The Quantellect Research Desk is a reader-run workflow for getting a short, sourced research brief from public market pages. You point your own AI assistant, such as ChatGPT, Claude or Grok, at the public pages you want to review. You then ask for a compact summary that keeps observations, dates, links and unknowns visible.
This approach treats an AI stock research assistant as a reading and organizing aid. It does not replace the source pages. It does not monitor them for you, send alerts or rerun a check automatically. When you want an updated brief, you return to the pages and run the workflow again.
The aim is disciplined evidence collection. A useful brief says what a page shows, when the information applies, where it came from and what remains unconfirmed. It should not turn a headline, a scheduled event or a disclosed transaction into a conclusion that the source does not support.
How to use ChatGPT for stock research
If you are learning how to use ChatGPT for stock research, begin with a narrow question. Ask for a short brief on a small set of tickers and a defined group of public sources. The narrower the request, the easier it is to inspect the result.
Open the relevant Quantellect pages first. Then give your assistant the page URLs, your factual criteria and a request for citations. ChatGPT can use web search when available, and OpenAI explains that feature in its ChatGPT web search help page. Claude and Grok can be used in the same reader-run pattern when they can access the public pages.
Ask for observations rather than a prediction. For example, the assistant can identify an event date shown on a calendar, summarize a disclosed transaction, or connect a news item to the source page where it appears. It should quote or paraphrase the page carefully and preserve the source URL.
AI assistants can misread a page, misunderstand a table or fail to open a URL. Treat a generated brief as a draft for verification. If a page is unavailable or the data looks stale, the brief should say so instead of filling the gap with an assumption.
Choose your stocks and research sources
Start with the stocks you already want to examine. A Quantellect ticker page follows the pattern /us/stocks/{exchange}/{symbol}, with the exchange and symbol written in lowercase. The supplied AAPL ticker page shows the format you can use as a starting point for a selected stock.
Next, decide which source pages match your question. The market movers page can be used when you are reviewing the top-gainers list. The market news page can be used when you are collecting recent news items. These pages help define the evidence set, but they do not define your personal factual criteria.
Give the assistant a clear time boundary when the page displays one or when your research question needs one. Ask it to distinguish an observation from an interpretation. “The page lists an earnings date” is an observation. “The date will produce a favorable result” is an unsupported conclusion.
For each stock, keep the same output fields. A consistent structure makes it easier to compare briefs and notice missing evidence. At minimum, capture the ticker, the sourced observation, the relevant event or trade date, the source URL, the time the data was checked, counter-evidence and unknowns.
Check insider and congressional disclosures
Use Quantellect’s insider trades page to review SEC Form 4 insider-trade disclosures. The point of this step is to record what was disclosed, not to assign a motive or turn one filing into a forecast. Ask the assistant to preserve the transaction details that are visible on the page and to link back to the source.
The congressional trades page covers disclosed congressional trades. Apply the same discipline: identify the ticker, describe the disclosed observation, record the relevant dates and cite the page. Do not treat a disclosure as proof of an investment thesis, a future outcome or a recommendation.
Pay close attention to trade date versus disclosure date. The trade date is when the transaction occurred. The disclosure date is when the transaction was reported or made public. They answer different questions. A brief that combines them into one date can create a misleading timeline, so label each date separately when both are available.
If a record does not show a date clearly, say that the date is unknown. If the page is inaccessible, do not infer the date from a headline or from another source. The reader can rerun the check later and see whether the missing information is available.
Track earnings, dividends and IPO dates
The earnings calendar is the source for earnings dates and EPS information in this workflow. IPOs appear on that calendar as well, so use it when you are checking IPO dates. There is no separate IPO page in the source set for this article.
Separate scheduled from reported information. A scheduled earnings date is an event expected on the calendar. A reported EPS figure is information that has already been posted. Ask the assistant to label which kind of evidence it found and to attach the date shown on the page. It should not treat a scheduled event as a reported result.
For dividend research, use the Dividend Center. Ask for the relevant dividend observation and its event date as displayed. Keep the wording descriptive. The existence of a calendar entry or dividend record does not, by itself, establish what will happen next.
Dates can change, pass or become stale relative to the moment you read them. Include a data timestamp in the brief: the date and time when you checked the page. If the page does not provide a current timestamp for a record, distinguish the page’s event date from your own check time.
Read news without inventing a cause
Use the market news page to collect the article or headline that is relevant to your question. Ask the assistant to summarize only what the source says. Keep the headline, the source URL and the date shown on the page together.
A news item may describe an event without proving that the event caused a market reaction. A brief should not invent a cause from timing alone. It should say “the page reports” or “the source describes” when that is all the evidence supports.
Counter-evidence matters here. If another source page in your selected set points to a different date, a different description or no confirmation, include that conflict. If the source set does not establish a cause, write “cause unknown.” That is more useful than a confident explanation that cannot be checked.
Paste your research desk prompt
Copy the prompt below into your own AI assistant after opening the public pages you want it to review. Replace the bracketed fields with your question and sources. Keep the list to no more than three items so the brief stays easy to audit.
Act as a careful research organizer. Review only the public Quantellect pages and other URLs I provide in this message. Return at most three items. For each item, include: ticker; one or more sourced observations; the event date, with trade date and disclosure date separated when relevant; the exact source URL; the data timestamp showing when you checked it; counter-evidence; and unknowns. Label information as scheduled or reported when relevant. Apply PASS, FAIL or UNKNOWN only against these factual criteria that I provide: [insert my criteria]. Do not turn a missing, inaccessible or stale page into an assumption: missing or stale evidence is UNKNOWN. Do not make trade recommendations, predictions or suitability judgments. Do not invent a cause for a news event. If you cannot open or verify a page, say that clearly and leave the affected item UNKNOWN.
The prompt deliberately limits the output. “PASS” means the sourced evidence meets the reader’s stated factual criterion. “FAIL” means the sourced evidence does not meet it. “UNKNOWN” means the available evidence is missing, stale, inaccessible or insufficient. Those labels should never become a hidden rating system.
Your factual criteria should be your own testable questions, such as whether a page shows a particular event type or whether a date is present. Avoid criteria that ask the assistant to decide whether a stock is good, safe or suitable. Those questions cannot be answered responsibly from this source set alone.
Verify the brief and repeat the check
Read the brief beside the source pages. Open every cited URL, confirm that the ticker and dates match, and check that the observation is actually present. Pay special attention to the difference between a page’s event date and the time you checked the page.
Check the unknowns before you accept the summary. An empty unknowns section is not evidence that nothing is uncertain. If the assistant failed to open a page, misunderstood a table or used a stale observation, correct the brief or mark the item UNKNOWN.
Repeat the research desk manually when your question requires fresh evidence. Reopen the relevant ticker, insider, congressional, calendar, dividend, market-mover or news pages, then run the prompt again. The workflow promises a repeatable reading process, not automatic monitoring or alerts.
Over time, the value of the Quantellect Research Desk comes from the habit of checking sources in a consistent format. Your AI assistant can help organize the reading, while you remain responsible for opening the pages, checking the citations and deciding whether the evidence answers your own factual question.

