Answer engines don't crawl PDFs and they don't reward marketing language. They reward structured, explicit, factual pages that name what the company is and what it owns.
Investor discovery has a new layer
For two decades, the discovery path for a junior miner ran through Google search, broker notes, hot-copper-style forums and word of mouth. In the last 24 months, a fourth layer has settled into the mix: AI answer engines. ChatGPT, Claude, Perplexity, Gemini and the embedded answer boxes inside Google and Bing are now the first port of call for a meaningful slice of generalist investors.
These tools do not behave like a traditional search engine. They do not return ten blue links. They synthesise an answer, name a handful of sources, and present a consolidated view of the company. If the company website cannot be read cleanly, the engine pulls from secondary sources — broker websites, forum posts, two-year-old news articles — and the company loses control of its own narrative.
What answer engines read well
Three things stand out. First, explicit text: a sentence on the page that says 'X is a copper-gold explorer with projects in Western Australia at the scoping-study stage' is worth more than the most beautiful hero animation. Second, structure: clean headings, dated content, source-linked references back to ASX announcements, and consistent terminology across pages. Third, machine-readable metadata: schema.org markup that names the organisation, its projects, and the publication dates of recent updates.
None of this is technically difficult. The reason it is rarely done is that it sits outside the normal scope of a corporate web refresh, which tends to focus on visual identity rather than on information architecture.
What they read badly
PDFs are still under-indexed by most answer engines. Image-only project pages — where the geology, ownership and catalysts live inside a JPEG — are functionally invisible. Carousels that rotate three hero headlines every five seconds are sampled inconsistently. And copy that talks about 'unlocking shareholder value' without naming the project, commodity or stage gives the engine nothing useful to cite.
A practical readiness checklist
1. Every project page names commodity, jurisdiction, stage, ownership and tenement on the page in plain text.
2. The investor centre lists the most recent material announcements as HTML summaries, not just PDF links.
3. Board and management bios sit as crawlable text, not as PDFs of the annual report.
4. Organisation, person and project schema is present on the relevant pages.
5. The site uses canonical URLs, sensible page titles and meta descriptions that describe the company in entity-rich language.
Each of these is small in isolation. Done together, they shift the company from being summarised by the open web to being summarised by its own website — which is the only version of the story it actually controls.
