Thirty companies. Six verticals. Two engines tested directly. The first independent measurement of whether the multi-million-dollar IR page infrastructure American public companies pay for actually earns AI engine retrieval surface.
By EPR Editorial Team. June 2026. Phase 1, Pattern-Confirmation Pilot. Edited Sep 6, 2026.
Key Findings: Estimated, Modeled
- Financial services: ~100% retrieval. Every tested bank and insurer's IR page surfaced in at least one engine's answer.
- Biotech: ~92% retrieval. Clinical-trial and pipeline pages cited almost as reliably as financial-services IR pages.
- Utilities: ~74% retrieval. Rate-case and regulatory-filing pages cited moderately.
- Consumer/retail: ~41% retrieval. IR pages competed with far more crawlable owned-media and press coverage for the same queries.
- Mega-cap tech: ~4% retrieval. Engines overwhelmingly cited press coverage and analyst notes over the companies' own IR pages.
- Junior mining: ~0% retrieval. No tested junior miner's IR page surfaced in either engine across any query.
What the Audit Tested
The audit ran a fixed set of investor-research queries, "what is [company]'s revenue guidance," "when does [company] report earnings," "what did [company] say about [recent event] in its latest filing," against thirty publicly traded companies across six verticals: financial services, biotech, utilities, consumer/retail, mega-cap technology, and junior mining. Two engines were tested directly, Claude and Google AI Overviews, chosen because both surface citations inline rather than only in a separate sources panel. Figures throughout are labeled "Estimated ~X% modeled" and should be read as directional, not as a live query log.
Why Financial Services and Biotech Retrieve, and Mega-Cap Tech Doesn't
The pattern isn't about company size or IR budget, it's about competition for the answer. A financial-services IR page is often the only structured, machine-readable source for a specific regulatory disclosure or earnings figure, so an engine has little choice but to cite it. Biotech IR pages carry a similar advantage: clinical-trial results and FDA-pipeline updates are frequently first disclosed on the company's own IR page before any outlet covers them.
Mega-cap tech is the opposite case. When a query touches a company in that tier, the open web is saturated with analyst coverage, press reporting, and aggregator commentary, all faster-updating and more heavily cross-linked than the IR page itself. The IR page still exists, still gets built, still costs money to host. It simply loses the citation to a louder, better-linked source covering the same fact.
Junior mining sits at the far end for a different reason: thin coverage everywhere. Few analysts, minimal press, and IR pages that in this sample were frequently missing structured markup, PDF-only filings, or hosted on templates that don't crawl cleanly. The directional read is that junior miners aren't losing the citation to a competitor, the query itself barely returns a confident answer from either engine.
The IR-Hosting Vendor Matters
Retrieval also tracked, loosely, with which IR-hosting platform a company used. Pages built on modern IR-hosting platforms with clean HTML, structured press-release archives, and crawlable filing indexes retrieved more consistently within their vertical than pages hosted on older templates relying on PDF-embedded financials or JavaScript-rendered filing tables. This is a directional pattern from a 30-company Phase 1 pilot, not a vendor-by-vendor ranking. A larger sample is needed before naming specific platforms with confidence.
What This Means for IR and Comms Teams
The practical read for IR and communications teams: in categories where the company's own disclosure is the primary source of a fact (financial services, biotech), the IR page's crawlability is doing real work and is worth auditing directly: clean HTML over PDF-only filings, structured headings, and up-to-date press-release archives. In categories crowded with faster-moving third-party coverage (mega-cap tech), the more effective lever is likely making sure that third-party coverage itself cites the company accurately, since the IR page is unlikely to win the citation directly.
Methodology
Phase 1 pattern-confirmation pilot: 30 companies across 6 verticals, a fixed investor-research query set, two engines tested directly (Claude, Google AI Overviews). Findings are modeled and directional, labeled "Estimated ~X% modeled" throughout, and intended to confirm whether a pattern exists before a larger Phase 2 sample. Full methodology: The GEO Citation Audit: Six-Step Methodology.
The AI Visibility Audit Cluster
Master index: The AI Visibility Audit Cluster 2026: Research series measuring how corporate communications pages surface in AI engine retrieval across seven verticals.
Direct siblings: Wire Service Citation Audit 2026 · Vertical Citation Audit 2026 · 100 Consumer Brands Citation Share Study · Podcast Citation Index 2026 · Paywall Visibility Index 2026 · AI Language Citation Audit (Arabic, Spanish, French, German)
Methodology anchor: How Citation Share Works: The Measurement Framework
Related: AI Communications · GEO (Generative Engine Optimization) · The Answer-Engine Era
