. Better research.

AI-generated research. RockTurners provides research and analysis for educational purposes only. Not investment advice. Consult a licensed financial professional before making investment decisions.

Educational research only. Not investment advice. Recreated screens; fictitious companies, invented figures.

WHAT PROBLEM DOES IT SOLVE?

Reading the filings eats evenings and weekends.

The numbers, the risks, what the executives get paid: it's all in the SEC filings. An annual report (for most U.S. companies, the 10-K) can run past a hundred pages, so you only ever get to a few companies. A numbers screen is fast, but it doesn't read the filings.

READING THE FILINGS YOURSELF

✓The whole story: the business, its risks, and how management explains the results.

✕Hours per company, so you get to a few.

A NUMBERS-ONLY SCREENER

✓Fast, across thousands of companies.

✕It ranks reported figures. It doesn't read what the filings say.

ROCKTURNERS

✓A screen across over 5,000 U.S.-listed companies.

✓Then an AI read of the top-ranked companies' filings, with the sections cited.

Limits: U.S.-listed companies only, and the filing read covers the top-ranked companies that pass, not every listing.

WHAT IS ROCKTURNERS?

An AI stock screener that also reads the filings

RockTurners is an AI research platform for investors who research individual companies. It works in four steps.

  1. 01 · DESCRIBE

    Say what you're looking for

    Start from one of the 10 built-in frameworks, or describe the companies you're looking for in plain English in the Strategy Lab. It drafts a research framework, and you review the rules before anything runs.

  2. 02 · SCREEN

    Screen over 5,000 companies

    The framework screens U.S.-listed companies on reported figures and market prices, and ranks the ones that pass. You confirm a run's estimated cost before it starts.

  3. 03 · READ

    AI reads the filings

    Its AI reads the relevant filing sections, fetched from SEC EDGAR, for the top-ranked companies that pass.

  4. 04 · CITE

    See the evidence and the gaps

    Citations name the filing section, so you can check them. When something's missing, the result names the gap, so you know where to dig in.

Ten-Year EPS Growth, Low-Debt Value Screen
SCANNING_
6/25 read
■ stop
SCREENED
5,348
CANDIDATES
426
DEEP-READ
6/25
ELAPSED
01:29
ETA
~9m 33s left
ACTIVITY
▸ stage 1 · quant screen (5,348 filers · cached SEC EDGAR XBRL)
NOPECO✗ filter miss
FILLCO✗ filter miss
✓ 426 passed raw filters → 25 advance to deep-read
▸ stage 2 · ai deep-read (targeted filing sections → model)
EXAMPL◆ GemstoneTen fiscal years of filed EPS growth above the 10% floor, with debt inside the limit
DEMOCO· SedimentEPS growth slowed to low single digits in the latest two 10-Ks
SAMPLE◈ GeodeMargins improved on pricing; the multi-year EPS record is incomplete
MOCKUP· SedimentNet debt rose after an acquisition disclosed in the latest 10-K
TESTCO◈ GeodeEPS growth meets the screen; recent gains tie to one contract
▸ FAUXCO · Fauxco Foods Inc. — deep read in progress …
>
FOUND SO FAR
◆ 1
GEMSTONE
◈ 2
GEODE
· 2
SEDIMENT
EXAMPLExample Industries Corp◆ Gemstone
↳ Ten fiscal years of filed EPS growth above the 10% floor, with long-term debt inside the framework's limit.
SAMPLESample Company, Inc.◈ Geode
↳ Margins improved on pricing actions, but the filings provided don't cover the full ten-year EPS record.
TESTCOTestco Holdings◈ Geode
↳ EPS growth meets the screen; the latest 10-K ties recent gains to one customer contract.
What steps 02 and 03 look like in the app: the screen runs first, then the AI's verdicts stream in. App recreation; fictitious companies, invented figures.
GEMSTONEThe filings support the framework's criteria.
GEODESome criteria hold; others can't be verified from the filings.
SEDIMENTCriteria not met. Kept as a turned-over stone.

A verdict describes fit, never an action to take. When the filings can't support a verdict, the result says Not scored instead of guessing. See how it works, step by step →

WHY AI MAKES THE RESEARCH BETTER

AI does the reading. The method makes it checkable.

Reading is the slowest part of research, and it's the part an AI model does quickly. RockTurners puts the model inside a fixed method, so every read follows the same rules, and citations name the filing section they rely on.

READS THE FILINGS

It reads what a numbers screen can't

The business, its risks, how management explains the results: the AI reads the filing sections each framework declares, for each top-ranked company that passes the screen.

ONE METHOD

Every company gets the same method

One documented framework, with the same prompt and rubric for every company it reads, so companies are compared on the same terms.

SHOWS ITS SOURCES

It shows where findings came from

Citations name the filing section they rely on, each is checked against the sections the model was given, and each filing opens on SEC.gov from the result.

ADMITS GAPS

It says what it doesn't know

What a filing doesn't disclose is flagged as unknown instead of filled in. When the filings can't support a verdict, the result says Not scored.

WHY SHOULD I CARE?

Less noise. More evidence. Better research.

Spend your evenings on the companies that fit. With the first read done, your time goes to the companies whose filings fit the framework, and to the questions those filings leave open.

01

Start with a shorter list

A screen of thousands narrowed to the top-ranked companies that pass, each with a first read of its filings done.

02

Know which section to read

Red flags sit next to the sections they came from, so your own reading starts where the evidence is.

03

Know where to dig in

The gaps a read names are the questions the filings left open, and where your own research starts.

It doesn't tell you what to do. It shows you what the filings say, and where they say it.

CHECK IT YOURSELF

Each verdict shows its work

A verdict isn't a number. It's a structured read of the filings, in four parts you can check:

01
What the filings show
A plain-English read of the filings, measured against what the framework checks.
02
What the framework checks
A score for each thing the framework checks. Together they decide where the verdict lands.
03
Red flags, beside the evidence
Red flags the read found in the filings, next to the sections it cited.
04
Unknowns, kept honest
What the filings don't disclose is flagged as insufficient data instead of filled in.

Each filing a read used opens on SEC.gov from the result, so you can read the section yourself.

Before a scan starts, you confirm its estimated cost.

Dissect a full sample verdict
SAMPLE GEMSTONE Illustrative sample
Sample Company, Inc. (fictitious) · Quality-Value Rank
CONFIDENCE
High
SCORE
14 / 15
ASSESSMENT CHART THE NUMBERS LATEST NEWS MY NOTES

Sample Company's rank is supported by recurring operating earnings and current financial filings. The 10-K filed March 3, 2026 (within the 14-month window) shows net income from continuing operations of $412.8 million for FY2025, with stable, recurring operating performance across its two reporting segments. Operating expenses grew proportionally at 4.1%, with no material one-time items identified in the provided MD&A or financial notes. The data integrity, earnings recurrence, and accounting quality scores all meet or exceed the minimum threshold of 3, with no disqualifiers triggered.

Evidence cited, checked against the filings provided
Item 7, MD&AOperating expenses grew in line with revenue, with no material one-time items.
Framework scores
Insufficient data
Enterprise value components not provided in source materials
Return on capital metric not provided in source materials
▍▎▊ What the numbers show · 12 → AINOT ADVICE

4 LENSES · 10 FRAMEWORKS

Published frameworks, applied identically

Each framework is a fixed screen plus a fixed evaluation prompt, in the tradition of published value- and growth-investing research. Every subscriber who runs a framework gets output from the identical criteria.

The four lenses and the frameworks under each
Lens Focus Frameworks
01 · VALUE

Cheapness with a margin of safety

Four ways of asking the same question — is this business worth clearly more than it trades for? — from statistical net-nets to durable-moat franchises.
  • Deep Value (Net-Net)
  • Quality-Value Rank
  • Moat Value: Pricing Power
  • Moat Value: Scale Economics
02 · GROWTH

Growth priced with discipline

Three ways to hold growth accountable: classify before comparing, insist the price stay proportionate to the growth rate, and require the leadership show up in the reported numbers.
  • Growth at a Reasonable Price
  • Growth Leaders
  • Momentum Leaders
03 · QUALITY

Durable returns on capital

Two depths of the same question — does this business earn high returns on capital, and can it keep doing so? — one quantitative, one read fifteen points deep in the filings.
  • Quality Compounder
  • 15-Point Qualitative
04 · INCOME

Payout streaks that can continue

One discipline, applied strictly: dividend growth streaks matter only when the coverage, cash flow, and balance sheet behind them currently support the streak.
  • Dividend Compounders

Framework names describe the method. Practitioner names appear only as "in the tradition of" attribution on each lens page. Browse all four lenses →

WHY NOT JUST ASK AN AI CHAT?

Change the model, not the framework.

RockTurners runs on the same kinds of AI models as popular AI chat apps. In a chat, the method is whatever you set up in the conversation. In RockTurners it's written down once, and every company is read against the same framework, whichever model you choose.

A general-purpose AI chat is a good tool for a question about one company you've already chosen. RockTurners is for finding which companies are worth the question.

Describes general-purpose AI chat apps as a category, as of November 2026. Features vary by product.

WHO IT'S FOR · WHAT IT ISN'T

A research instrument, nothing else

BUILT FOR

Investors and traders who want to know the business behind a ticker, and who research individual companies before they decide.

ALSO USED FOR

Independent analysis, investment-club research, and writing that cites filings.

NOT FOR

Timing entries or exits, intraday signals or options pricing; ETFs, funds and listings outside U.S. exchanges; stock tips or price targets.

ROCKTURNERS IS
→An AI research platform that applies published investing frameworks identically to every company in its screen of over 5,000 U.S.-listed companies.
→Grounded in SEC filings: citations name the filing section they rely on.
→Honest about gaps — what a filing doesn't disclose is flagged as unknown instead of filled in.
ROCKTURNERS IS NOT
✕Investment advice — verdicts describe framework fit (how well a company's filings satisfied a framework's criteria), never an action to take.
✕Personalized — nothing about who you are is an input.
✕Predictive — no targets, no forecasts, no claims about future returns.

Find the companies worth a deeper look.

Try it on a company you know well, and check each cited section against the filing. Free to browse; paid plans include a 14-day first-time trial.

Sign up See pricing