The difference in one table
| Question | General-purpose chatbot, such as ChatGPT | PreTrAIde |
|---|---|---|
| Main purpose | Conversational help across many subjects, including financial concepts and documents | Pre-trade analysis of a proposed stock basket |
| Data freshness | Depends on the chatbot, its settings, connected tools and information supplied by the user; verify dates and sources | Enriches each stock with market data, including end-of-day prices and volumes, plus recent ticker news and analyst ratings; these are not live order-book quotes |
| Numerical work | Can calculate from supplied inputs, and some configurations have calculation tools; methods and inputs need checking | Uses a deterministic, reproducible quantitative layer for defined estimates and scores |
| Execution costs | Can explain or calculate them if given suitable data and a specified method | Estimates bid-ask spread, participation in daily volume and market impact |
| Basket analysis | Can discuss a basket when given its holdings and relevant data | Analyzes each stock and the basket, including concentration and portfolio breakdowns |
| Output | Flexible conversational response | GO, NO-GO or MODIFY for the basket and each line, with a model-estimated success probability, risk assessment, suggested execution strategy and plain-language reasoning |
| Personalization | Depends on the chatbot’s features, settings and information provided | Optional per-account memory from trades users log; it affects the AI layer, not quant scores |
| Trading | Capabilities vary by product and connected tools | Does not receive, transmit or execute orders |
Neither category replaces checking the inputs behind a consequential decision. A polished explanation can be wrong; a calculated estimate can also be uncertain or based on incomplete data.
What each tool is built to answer
A general-purpose chatbot is useful for questions such as “What does the bid-ask spread mean?” or “Summarize the risks described in this filing.” It can help organize a research process, explain a formula, compare definitions and turn technical language into plain English. ChatGPT, Claude and Gemini are examples, but their available data and tools vary by product and configuration.
PreTrAIde starts with a different question: “What are the measurable risks and execution considerations for this proposed basket, under these quantities and conditions?” You enter one or more stocks, each with a buy or sell side, quantity and execution conditions. PreTrAIde runs its PreTrade analysis before any order is placed. It is analysis software—not a broker, investment adviser or portfolio manager—and you place any trades yourself through your broker.
Its verdict is decision support, not an instruction to trade. GO, NO-GO and MODIFY are structured outputs of the analysis, not promises about what will happen.
Data freshness: check the timestamp, not the tone
For stock analysis, “current” is not a single standard. A last closing price, a recent news story, an analyst rating and a live bid-ask quote describe different things at different times. A chatbot may have access to connected tools, may rely on material you provide or may answer without up-to-date market data. Ask what source and timestamp support any time-sensitive claim.
PreTrAIde enriches stocks with about 57 fields per stock, drawing on end-of-day market prices and volumes, company profiles, recent news per ticker and analyst ratings. That gives its analysis defined inputs, but end-of-day data are not a live view of the order book. An estimated spread is not a firm quote, and an estimated fill cost is not a broker’s execution report.
The practical comparison is therefore not “AI versus live data.” It is whether a particular answer identifies its inputs, their dates, its calculation method and the limits of what those inputs can establish.
Computed numbers versus generated explanations
A chatbot can perform useful arithmetic. If you supply position values, it can calculate weights; if a calculation tool is available, it may run more elaborate computations. But a number in a fluent response should still be checked for units, assumptions and arithmetic. Prompting it to estimate trading costs does not, by itself, provide a defensible spread or market-impact estimate.
PreTrAIde separates a deterministic, reproducible quantitative layer from an AI reasoning layer. Its calculations and estimates include a liquidity-tier spread model with a volatility uplift and one-tick floor; daily-volume participation; Almgren-Chriss-style market impact; blended daily-volatility estimators; and one-day 95% value at risk (VaR). It also produces 0–100 liquidity, volatility and execution scores for each stock and the basket. Reproducible means the defined quant layer can produce the same result from the same inputs; it does not mean the estimate equals a future realized outcome.
For a clearly hypothetical example, suppose an order is for 2,000 shares and the stock’s average daily volume used in the analysis is 100,000 shares. Its participation is 2,000 ÷ 100,000 = 2% of daily volume. That ratio helps describe the order’s size relative to typical trading activity. It does not say what price the order will receive.
Execution costs and whole-basket risk
The quoted stock price is not necessarily the price at which a trade fills. The bid-ask spread is the difference between the quoted ask and bid, measured in currency per share. Market impact is a separate estimate of how an order may affect its execution price; it depends on assumptions and can differ from realized costs. PreTrAIde estimates both spread and impact, as well as participation in daily volume, to inform its suggested execution strategy.
A chatbot can explain these ideas well. With appropriate inputs and an explicit model, it can also work through an illustrative calculation. PreTrAIde’s narrower purpose is to apply its defined quantitative methods to the proposed order, rather than merely describe why execution costs matter.
Basket-level risk adds another dimension. PreTrAIde breaks a portfolio down by country, sector, region, currency and market capitalization, and calculates a Herfindahl-Hirschman index (HHI) for concentration. For basket weights expressed as decimals, HHI = Σ(weight²). A hypothetical basket with two equally weighted positions has HHI = 0.5² + 0.5² = 0.50; one position holding the entire basket has HHI = 1.00. The index describes concentration, not whether either basket is suitable for a particular person.
Verdicts, probabilities and guardrails
PreTrAIde’s AI agents reason about each stock and the basket using the quant scores, news and analyst ratings. The resulting GO, NO-GO or MODIFY verdict includes a success probability, risk assessment, suggested execution strategy and explanation. The probability is a model estimate bounded between 5% and 95%. It is neither a guaranteed chance of a profitable trade nor a forecast of returns.
Guardrails correct AI output that conflicts with specified data and rules. A suspended stock is always NO-GO; a GO with a success probability below 50% is downgraded; probabilities are clamped to the 5–95% range. These corrections only move a verdict in the conservative direction. Guardrails reduce those defined inconsistencies; they cannot remove uncertainty from market data or future events.
A chatbot’s open-ended format is an advantage when you need to ask follow-up questions or explore alternative assumptions. A fixed verdict format is useful when you want to review the same categories for each proposed basket. Neither format is an investment recommendation.
Personalization and what happens after analysis
A chatbot’s personalization depends on the specific product, settings and information you choose to provide. PreTrAIde offers a different, defined mechanism: users can log whether they executed, partially executed or skipped a trade, along with real fill prices. PreTrAIde measures outcomes from market data and builds a per-account memory that personalizes later analyses in the AI layer only; identical inputs still produce identical quant scores. Memory levels unlock at 1, 5, 15 and 40 logged trades, and users can pause or erase their memory.
After a verdict, users can also open a paper-trading simulation of the basket to follow how it would have performed, without real money. These features add context for reviewing a decision; they do not establish that a future trade will succeed.
When to use which
Use a general-purpose chatbot to learn terminology, draft research questions, summarize a filing you provide or test whether you understand a model’s assumptions. Check summaries against the original document, especially where wording or numbers matter.
Use a specialized pre-trade tool when the question concerns a specified basket’s quantities, concentration, liquidity, estimated execution costs and structured pre-order review. PreTrAIde covers US stocks and major European exchanges. Its free tools also include a trading-cost calculator, portfolio concentration checker and pre-trade checklist, available without sign-up.
The tools can complement each other: explanation helps you understand an estimate; a defined calculation helps you inspect a proposed trade. Neither knows your full circumstances simply because it produces a confident answer.