The AI Answer Gap: What Investors Hear About Your Stock

7 min readAshton
A junior mining headframe at dusk beside an investor reading an AI answer, showing the AI answer gap about a stock

An investor researching your ticker may now ask an AI assistant before opening your latest filing, and most junior miners have never checked what that assistant says.

When a retail investor asks an AI assistant about a small-cap mining ticker, the answer comes from whatever the model can retrieve most easily: an old press release, a stock-forum thread, a dated resource figure. That is not a marketing problem. It is a disclosure-hygiene and information-accessibility problem, and the fix requires no new claims, only making what is already disclosed easy to find and parse.

Why would an AI assistant get basic facts about our company wrong?

A model answers by retrieving and summarizing whatever text about your company it can parse, with no way to know that text is outdated unless a date is stated plainly. If your current technical report sits behind a regulator portal as a scanned PDF, and an old stock-forum thread is plain, crawlable text, the forum thread is often the easier source for the model to lift, regardless of which one is accurate.

This echoes a shift we have described before: an assistant does not rank ten sources for the reader to sift through, it picks one source and quotes it. A widely covered company can afford to be the second choice. A thinly covered junior explorer may have only one or two sources indexed, and neither may be the company's own words.

Is this an investor relations issue, or a marketing one?

Treat it as investor relations and disclosure hygiene, not marketing: the question is whether accurate, already-filed information is findable, not whether more people hear about the stock. A CEO who would never let a stale fact sheet reach a fund manager should hold what a retail investor's AI assistant retrieves to the same standard.

Many junior miners spend heavily on investor conferences and newsletter placements to reach this exact audience. Meanwhile, the fastest-growing research channel, an assistant answering a direct question, returns whatever it can find, sometimes a dated resource estimate or a promotional forum post. Closing that gap is not a bigger marketing budget. It is making sure the accurate version already on file is the version a model actually retrieves.

Where would an assistant actually be pulling information about us from?

An assistant draws on whatever is indexed and readable as text: your own website and press releases where those are plain HTML, financial news wires, data aggregator sites, and retail-investor forums discussing the stock. Filings on SEDAR+, the ASX Market Announcements Platform, or the London Stock Exchange's RNS service are the authoritative record, but are often image-based PDFs, which are harder for a model to retrieve cleanly than a forum post.

  • Company website and press releases, when published as readable text rather than scanned images.
  • Financial data aggregators and news wires, carrying whatever was last submitted, current or not.
  • Retail-investor forums and social platforms: plain text, frequently updated, easy for a model to retrieve regardless of accuracy.
  • The regulatory filing itself: authoritative, but often the hardest of the four for a model to parse.

What can we actually do about this without it reading as promotion?

Start by asking several AI assistants what they say about your company and your project, using the terms a retail investor would type, and compare the answers against your current filings. Where the gap is a formatting problem, meaning accurate information that exists only as a scanned PDF, republish that same already-disclosed content as plain, crawlable text on your own investor page, through your normal disclosure process.

This adds no new claim and takes no position on the stock. It is the same current information in a form a model can retrieve: resource and reserve figures with their reporting code and date, the qualified person or competent person of record, a link to the technical report and latest MD&A, and when each was last updated. Every fact already exists in a filing. Only its readability changes.

This is the same lens behind how we think about investor visibility for junior miners: the work is accuracy and access, not amplification.

Who inside the company should own checking this?

Assign it the way you would any other disclosure control: to the IR lead or corporate secretary, on a fixed schedule tied to your reporting calendar, not as a one-time project. Whoever owns continuous disclosure filings is the right person to also own checking what AI assistants say about the company, since the underlying discipline is the same: comparing a public claim against the current record.

Log what each assistant said, the date checked, and what was corrected as a result. That record shows a consistent, documented process if a regulator or an investor ever asks how the company handles public accuracy, and turns an open-ended worry into a recurring, bounded task with a named owner.

Key Takeaways

  • Retail investors researching small-cap mining stocks increasingly meet an AI assistant's summary before or instead of the company's own filings, and most issuers have never checked what that summary says.
  • What an assistant retrieves depends on what is easiest to parse, so a plain-text forum post can outcompete an accurate but scanned technical report for a model's attention.
  • The AI-answer gap is a disclosure-hygiene and accessibility problem, not a marketing opportunity, and the right response adds no new claim about the company or its project.
  • Any public content meant to close the gap between a filing and what an AI assistant repeats must go through the issuer's normal disclosure process and qualified person or competent person review before publication.

FAQ

Does checking what an AI assistant says about us count as continuous disclosure?

Checking is not disclosure. Publishing anything as a result of that check is, and it must go through the normal disclosure process, including qualified person or competent person review of technical content, before it goes live. Treat a correction to your investor page like any other public statement about the company.

Can we contact the AI companies directly and ask them to correct what they say?

Most AI assistants have no formal correction channel comparable to a press correction, and answers can change between queries and providers. The reliable lever is making the accurate, current, primary-source information easy for any system to retrieve, not petitioning a vendor to fix one answer.

Is any of this legal, disclosure, or investment advice?

No. This article is general commentary on information accessibility, not legal, disclosure, or investment advice, and it makes no claim about the prospects, valuation, or merits of any security. Speak with securities counsel and your qualified person or competent person before publishing anything about a specific project.

Who is behind Second Brain's view on this?

Second Brain is a growth consultancy, not an investor relations firm or a law firm, and this article follows the sourcing standards in our editorial policy. Read more about who runs Second Brain before treating anything here as more than a starting point for a conversation with your own advisers.

Sources