AstraEye Labs
Sign InEnter the Lab
AstraEye Labs · Guide · 2026-11-14

AI Stock Analysis: How to Analyze a Stock with AI, Step by Step

By , Founder, AstraEye Labs

The short answer

Ask one question at a time and have each answered from market data and filings, not from a chatbot's memory: where the stock's trend stands, how far it moved after its past earnings reports, how volatile it has been, what its financial statements and insiders' filings show, and what was published about it. An AstraEye Labs report does this for U.S. stocks in one run, each measurement computed in code by a specialist AI scientist, with its data sources named; for an ETF it runs the analyses of its price and its news, not those about a company. It describes the stock; it does not say whether to buy or sell it.

A stock can be read many ways: its chart, its earnings, its filings, its news. This guide takes them one question at a time, names the analysis that answers each and links its method, then shows them together on a real published report. Its earnings and volatility steps use next-gen methods AstraEye built itself: computed in its own code from prices and SEC filings, each with its method written out.

  1. 01
    Trend
  2. 02
    Earnings moves
  3. 03
    Volatility
  4. 04
    Filings
  5. 05
    News
  6. 06
    Together
01The trend

Where does the stock's trend stand?

Start with the price: whether it sits above or below its 20-, 50- and 200-day moving averages, where its nearest support and resistance are, and whether its daily, weekly and monthly trends agree. Then compare it with its sector, so a stock rising with its whole sector is told apart from one moving ahead of it. In AstraEye, ATLAS, the market structure and momentum desk, computes these readings in code from daily price history.

02Earnings

How far has it moved after its past earnings reports?

A company's earnings release arrives on a date known in advance and can move its stock sharply. Find when each past report reached the market, measure the first session that could react against the close before the report, and see how long the stock's daily swings took to settle afterwards. Read the median beside the largest move: most reports move a stock modestly and a few move it a great deal. In AstraEye, ATLAS measures this from the moment the SEC accepted each earnings filing and daily prices, without a language model. It describes past reports only, never the next report's move or direction.

Measuring how a price reacted to news around its date is an established method: the event study, introduced by Fama, Fisher, Jensen and Roll in 1969. A U.S. company usually furnishes its earnings release to the SEC on Form 8-K, under Item 2.02, and you can read its filings yourself on the SEC's EDGAR database.

03Volatility

How volatile has it been?

Measure how much the stock has actually moved, not which way, usually quoted as a yearly rate. A day's high and low say more about that than its close alone. Place the reading in the stock's own history and beside the market's, so a quiet stock in a stormy market reads differently from a volatile stock in a calm one. In AstraEye, ATLAS reads realized volatility from each session's open, high, low and close, for the stock and for the market around it.

Reading volatility from a day's range rather than its close goes back to Parkinson in 1980, whose estimator ATLAS averages with two later ones.

For how much it may move next, the Volatility Outlook forecasts a stock's volatility over the next 5 and 21 sessions in code, from its daily prices and, when a report projected from the company's release a year earlier falls inside that window, its own past earnings reactions. It forecasts the size of moves, never their direction. Its forecasts are recorded in a public forward test judged on 17 May 2027.

04Fundamentals

What do its financial statements and filings show?

Read what the company reports: whether its profit is backed by cash, and how much debt it carries against what it earns. Then read what its insiders reported buying and selling: insiders sell for many reasons unrelated to a view of the company, such as taxes or scheduled plans, so their sales are noisier to read than their purchases. Both come from filings with the SEC that you can read yourself. In AstraEye, FELIX, the fundamental and macro desk, reads the reported financials, and NEXUS, the on-chain and alternative data desk, reads insiders' Form 4 filings.

05The news

What was published about it?

Read the news from recent days and the company's own filings, such as an earnings release or a report of another material event. A move that follows a filing within hours reads differently from one with no news at all. In AstraEye, NOVA, the news, catalyst and sentiment desk, reads the news coverage and classifies a company's 8-K filings by the SEC's own item codes.

06Together

Where do the readings agree, and what is missing?

No single reading describes a stock. Put them together, note where they conflict, and note what could not be measured. In an AstraEye report, SAGE, the research guide, writes the bottom line from the desks' findings and answers the questions your project set. Before the report is delivered, code checks the figures in SAGE's summary against the run's own data, and its dates too when every analysis's findings were read in full, and the report ends with the data sources it used, so you can check any figure against them.

How a report is checked, and what those checks cannot catch, is on the research standards.

◆A real example

What does AI analysis of a real stock look like?

These are the parts of one published report, as its public page shows them: the bottom line, the questions its project set, the analyses it used with their methods, and its data sources.

AstraEye Research · published 2026-11-12
Cisco (CSCO) reports fiscal Q1 2027 after the close on Nov 12, at $100.76 after a 7.4% three-session rebound: expectations, estimates, past report moves, trend and news
U.S. stocks3 desk sectionsSources credited
Impersonal research. AstraEye Labs is not a registered investment adviser, does not give financial advice, and does not know your goals or risk tolerance. This research does not account for your individual goals, risk tolerance, financial situation, or other circumstances. Past performance does not predict future results.

Bottom line

Main conclusions

  • Consensus EPS and revenue align almost exactly with Cisco's midpoint guidance, shifting the burden of the reaction to forward guidance and commentary.
  • Technical weakness into the report amplifies the stakes: the 200-day average sits just 0.8% above, making it a near-term fulcrum.
  • The 90-day upward revision cycle that lifted estimates by more than 6% has stalled completely in the past 30 days.

A meaningful guidance raise or cut for fiscal year 2027, or a quarterly result outside the $1.32 to $1.34 EPS and $18.0 to $18.2 billion revenue range, would reset both the estimate and technical picture.

Questions and answers

Questions this research set out to answer

  • What do analysts expect for the quarter (revenue and earnings per share)?

    Analysts expect Cisco to report fiscal Q1 2027 EPS of $1.33 (adjusted basis) and revenue of $18.12 billion, based on a consensus of 15 analysts as of November 12. Quarterly consensus revisions over the past 30 days have been minimal: EPS +0.10% and revenue +0.07%.

  • How does that compare with the guidance Cisco gave for it?

    Consensus sits within the company's August guidance on both measures. Cisco guided Q1 revenue at $18.0 to $18.2 billion and non-GAAP EPS at $1.32 to $1.34. Consensus EPS of $1.33 falls near the midpoint of that range, and consensus revenue of $18.12 billion likewise sits near the center.

  • How have estimates for the quarter and the fiscal year moved over the past 30 and 90 days?

    Over 30 days, fiscal-year EPS estimates rose just 0.07% and revenue 0.06%, both essentially flat. Over 90 days the picture is markedly different: EPS rose 6.62% and revenue 6.15%, indicating a strong upward revision cycle that has since stalled. Quarterly consensus revisions mirror the 30-day flatness at +0.10% EPS and +0.07% revenue.

  • How has Cisco stock moved after its past quarterly reports?

    Eight past post-earnings reactions are on record. The three most recent alternated sharply in direction: a decline of 8.4% after the August 12 report, a gain of 13.41% after May 13, and a decline of 12.32% after February 11. The median number of sessions to settle was 5.5.

  • How large were those moves?

    The median absolute earnings-day move across the eight reactions was 4.73% and the mean was 6.17%. The three most recent reactions were all well above the median, each exceeding 8% in absolute terms, with the largest a 13.41% gain after the May report.

  • Where do trend and momentum stand against the 50-day and 200-day averages going into the report?

    Trend reads sideways and momentum reads bearish, a disagreement. The stock closed 6.7% below its 50-day average and 0.8% below its 200-day average, with RSI at 43.01 in the subdued band. All four momentum votes are bearish. Support is 15.5% below at 85.14 and resistance 11.8% above at 112.69.

  • What has coverage said about orders, AI networking demand and margins?

    Coverage reports Cisco integrated Anthropic's Claude Mythos into its AI Defense framework to enhance enterprise AI security, including adaptive red teaming and tools for monitoring AI spend. Coverage also notes ETFs added $83.5 million in Cisco holdings on November 10. The news window does not specifically discuss orders, broader AI networking demand trends, or margin commentary.

How this was computed

The analyses behind this report, each linked to its own public page.

2 more analyses in this report are not in the public catalog, so their methods are not listed here.

All research methodsMethods AstraEye built itself

Data sources

◇More reports

Which stock reports has AstraEye Labs published lately?

The newest reports in the research library that cover a U.S. stock or ETF, each readable without an account and each ending with the data sources it used.

  1. New · Published

    Tesla (TSLA) after the Cybertruck phase-out news: estimates, analysts and trend (Nov 9, 2026)

    Stock
    • FELIX
    • NOVA
    • ATLAS

How to read one, part by part, is in the guide How to research U.S. stocks and crypto with AI: one real report, step by step. For a coin, the same questions are in the guide AI crypto analysis: how to analyze a coin with AI, step by step.

◇Limits

What can't this analysis tell you?

Figures are dated, not live

Every figure is as of the data it was computed from, and depending on the source that data can be real-time, delayed, end-of-day or periodic. A report is a record of the day it was researched and is not updated afterwards, so its figures age from that day.

The prose can be wrong where the figures are not

The figures are computed in code; the sentences around them are written by an AI model. The checks above catch a figure or a date with no source in the run, and some contradictions, but they cannot prove every sentence right: a correct figure can still be described wrongly, and the conclusion review does not always run or reach every conclusion. Check anything important against the source the report names.

Securities: U.S. only

Its stock, ETF and mutual fund coverage is U.S. only, and a mutual fund is covered only once it is tracked, without the company analyses or the market-wide scans.

Priced per run

Each research run is priced in Astra Credits, so it is built for reports, not for many quick lookups through the day.

No advice and no trades

It does not say what to buy or sell, and it does not place trades.

✦Your own research

Analyze a stock you follow

Read a published report on a stock first, no account needed. Then choose the U.S. stocks and ETFs you follow and run the analyses you choose: every run shows its price in Astra Credits before it starts.

?Questions

Common questions

How do I analyze a stock with AI?

Ask one question at a time and have each answered from market data and filings, not from a chatbot's memory: where the stock's trend stands, how far it moved after its past earnings reports, how volatile it has been, what its financial statements and insiders' filings show, and what was published about it. An AstraEye Labs report does this for U.S. stocks in one run, each measurement computed in code by a specialist AI scientist, with its data sources named; for an ETF it runs the analyses of its price and its news, not those about a company. It describes the stock; it does not say whether to buy or sell it.

Is there an AI tool that analyzes stocks from real data?

Yes, if its figures come from market data and filings rather than a language model's memory, which may be out of date or wrong. Look for a tool that names the data sources behind its figures and shows how each analysis is computed. In AstraEye Labs, specialist AI scientists compute each measurement in code from licensed market data and public sources such as SEC filings, each analysis has a public page stating what it answers and, where one is published, its method, and every report lists its data sources.

How much does a stock usually move after earnings?

It differs from company to company and from report to report, so measure it on the company's own past reports: the first session that could react to each one, against the close before it. The median says more than the average, because a few reports move a stock much more than most. Past reactions describe the past; they do not say how the next report will move it.

Does AI stock analysis tell me whether to buy a stock?

Not from AstraEye Labs. Its research is impersonal and descriptive: it reports what the data shows about a stock and how each figure was measured, and it does not say whether to buy, sell or hold anything.

Can I use AstraEye Labs inside ChatGPT, Claude, Gemini, Perplexity, Grok or Glama?

Yes. AstraEye Labs connects to ChatGPT, Claude, Gemini, Perplexity, Grok or Glama as an app, so the assistant can read your finished reports and their methods, ask SAGE about them, and price and start new research.

How much does AstraEye Labs cost?

There is a free plan, and paid plans start at $19 a month (Space Analyst). New accounts get a 7-day trial with no card. Every research run shows its price in credits before it starts.

§Sources

Sources cited

  1. Eugene F. Fama, Lawrence Fisher, Michael C. Jensen and Richard Roll, International Economic Review, vol. 10, no. 1, pp. 1-21, The Adjustment of Stock Prices to New Information, February 1969.
  2. U.S. Securities and Exchange Commission, Investor.gov, Using EDGAR to Research Investments, accessed November 2026.
  3. Michael Parkinson, The Journal of Business, vol. 53, no. 1, pp. 61-65, The Extreme Value Method for Estimating the Variance of the Rate of Return, January 1980.
→Read next

More guides