By Alexey Kovalyov, Founder, AstraEye Labs
If an AI chatbot keeps giving you stock or crypto figures you cannot trace, look for a research tool whose numbers come from market data, with the source and the method attached. These seven checks work on any such tool. Each comes with a test you can run yourself, and with how AstraEye Labs answers it, limits included.
A language model can write a price or a ratio from memory and state it with confidence. A research tool you can rely on computes each figure from market data and can say which data it used.
Ask for a figure you can look up yourself, such as last week's closing price of a stock you follow, and ask where it came from. Compare it with the exchange or a price provider. Then ask again in a new session: a recalled figure can change between answers, while one computed from data for a fixed date does not.
Every figure in an AstraEye report is computed in code by a specialist AI scientist on licensed market data. The language model explains the results; it does not produce the numbers.
Limits: The prose that explains the numbers is written by a language model, so read each figure with its window and date, not only the summary's wording.
The questions to ask about a single figure are in the guide Seven checks before you trust an AI's stock or crypto numbers.
A figure is only as checkable as its source. A tool that names the provider behind its numbers lets you open the same source and compare.
Pick one figure and ask which provider it came from and as of which date. An answer such as "market data", with no provider and no date, cannot be checked.
Every AstraEye report ends with a Data sources section that credits each provider it drew on.
Limits: Sources are credited for the report as a whole, not beside each figure. To trace one number, read how its analysis works and compare the number with the provider credited for that kind of data.
AstraEye's providers and the terms behind them are on the data sources page.
Two tools can give the same name, such as momentum or volatility, to different calculations. The method tells you which one you are reading, and its limits tell you what the reading does not support.
Ask how one reading was calculated: the rule, the window and the data it used. Then ask what the method cannot tell you. A tool with a real method answers both; a tool without one restates its conclusion.
Each AstraEye analysis states what it measures, the rule it applies and what its method cannot support, and a report's Methods section says how each analysis in it worked. SAGE, the research guide, answers questions about any of it.
Limits: The published report pages anyone can open show the findings and the data sources. The Methods section is in the full report inside AstraEye.
No data source covers every asset on every day. A tool that reports a gap as a gap is telling you something; one that fills the gap with an estimate looks complete and is not.
Ask for research on something thinly covered, such as a small coin or a recent listing, and see whether the answer says which readings were missing and why.
AstraEye reports state their coverage per analysis and name what was not measured and why, rather than reporting a starting value as a reading. An analysis that does not apply to an asset, such as a company-fundamentals analysis on an ETF, is not run for it.
If you follow both markets, research split across two tools leaves you reconciling different dates, windows and definitions by hand, and questions about how the two markets relate go unanswered.
Ask for one report on a stock, an ETF and a coin you follow. Check that all three are covered, and that each figure states its window and date so the three can be compared.
One AstraEye report can cover U.S. stocks, ETFs and crypto, from desks for technical structure, fundamentals and macro, news and sentiment, on-chain and alternative data, derivatives, and discovery screens.
Limits: Stock and ETF coverage is U.S. only. Commodities are covered only through commodity ETFs, whose research describes the fund, which can drift from the commodity; commodity prices and futures are not covered. Analyses about companies, such as fundamentals and earnings, do not describe an ETF.
What one report across both markets looks like is in the guide Stocks, ETFs and crypto in one AI research report.
If you already ask an AI assistant about markets, research you can reach from inside it is research you will use. What matters is what the assistant receives: research computed from data, or more of its own memory.
Look for the tool in your assistant's own directory, connect it, and ask about one report. Then ask where a figure came from: the answer has to point to the tool's research, not to the assistant's memory.
AstraEye Labs connects to ChatGPT, Claude, Gemini, Perplexity, Grok or Glama as an app. From the assistant you can read your finished reports, ask SAGE about them, and price and start new research, within a daily credit limit you set.
Limits: The assistant receives written research, not raw data: the summary, the findings and, where a report has them, its methods. Charts, tables and data fields stay in AstraEye. A connection needs a personal AstraEye account on a plan that includes the Research API. The Research API is available on the Space Analyst plan and above.
Connecting, asking and reading a report back, step by step, is in the guide How to research stocks and crypto inside ChatGPT or Claude, with real data.
Some tools charge a flat monthly fee and some charge for each use. Either way, the cost of a piece of research is worth knowing before you start it, not after.
Before running anything, ask what one run will cost. Then check what a plan includes each month, and what happens when that runs out.
AstraEye research is priced in Astra Credits. A run's price depends on the analyses it includes and the assets they cover, and every run shows its price in credits before it starts. Paid plans include credits each month, and credits you buy never expire.
Limits: Every run uses credits, on any plan. When a run needs more than you have, you can buy credits, change plan, or run fewer analyses.
The plans and the credits each includes are on the pricing page.
AstraEye Labs is built for research on the U.S. stocks, ETFs and coins you follow, in one report, where every figure is computed in code from licensed market data and each analysis states its method. You read it in AstraEye and ask SAGE, the research guide, about it, or reach it from ChatGPT, Claude, Gemini, Perplexity, Grok or Glama.
It is not built for placing trades: AstraEye does not manage money, route orders or transact. Its research is impersonal and does not know your goals or risk tolerance. It does not cover stocks outside the U.S. or commodity prices and futures, and an AI assistant connection carries written research, not raw data.
Open a published AstraEye report, no account needed, and run the checks yourself: the data sources at its end, the window and date on each figure, and what it says it could not measure.
Hold each tool to the same seven checks: whether its numbers are computed from market data or recalled by a language model, whether it credits its sources, whether it shows its method and that method's limits, whether it reports what it could not measure, whether one report covers stocks, ETFs and crypto, whether it works inside the AI assistant you use, and whether a run's price is shown before it starts. Test each one on a figure you can look up yourself.
Test them on numbers you can check. Ask each tool for figures you can look up yourself, such as a recent closing price or a company's reported revenue, and compare them with the price provider or the filing. For anything you cannot look up, ask for its source, window, date and method: a tool that cannot say where a number came from has not shown that it is right.
Yes. In AstraEye Labs, specialist AI scientists compute every figure in code on licensed market data, each analysis states its method and what it cannot support, and each report ends with the data sources it used. One report can cover U.S. stocks, ETFs and crypto.
Its stock and ETF coverage is U.S. only, and commodities are covered only through commodity ETFs, whose research describes the fund; commodity prices and futures are not covered. AstraEye does not manage money, route orders or transact, and its research is impersonal: it does not know your goals or risk tolerance.
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.
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.
The AI assistants named here are third-party products with their own terms. AstraEye Labs is not affiliated with their makers.