How to Verify a Stock Thesis with AI, Claim by Claim

To verify a stock thesis with AI, split the thesis into separate factual claims, ask the AI to find evidence for and against each claim in the company's filings and transcripts, and require a source for every answer. Then make it argue the bear case and list what it could not find. What you end up with is a map of which parts of your idea are supported, which are contested and which are assumptions.

The aim is not to get the AI to agree with you. Language models are inclined to be agreeable, and a vague request such as "is this a good investment?" tends to produce a balanced-sounding essay that confirms whatever you seemed to want. The method below removes that room.

Why verify a thesis before buying?

Because the cheapest time to find a flaw is before the money is committed. After you buy, you have a financial and emotional stake, and the research tends to become a search for reassurance.

Verification also makes later monitoring possible. If you know which claims your thesis rests on and where the evidence sits, you know exactly what to watch in the next filing. Our investment thesis tracker guide explains how to record that.

Step 1: How do you break a thesis into checkable claims?

Write the thesis as you would explain it to a friend, then pull out each statement that could be true or false.

Take an example thesis: "This software company is a buy because customers keep spending more, margins are expanding, and the new AI product will add a large revenue stream."

That breaks into:

Separating facts from forecasts is the most useful thing in the exercise. Facts can be verified. Forecasts can only be tested for plausibility against evidence.

Step 2: How do you check each claim against primary sources?

Give the AI one claim at a time and the documents to check it against. If the tool can retrieve filings itself, ask it to name the filing and section. If not, paste or upload the relevant 10-K, 10-Q or transcript.

A prompt that works:

Check this claim against the attached 10-K and the last two 10-Qs: "Gross margin has risen in each of the last four quarters." Quote the figures with their reporting periods and the page or section they appear in. If the documents do not support the claim, say so plainly. Do not use outside knowledge.

Then check the quoted figures yourself. It takes a minute and catches the most common errors: the wrong period, a non-GAAP figure presented as GAAP, or a segment figure presented as a company total. For a refresher on where things live in a 10-K, see how to read a 10-K with AI.

Step 3: How do you force an honest bear case?

Ask for the opposing argument explicitly, and constrain it to evidence.

Act as an analyst who is short this stock. Using only the attached filings and transcripts, give the five strongest arguments against the thesis below. For each, cite the passage. Then state which of my claims is the weakest and why.

Useful places for the bear case to look:

• Risk factors in Item 1A of the 10-K, especially any that are new or reworded
• Customer concentration disclosures
• Changes in accounting estimates, such as useful lives or revenue recognition
• Cash flow versus reported earnings, particularly if earnings rise while operating cash flow does not
• Stock-based compensation and share count growth
• Management tone shifts between calls

If the AI cannot produce a credible bear case from the documents, that is mildly reassuring. If it produces a strong one, you have learned something important before buying. We cover this angle further in using AI as a devil's advocate.

Step 4: How do you find the evidence gaps?

Finish by asking what is missing:

List the claims in my thesis that the attached documents neither confirm nor contradict. For each, say what information would be needed and where it might be disclosed.

Gaps are not failures. They are the parts of your thesis that are assumptions, and you should know which those are. Often a gap is a metric the company does not disclose, which is itself worth noting.

What does a finished verification look like?

Summarise the result in a short table you can keep:

That table is your thesis in checkable form. The "watch next" column is what a monitoring system should track for you after you buy.

Which AI tools are suited to thesis verification?

Any capable model can help with the reasoning. The difference lies in sources:

For a fuller discussion of reliability, see can you trust ChatGPT for stock research?.

Where does Ecomerate fit?

Ecomerate is an autonomous AI research and portfolio-monitoring agent for DIY investors, free to start. Its answers are evidence-backed with sources, which suits claim-by-claim verification. Once you own a stock, it keeps monitoring the thesis against new SEC filings and earnings and sends alerts by iMessage when a claim changes status. Chains, its dependency-mapping feature, also checks the suppliers and customers your thesis relies on.

This article is for general information and research education only. It is not investment, tax or legal advice, and it is not a recommendation to buy or sell any security. Check material facts against current primary sources before making a decision.