---
title: "Why your IR website is the most undervalued asset in your capital markets program · ARC"
description: "Most public companies treat the IR website as a compliance requirement. As of September 2026 your IR platform can show you exactly how AI crawlers read it. Here is what that number means, and what to do about it."
canonical: "https://staging.arcengineered.com/insights/ir-website-agentic-optimization/"
markdown: "https://staging.arcengineered.com/insights/ir-website-agentic-optimization/index.md"
author: "Rachel Carroll"
published: "2026-10-08"
modified: "2026-10-20"
language: "en"
shareDescription: "Your IR platform can now show you how AI crawlers read your site. Most companies will not like the number. Here is what to do about it."
image: "https://staging.arcengineered.com/images/cards/ir-website-agentic-optimization.5791614172.jpg"
---

# Why your IR website is the most undervalued asset in your capital markets program

Ask a CFO what the IR website is for and the answer is usually "filings." That was adequate in 2018. Today it is the primary source the AI systems that pre-screen investment decisions read first, and it competes with data-vendor feeds for the right to describe your company. This piece sets out how agents use the site, the seven mistakes that cost companies most, and what a site built to be read correctly looks like.

Rachel Carroll

- Number Insight 02
- Published 8 Oct 2026
- Reading 7 min
- Sources 5, all public
- Series Signal engineering

Ask most companies what their IR website is for and the answer is that it is where investors go to find filings. That answer was adequate in 2018. Today, your IR website is the single most important piece of structured content your company owns. It is the primary source document for the AI systems that pre-screen investment decisions for a significant and growing share of the capital that determines your valuation.

Until September 2026, that was an argument. It is now a line item on your IR platform's dashboard: [Notified added AI crawler reporting to its IR Hub](https://press.notified.com/posts/pressreleases/notified-ir-hub-now-tracks-what-ai-tells-inve) in September 2026, and the other major IR website hosts are following. The report shows how many agents visited and which pages they reached. It does not show what they concluded. It cannot tell you whether the agent reconstructed your investment thesis or a competitor's, whether the figures it extracted were from the right period, how your own site weighed against the data-vendor feeds and news coverage it read alongside, or what a different structure would have produced. Traffic is the question the platform can answer. Comprehension is the one that moves capital, and it is measured differently.

Then on September 10, 2026, OpenAI launched [ChatGPT for Financial Services](https://openai.com/index/introducing-chatgpt-financial-services/), and the stakes for the IR website changed again. The product builds in data from Daloopa, [Quartr](https://quartr.com/newsroom/press-release/quartr-becomes-a-data-partner-to-openai-s-chatgpt-for-financial-services), LSEG News, and PitchBook and connects to FactSet, so that an equity researcher's agent reconstructs your company from data-vendor feeds, transcripts, and filings, alongside and in competition with what you publish yourself. Your IR website is no longer just a source. It is a candidate for canonical source, and if it is stale, fragmented, or unstructured, the agent will settle the question in favor of the vendor feed.

So the question has moved. It is no longer whether AI systems are reading your IR website. It is what you are going to do about the number your platform is about to show you, and whether your own site wins the argument about what your company is.

## How agentic systems use your IR website

When an agentic system builds a company profile to answer an investor's research query, it draws on multiple sources at once: filings, earnings-call transcripts, news coverage, analyst research, third-party data. Your IR website occupies a particular position in that hierarchy that makes it disproportionately influential: it is assumed to be your intentional, current, authoritative self-presentation. Filings are what you were required to say. The IR website is what you chose to say.

The systems also evaluate architecture, not just text. Where information sits on the page, how it is structured, whether metadata accurately describes the content, whether the language matches your filing disclosures: all of these are inputs into how the system weights and extracts what you have published. A thesis stated once, on a page the crawler reaches, in language that matches your 10-K, is retrievable. The same thesis scattered across a CEO letter, an investor deck, and a press-release archive is three partial signals that may not reconcile.

## Seven IR website mistakes that cost you capital

1. **No explicit investment thesis.** The most common failure and the most expensive. If the site does not state, in one place, why an investor should own the stock, the model constructs a thesis from whatever else it finds, usually a generic sector description. Ask an AI assistant "tell me about \[your company\] as an investment" and read the answer. If it could describe any of your peers, this is why.
2. **Stale content.** A "latest news" module showing a release from two quarters ago, an investor presentation from last year's conference season, a fact sheet with prior-year figures. Freshness is a ranking input for retrieval systems. Stale content is read as either inattention or decline, and the model does not distinguish.
3. **Fragmented messaging.** Different framing of the same strategy on the homepage, the IR landing page, and the latest deck. Humans reconcile these instinctively. Models score them as inconsistency, the same narrative-drift penalty that applies to earnings releases, applied to your permanent record.
4. **Poor content hierarchy.** The thesis in a PDF, three clicks down. Governance boilerplate above the strategy. Extraction weight follows structure; if your most important content sits below your least important, that is the weighting the system inherits.
5. **Missing structured data.** No schema markup on earnings releases, events, leadership, or financial data. Structured data is how you tell a machine what a page is rather than leaving it to infer. Most IR websites have none, which is a straightforward technical fix with outsized effect on retrievability.
6. **Generic executive biographies.** "Twenty years of experience in the industry." Models use leadership pages to answer questions about execution credibility. A biography that does not connect the executive's track record to the current strategy contributes nothing to that answer.
7. **No content cadence.** A site that changes only at quarter-end tells retrieval systems that nothing happens between quarters. A site with a steady cadence of thesis-relevant content (a note on a customer milestone, a perspective on a regulatory change, a data point on a market) is continuously re-crawled and continuously re-weighted. Cadence is the difference between being indexed and being current.

## What an agentic-optimized IR website actually looks like

The IR websites that generate the strongest agentic signals share four characteristics.

- **A dedicated investment thesis page:** the case for the stock, in plain language, in one place, consistent in substance with the thesis restated in every earnings release.
- **A regular content cadence:** a defined rhythm of thesis-relevant updates between reporting events, so the site is always current when crawled.
- **Perfect consistency across channels:** a controlled vocabulary for segments, metrics, and strategy that holds across the website, releases, filings, and transcripts.
- **Technical structure for extraction:** clean URLs, accurate metadata, schema markup on every financial and governance page, and content the crawlers you want are actually permitted to reach.
- **A canonical KPI record:** one historical table of recurring metrics, each labeled by period, currency, and definition, with links to the filing it comes from. The [FinRank benchmark](https://arxiv.org/abs/2608.07400) published in August 2026 showed that retrieval systems reading SEC filings lose 13 to 20 points of accuracy when they encounter near-identical passages from adjacent periods or peers. A single, labeled, linked source of truth on your own site is how you stop being one of those confusable passages.

None of this is exotic. It is the discipline of treating the IR website as a publication rather than a repository, and it is now auditable, by you and by anyone else with access to the same tools.

## The asset you already own

Your IR website is not where investors go after they have decided to look at your company. For a growing share of the capital that matters, it is where the decision to look at your company is made, by a system that visited it before any human did. For many companies it is also the only analyst there is: about a quarter of US-listed companies have no analyst coverage and a third have one analyst or none, [according to Nasdaq in October 2026](https://www.nasdaq.com/economic-institute/the-print/quarter-of-companies-no-analyst-coverage). For those companies, what the agent reads on the IR website is the research.

It is an asset you already own, already host, and already pay to maintain. As of September 2026, it is also an asset your own platform will score for you. The question is whether you treat it as the most important investor communication your company produces, or as the place where you keep your filings.

> The crawler report is coming whether or not you asked for it. Better to know what it will say.

## ARC tests how AI research tools read your company.

In a workshop, we present our current findings and apply them to your own materials.

[Request a workshop](https://staging.arcengineered.com/contact/)

## Keep reading

### [The first investor to read your earnings release isn't human](https://staging.arcengineered.com/insights/earnings-release-agentic-ai/)

### [The agentic investor: What the new capital allocator actually looks like](https://staging.arcengineered.com/insights/agentic-investor-capital-allocator/)

[All insights](https://staging.arcengineered.com/insights/)