Can SEBI Research Analysts Use ChatGPT to Draft Reports? (2026)
Short answer: yes, nothing in SEBI's rules bans using ChatGPT or another AI tool to draft a research note. The part that does not change is accountability. The registered analyst owns every fact, every number, and every recommendation that goes out, no matter which tool typed the first draft.
Not legal advice. AI-disclosure guidance for SEBI intermediaries is still developing. Confirm current wording against SEBI's official circulars before you finalise your workflow.
What SEBI actually regulates here
SEBI regulates Research Analysts, not word processors. Its rules were written around the person and the firm holding the registration, the content they publish, and the process behind it, not the specific software used to produce a draft. There is no clause anywhere in the RA regulations that names ChatGPT, Claude, or any other AI tool, and there does not need to be. The obligations that already exist, accuracy, substantiation, disclosure, record-keeping, apply to the finished report regardless of how it was written.
That is the useful frame for the whole question: AI changes your production process, not your regulatory position. Everything SEBI already expects from a report you wrote by hand, it expects from a report AI helped you write.
Where the risk actually sits
You stay the author of record
Every report goes out under your registration number. SEBI does not ask what tool drafted a paragraph; it asks whether the registered analyst stands behind every claim in it.
Fact-check numbers against the primary filing
LLMs summarise and occasionally invent. Cross-check every figure, date, and quote against the BSE/NSE filing or annual report it claims to come from before it reaches a client.
Keep your own reasoning in the recommendation
AI can draft the "what happened" section faster than you can type it. The "why this matters" and the call itself should still be your analysis, substantiated the way SEBI expects.
Log the workflow in your audit trail
If AI touched the draft, note it in your internal record-keeping the same way you would note a research assistant's contribution. It costs nothing to log and it answers the question cleanly if SEBI ever asks.
Drafting aid versus advice engine
The distinction that matters is what the AI is doing for you. Using a model to summarise a BSE filing into plain English, tighten a paragraph, or generate a first pass at a note you then rewrite and verify is drafting assistance. You are still the one forming the view and checking the facts. Letting a model read a filing and hand you a buy or sell call with no independent verification is closer to outsourcing the analysis itself, and that is a much harder position to defend if SEBI asks how a specific recommendation was reached. Keep the model on the drafting side of that line.
This is also where the disclosure question gets real: SEBI's current guidance treats AI-assisted content differently from AI-generated advice, and where exactly your workflow sits on that spectrum determines whether you need to say anything about it. We cover the regulatory framing in more detail in SEBI's AI disclosure rules for Research Analysts.
The fact-check step you cannot skip
LLMs are fluent, and fluent is not the same as correct. A model can produce a confident, well-structured paragraph that misstates a quarter, invents a subsidiary name, or rounds a number in a way that changes the conclusion. None of that shows up as an error to a reader unless someone checks it against the source. Before any AI-drafted paragraph reaches a client, verify every figure and claim against the primary filing, not against the model's own summary of it. We go deeper on exactly what to check in AI hallucination risk in equity research.
The practical workflow
Most independent RAs who use AI well settle on a simple split: AI drafts the descriptive sections (what happened, the numbers, the context), and the analyst writes or rewrites the interpretive sections (why it matters, what to watch, the call). The analyst then reads the whole note end to end against the source filing before it ships. That single read-through, done by a human who is accountable for the result, is the step that actually protects you. Skipping it is the mistake, not using AI in the first place.
FAQ
Is it legal for a SEBI Research Analyst to use ChatGPT?
Yes. SEBI does not prohibit Research Analysts from using ChatGPT or any other AI tool to help draft a report. There is no rule against the tool itself. What SEBI does not relax is who is accountable for the content: the registered analyst remains fully responsible for every factual claim, recommendation, and disclosure in a report, regardless of what wrote the first draft.
Do I need to disclose that I used AI to write a report?
SEBI has issued guidance distinguishing AI-assisted drafting from AI-driven advice. Using AI as a drafting or summarisation aid under your review is different from letting a model generate recommendations without human verification. If your workflow leans toward the latter, or if AI materially shapes the recommendation rather than just the prose, disclose it and confirm current wording against SEBI circulars before publishing, since this is an area still being clarified.
What happens if ChatGPT gets a number wrong in my report?
You are liable, not the tool. SEBI holds the registered Research Analyst responsible for the accuracy of every report published under their registration. An incorrect EPS figure, a misstated quarter, or a fabricated data point that reaches a client is a compliance failure attributable to the analyst, the same as if a human trainee had made the error and it went out unchecked.
Can AI make the buy or sell recommendation, or just write the note?
Treat AI as a drafting and research-assistance tool, not a decision-maker. The recommendation, the rationale, and the risk factors need to reflect the analyst's own judgement and be substantiated the way SEBI expects any research call to be. Letting a model generate the call itself, with no analyst reasoning behind it, is a harder position to defend if SEBI ever asks how a recommendation was arrived at.