When investors ask an AI about your company, where do the answers come from?
When an investor asks a general AI model about your company, you cannot choose which sources it draws from or whether it reaches your own corporate information. A source-grounded IR assistant answers only from approved company content and always cites each point.
In short
More investors now research listed companies through AI tools instead of the IR website. When an investor asks ChatGPT, Claude or another model about your company, you cannot control which sources it draws on, whether it reaches your own corporate content, or whether it shows the investor where each answer came from. A source-grounded investor relations assistant answers differently: only from your own approved, published content and disclosures, citing each point back to the document, and honest about what the filings do not say. The difference is not a nicer answer. It is the source, and who controls it.
Key facts
- Investors increasingly research companies through AI tools alongside the IR website.
- Traffic to IR websites is decreasing, leaving companies blind about what resonates with investors and what they are looking for.
- With general AI, a company does not control which sources are used or whether its own documents are cited.
- A source-grounded assistant answers only from the company's approved documents and cites each point.
- It is honest about what the filings do not say, rather than filling the gap.
Part of investment research increasingly happens inside an AI assistant: investors type their question, read what comes back, and form a first impression. The answer they get may be perfectly reasonable. The question for an IR team is a different one: where did it come from, and could your company stand behind this answer?
Where do AI tools get their answers about a company?
From whatever public sources the model chooses to draw on, not the ones you would choose. When an investor asks a general-purpose AI about a listed company, it draws on the public material it can easily reach: financial aggregators, news and analyst commentary, reference sites, and your own disclosures where it finds and uses them. Sometimes that material is current and accurate. The point is that the model decides which sources to lean on, and you have no say in that choice and usually no visibility into it.
There is a second mechanic worth understanding. When someone asks an AI model a broad question, such as whether a company is worth investing in, the system breaks it into smaller sub-questions and answers each one separately: what the company does, how it has performed, what the risks are, what the latest news says. Each fragment is answered from whatever source the model surfaces for it. The investor receives a single, confident summary stitched together from several sources, and typically without any indication of which claim came from where.
There is a further complication some IR teams do not realise applies to them. A number of company websites, IR sections included, are configured to block AI crawlers outright, often as a side effect of a general security setup rather than a deliberate IR decision. When that is the case, the model cannot reach the company's own disclosures at all, however current or well written they are, and falls back entirely on secondary sources: aggregators, commentary, older material. The company has not just lost control of the source mix. It has completely removed its own narrative from it.
Can a company control what an AI says about it?
On the open models, not directly. You cannot edit how a general AI model summarises your strategy, and you cannot decide which sources it consults when it answers. You can try to improve the raw material by keeping your IR site clear and well structured, and that is definitely worth doing. But you are still handing the final answer to an external system that chooses its own sources, interprets them, and presents them as it sees fit.
What you can control is the assistant you place in front of investors on your own channels. If investors can get an instant, accurate answer directly from you, drawn only from your approved disclosures and citing them, many will use it, in the same way they would rather hear something from the company than from a stranger, if the company made it just as easy to ask.
Why does the source of the answer matter for a listed company?
Because for a listed company, the standard is not just "roughly right." Disclosure is a regulated obligation to be fair, accurate, and even-handed, and provenance is part of that. When an answer about your company is assembled from sources you did not choose and cannot point to, you cannot demonstrate where it came from or stand behind it as your own.
There is a quieter cost as well. When investors ask a general AI model instead of reaching out to you directly, you lose sight of what they are asking: the topics they keep probing, where they hesitate, where the uncertainty sits. That signal used to reach IR through mailings, calls and meetings. When it moves into a general assistant, you lose a direct read on what the market actually wants explained.
What is a source-grounded IR assistant, and how does it avoid creating disclosure risk?
A source-grounded assistant answers questions using only the documents a company has approved and published, and nothing else. Every point it makes is traceable to a specific disclosure. When the filings do not contain something, such as a valuation multiple or a formal target, it says so plainly rather than inventing a plausible-sounding figure.
That design is also why it does not create new disclosure risk. It can only repeat what is already public, worded as the company worded it, and point back to the source. It is not producing new statements about the company; it is retrieving and citing existing ones. This is what MIRA is built around. The value is not a cleverer answer than a general model might produce. It is a sourced one: the version an IR team can stand behind, and whose provenance an investor can check. It answers in plain language, in many languages, and stays honest about its limits.
What should an IR team do next?
Start by seeing what AI currently says about your company, and asking where those answers come from and whether you could stand behind them. Then decide whether you would rather investors rely on that, or on an assistant that answers only from your own record and cites it. The practical first step is to watch a source-grounded assistant work on your actual disclosures: how it cites, how it handles a question the filings do not answer, and what it shows you about the questions investors ask most.
The bottom line
Your company is being described to investors with or without you. The choice is a narrow one: leave the answer to a general system that chooses its own sources, or give investors an assistant that answers only from what you have actually published, that explains your story how you want it to be told and points back to it every time.
See what MIRA does with your own disclosures. Book a demo and we show you MIRA answering real investor questions, grounded only in your published documents, with a citation behind every line.
MIRA is an investor relations assistant that answers investor questions from a company's own approved, published disclosures, part of the Finvictum group's tools for direct, compliant communication between companies and their investors.