How to Use AI to Read Earnings Releases

Part 1 in a series on using LLMs for financial research

LLMs are incredibly good at summarizing documents, and investors and analysts around world are asking ChatGPT or Claude to summarize the 5-10+ page PDFs of earnings releases.

But LLMs will capture all the interesting-seeming information. And they may miss the boring-but-critical details that expert humans notice. How can you tell?

We analyzed the all the filings from every company in the S&P 500, and correlated them again high quality long-term forecasts of company fundamentals. We took the top 50 (10%) earning releases we deemed most forecast-relevant, and here we share what we learned about how to get LLMs to highlight what actually matters in earnings releases.

Key Takeaways

• When summarizing earnings releases, LLMs gravitate toward CEO quotes, vision statements, and product launches, which rarely predict actual business outcomes
• LLMs omit the seemingly boring info that human experts sense is important, in five categories covered below
• To get the right information from LLMs, you need to give certain data as context, and prompt for specific findings. We provide examples at the end.

5 Earnings Release Patterns to Flag

Pattern 1: Segment Divergence - When Headlines Hide Reality

We found that the most valuable information in earnings releases is when top-line growth masks segment-level deterioration, or when strong margins hide collapsing volumes. This divergence appears first in earnings releases because they combine current-period actuals with segment detail, and by the time the 10-Q is filed, the market has moved.

So if you're reading these right when they come out, pay attention. If you wait until the 10-Q is out, pass the 10-Q in as context to the LLM to contextualize the earnings release.

Found in 34 of 50 sample companies.

Pattern 2: Capital Deployment - Actions Over Words

We found that changes in capital allocation, especially when contradicting stated strategy, prove remarkably predictive of management's true confidence in the business. Capital allocation changes appear in releases before strategic rationale is fully explained. Management can spin strategy in calls, but cash deployment reveals true priorities.

Found in 29 of 50 companies.

Pattern 3: Guidance Mechanics - Structure Over Numbers

It's not in the guidance itself, but how guidance is constructed is revealing to management's confidence, visibility, and priorities. What's excluded or heavily qualified often matters more than what's included.

Found in 27 of 50 companies.

Pattern 4: Quantified Strategy - Specifics Beat Spin

We find that some earnings releases quantify strategic shifts, and when they do, this is much more valuable than qualitative fluff about strategic shifts. Strategic pivots are cheap talk until quantified.

Found in 23 of 50 companies.

Pattern 5: One-Time Costs Are Strong Signals

Finally, we find certain one-time charges, settlements, or gains often reveal far more about underlying business quality than management acknowledges. Non-recurring items are often dismissed as noise, but their specific causes reveal operational vulnerabilities or dependencies that affect long-term forecasting.

Found in 19 of 50 companies.

What to Ignore

Five information types consistently proved non-predictive, that LLMs naturally over-weight:

1. CEO Quotes & Strategic Vision

Present in 50 / 50 companies in this sample, predictive in approximately 8 cases. Management's qualitative assessment rarely adds information beyond the numbers. "We're pleased with results" and "focused on executing our strategy" add zero information—these sections exist for legal compliance and PR, not analysis.

Exception: First-time mentions of specific challenges or changes in tone (e.g., moving from "confident" to "monitoring challenges"). To get this, you'd need to pass the previous earnings releases into the context window to make it possible for the LLM to notice this.

2. Market Share Claims

Companies are creative about defining markets to claim leadership. "Leading provider of X in Y geography among Z customer type" is usually cherry-picked. These look like substantive claims to LLMs, but they aren't, so instruct the LLM to ignore them.

3. Product Announcements

"Launching new product X" is meaningless without: pricing information, addressable market size, expected revenue contribution, and competitive positioning. Frequently these launches aren't material until they have numbers later, but LLMs sometimes think they're critical.

4. Partnership/Customer Testimonials

Directionally positive but rarely material to forecasts unless quantified with contract values, exclusivity terms, and revenue impact timelines.

5. Y/Y Comparisons Without Context

"Revenue up 15% vs. prior year" requires knowing: Was there an acquisition? What was the prior year comparison? What's the sequential trend? What's the organic vs. inorganic split? If you want the LLM to include such information, pass in the requisite background documents. Otherwise tell the LLM to ignore this.

Conclusion: How Prompt AI for Earnings Analysis

You may expect us to guide you to break down into many queries and then reconstruct them into a top level summary. In fact, we've found LLMs are very smart and very good with working with long context. So a single good prompt, passing in the right background documents, should be all you need, if you're using a frontier model as of late 2025.

Here is some simple guidance on how to ask for what's critical and ask to avoid what isn't.

Step 1: Prepare your context

We find that 10-Q, 10-K, last quarter's earning's release, and other basic financials (such as the industry growth rate, or industry news) all help. See the bottom for links to some of our other pieces on how to get and extract insight from these documents, so that you don't flood the context window.

Step 2: Prompt for the things not to miss:

Here's some language to include on segment divergence:

Extract all segment-level financial metrics from this earnings release.
For each segment, provide:
- Revenue ($ amount and YoY % growth)
- Operating income/margin (if disclosed)
- Key operational metrics (customers, volumes, etc.)

Flag any divergence where:
- Growth variance >10 percentage points between segments
- Margin variance >200 basis points between segments

For capital allocation changes:

Extract all capital allocation activities from this earnings release:
- Share repurchases ($ amount, shares, avg price if disclosed)
- Dividends ($ amount per share, any changes vs. prior)
- M&A announcements (deal size, target, financing)
- Debt paydown or issuance ($ amount, terms)
- CapEx ($ amount, any changes vs. guidance)

Flag changes where:
- Buybacks changed >25% quarter-over-quarter
- Any new capital allocation activity appears
- Any stated capital allocation activity is absent
- Stated strategy contradicts actual deployment

List contradictions between stated strategy and actual capital deployment.

For guidance mechanics:

Do NOT summarize the guidance numbers. Instead, analyze the guidance structure:

1. What specific metrics are included in guidance?
2. What metrics were guided previously but excluded now?
3. What new metrics appear in guidance that weren't there before?
4. What qualifications or contingencies are mentioned? ("excluding," "adjusted for," "subject to")
5. What's the format? (specific number, range, qualitative only)
6. What's the language? ("expect," "target," "anticipate," "committed to")

If the prior earnings release(s) are included, flag any changes.

For quantified strategy:

Extract ONLY statements that contain specific numbers about operations. Include:
- Customer counts, growth rates, retention rates
- Volume metrics (units sold, subscribers, transactions)
- Pricing metrics (ASP, yield, price increases/decreases)
- Operational metrics (stores, rigs, capacity, utilization)
- Efficiency metrics (cost per X, revenue per Y)

Ignore:
- Any statement without a specific number
- Percentages without underlying base amounts
- Vague quantifiers ("significant," "strong," "increased")
- Strategic language without quantification

For one-time costs:

Identify all non-recurring items mentioned in this release:
- Restructuring charges
- Impairments or write-downs
- Legal settlements
- Tax benefits/charges
- Gain/loss on asset sales
- Unusual regulatory payments or refunds

Skip those representing <5% of quarterly earnings. For those above:
1. What is the item? ($ amount)
2. What caused it? (be specific)
3. What does this reveal about the business?
   - Strategic failure? (impairment of acquired assets)
   - Operational vulnerability? (litigation, cybersecurity breach)
   - Hidden dependency? (regulatory payments, one-time tax benefits)
   - Balance sheet cleanup? (accelerated write-downs)

Further Reading: Earnings Release vs other documents

This is the first in a series of posts about how to use AI to read other documents, like earnings calls, 10-Ks, and 10-Qs. In brief:

Earnings Releases vs. 10-Ks/10-Qs

Earnings releases provide simultaneous disclosure of current results + management response + updated guidance. This reveals real-time disconnects before 10-Q filings add detail.

The 10-K discusses strategy; the 10-Q provides comprehensive numbers. But earnings releases uniquely capture the immediate collision between strategy and operational reality. By the time the 10-Q is filed (30-45 days after quarter-end), the market has already moved on the earnings release information.

Earnings Releases vs. Earnings Calls

Numbers in the release are fixed; management explanation comes after in the call. The release shows what happened; the call explains why it happened.

Management can frame narrative in the call ("we're investing for the future," "temporary headwinds," "strategic pivot"). But the release has already quantified the magnitude of underperformance, capital allocation shifts, or guidance changes. The facts are locked in before the spin begins.

The companies that provide quantified, segment-level operational detail—whether intentionally or inadvertently—enable far superior forecasting. The companies that bury this information in vague language, exclude it entirely, or wait to disclose it on calls reveal themselves as either:

• Having something to hide
• Not understanding their own business deeply enough to quantify it
• Prioritizing legal protection over investor information