What is AEO (Answer Engine Optimization) and how is it different from SEO?

Answer Engine Optimization (AEO) optimizes digital content so AI systems — not search engines — correctly retrieve, cite, and describe your company. SEO targets ranked links for human users; AEO targets the factual layer inside Large Language Models. When an investor asks ChatGPT about a company, SEO determines page rank; AEO determines whether the AI mentions you at all, and with what framing.

Why do publicly listed companies need AEO?

Because the due diligence workflow has already changed. Bloomberg terminals embed GPT, analysts use Claude for instant summaries, fund-of-funds screen with Perplexity, and retail investors ask Gemini about stocks. AEO ensures your corporate narrative travels intact through the AI pipeline that now mediates the majority of investment research.

How do LLMs form opinions about stocks and companies?

LLMs retrieve from a probabilistic corpus built from structured databases, web content, financial filings, news, and social sentiment. An answer is shaped by three factors: structured data signals, surface-area and authority of web content, and consistency checks across multiple sources. AEO aligns all machine-readable surfaces to ensure accurate representation.

What is Machine Capital and the B2A paradigm?

Machine Capital is capital allocation executed, filtered, or guided by algorithmic and AI-driven systems. B2A (Business-to-Agent) recognizes that a growing share of investment decisions flows through AI agents. IRAgent is the world's first B2A platform, bridging traditional IR disclosure with the structured, machine-readable layer that AI capital demands.

How is AI-driven due diligence different from traditional analyst research?

AI due diligence is parallel, probabilistic, and persistent. It cross-references dozens of sources simultaneously and never forgets past negative news cycles. You cannot spin an AI — you can only ensure the signal-to-noise ratio favors accurate information through structured data and authoritative content seeding.

How does IRAgent influence what AI models say about a company?

Through four white-hat channels: (1) Schema and Structured Data Injection — embedding JSON-LD into web properties; (2) High-Authority Content Seeding — publishing entity-optimized content; (3) Vector Database Alignment — ensuring content embeddings are present in RAG pipelines; (4) 24/7 Monitoring and Correction — detecting anomalies and triggering re-injection. No model weights are manipulated.

What are structured data and Schema.org injection?

Structured data is a standardized vocabulary (Schema.org) that describes entities in a format machines can parse without ambiguity. JSON-LD markup declaring company name, ticker, industry, revenue, and leadership allows AI crawlers to ingest ground truth rather than guessing from unstructured text — preventing hallucination.

What is AI Share of Voice (AI SOV) and how do you measure it?

AI Share of Voice measures a company's presence, sentiment, and accuracy across AI-generated answers relative to its peer set. We track four dimensions: Mention Rate, Sentiment Polarity, Factual Accuracy, and Hallucination Rate — benchmarked against a self-selected competitor set in a quarterly AI Perception Report.

How do you measure and correct AI hallucination about companies?

We query each target AI platform daily with a standardized prompt battery, parse responses, diff against ground-truth data, and score for hallucination. When detected, correction triggers multi-surface re-injection: updated structured data, refreshed authoritative content, and targeted entity disambiguation signals.

Is AEO compliant with securities regulations and listing rules?

Yes. All work operates within standard IR disclosure principles. We do not create false statements or fabricate data. Our role is to ensure accurate, verifiable, publicly disclosed information is correctly structured and surfaced for AI systems. We recommend client legal counsel review all content prior to deployment.

What does a typical IRAgent engagement look like?

Four phases: (1) AI Perception Audit (5 business days, complimentary) — mapping how every major LLM describes your company; (2) Schema and Vector Injection (2–4 weeks) — structuring corporate data; (3) Narrative Translation (ongoing) — converting IR materials into AI-optimized formats; (4) Continuous Monitoring (retainer) — 24/7 AI SOV tracking with monthly reporting. Full programs typically 3–6 months.

Which AI platforms do you cover?

Standard coverage includes OpenAI GPT-5.5, Claude Opus 4.7, Google Gemini 3.1 Pro, Perplexity, and Bloomberg GPT. Platform coverage expands based on client industry and target investor geography. Enterprise clients receive coverage for any new major LLM release within 72 hours.

How is this different from hiring a traditional PR or IR agency?

PR and IR agencies write for human audiences. IRAgent output is architected for machines — deliverables include structured data, embeddings, and FAQ schemas. Measurement uses AI SOV and hallucination rate instead of media impressions. Many clients retain both: an IR agency for traditional media and IRAgent for the machine layer.

What happens when AI models update and change their descriptions of us?

Model updates are the single largest source of volatility in AI-generated corporate descriptions. Our 24/7 monitoring detects anomalies within hours and triggers re-injection protocols to restore accurate signals. AEO is not a one-time project — it is a persistent operational requirement.

What qualifies a company for your services?

We work with publicly listed companies, pre-IPO firms in active fundraising, and companies with market capitalizations above approximately $500M USD. Exceptions are made for high-growth companies in sectors with heavy AI analyst coverage (SaaS, fintech, biotech, clean energy).

What is the pricing model?

Custom-quoted based on: number of AI platforms covered, complexity of the entity graph, and cadence of monitoring and reporting. Initial AI Perception Audit is complimentary (5 business days). Full engagement pricing is discussed after the audit when both parties have data-driven understanding of the AI perception gap.

Do you support multi-market and multi-language IR (HKEX + English + Chinese)?

Yes. Multi-market and multi-language coverage is core to the platform. We optimize across English, Simplified Chinese, and Traditional Chinese — covering major AI platforms in each language ecosystem. For HKEX-listed companies, this ensures correct representation in both English and Chinese LLMs. Native Chinese-speaking AEO specialists maintain separate quality benchmarks for each language track.

Why can't AI read PDF investor decks?

Most AI crawlers and LLM retrieval pipelines strip rich-media formats on ingestion. A 50-slide PDF deck is typically processed as a single unstructured text blob — charts, tables, and design hierarchies are flattened into noise. The AI does not 'see' the beautiful layout your agency spent weeks refining; it sees a wall of de-contextualized words. Without structured data markup, key metrics like ARR, gross margin, and growth rates are indistinguishable from boilerplate disclaimers.

What formats do AI systems ingest well?

HTML with JSON-LD Schema.org markup is the gold standard. Clean, semantic HTML paragraphs with proper heading hierarchy ( through ), paired with inline structured data, give AI crawlers a parseable data model. Supplement with XML sitemaps marking lastmod dates, FAQ and Article schema, and structured earnings data via FinancialProduct or Corporation schema types. For numeric financial data, machine-readable formats like XBRL or CSV downloads are preferred over image-based charts.

Should we stop producing PDF prospectuses?

No — PDFs still serve a critical role for human investors, lawyers, and printing physical copies. The strategy is not replacement but dual-format publishing: maintain your traditional PDF IR materials for human stakeholders while simultaneously publishing an AI-native HTML version with embedded structured data. Think of it like subtitles on a video — the video serves one audience; the subtitle track serves another. Both can and should coexist.

How much work is it to convert our IR materials to AI-native formats?

The initial conversion of a company's core IR corpus — annual report, investor presentation, earnings transcripts, and ESG report — typically takes 2–4 weeks with a specialized team. The heavy lift is in the first pass: extracting structured data from PDFs, mapping it to Schema.org vocabularies, and authoring the AI-optimized HTML. After the baseline is established, maintaining AI-native disclosure alongside traditional filings adds roughly 10–15% overhead to your IR production pipeline.

Analysis6 min2026-06-01

The Death of the PDF Prospectus — Why AI Doesn't Read Your IR Deck

You spent $400K on the investor deck. But when ChatGPT summarizes your company, it never sees those slides. Here's why traditional IR materials are invisible to the capital markets' newest gatekeeper — and what to do about it.

IRAgent Team

Published 2026-06-01

Let's run a thought experiment. Your company just filed a 300-page IPO prospectus. The PDF is immaculate — custom typography, branded color palette, executive photography, infographics that cost $40K each. Your board reviewed it three times. Your lawyers approved every comma.

Now an institutional analyst opens ChatGPT and types: "Summarize the financial outlook and key risks for [Your Company]."

Here is what happens: the AI does not open your PDF. It never sees your deck. Instead, it retrieves fragments from a half-dozen news articles, a Wikipedia stub, two Seeking Alpha posts, and a Reddit thread from 2023 — stitching together a composite description that may be 40% accurate and 60% hallucination.

The Format Problem: Rich Media Is Machine-Invisible

Most AI retrieval systems work by converting content to text embeddings — numerical vectors that represent semantic meaning. When you feed them a PDF, they run OCR or text extraction. The result is a single flat text string with zero structural hierarchy:

[
  "Page 1: OUR COMPANY We are a leading provider...",
  "Page 2: (image: revenue chart, alt text: '')",
  "Page 3: RISK FACTORS The company faces competition...",
  "Page 4: OUR TEAM John Smith, CEO (image: headshot)",
  "..."
]

Everything valuable about your document — the hierarchy of information, the relationship between revenue data and risk disclaimers, the distinction between audited financials and forward-looking statements — is flattened into an undifferentiated text stream. The AI cannot distinguish between your Q4 revenue figure and your cafeteria menu if both appear in the same extraction.

THE COST OF FORMAT INVISIBILITY

A 2025 study of 200 pre-IPO companies found that firms with no structured-data presence on their corporate websites had an average AI hallucination rate of 34% on key financial metrics — meaning roughly one in three AI-generated statements about revenue, margins, or growth were provably false. Companies with full JSON-LD schema deployment reduced this rate to under 8%.

What AI-Native Disclosure Looks Like

AI-native disclosure transforms your IR content into a format that machines can parse, index, and cite with precision. It consists of three layers:

Structured Data Layer — JSON-LD schema markup embedded in your HTML pages that explicitly declares your company's legal name, ticker, industry, NAICS classification, financial metrics, leadership, and corporate actions. This is the ground-truth layer that AI crawlers ingest as canonical information.
Semantic HTML Layer — Clean, accessible HTML pages with proper heading hierarchy, semantic sectioning, and machine-readable data tables. Charts are accompanied by structured CSV or JSON data downloads. Every image has descriptive alt text. Every financial figure has context.
Entity Graph Layer — Cross-referencing your company's structured data across multiple authoritative domains (your corporate site, investor relations portal, regulatory filings page, LinkedIn, Bloomberg profile) to create a consistent entity identity that AI models recognize across the web.

The Dual-Format Strategy

We are not advocating for the abolition of the PDF. The PDF prospectus remains the legal document of record, the format that lawyers and regulators understand, and the artifact that human analysts are accustomed to reviewing. The strategy is dual-format publishing:

PDF for humans — lawyers, analysts, regulators, print distribution
HTML + Schema.org for machines — AI crawlers, LLM retrieval pipelines, financial data aggregators

Think of it like subtitles on a video. The video serves one audience; the subtitle track serves another. Both coexist. Neither is optional in a multi-audience world.

Practical Steps for Your IR Team

Audit your current IR web presence. Run your investor relations page through Google's Rich Results Test and Schema.org validator. Count how many structured data types are detected. If the answer is zero, you have a format gap.
Deploy JSON-LD schema on your IR site. Start with Organization and FAQPage. Add FinancialProduct and NewsArticle for earnings releases. Add Event for investor day and AGM dates.
Build an AI-optimized HTML version of your annual report and investor presentation. Clean semantic HTML, structured data, machine-readable tables. Publish it alongside your PDF — same URL, different format for different audiences.
Monitor the results. After 30 days, run an AI Perception Audit: query ChatGPT, Perplexity, and Gemini with standardized prompts about your company. Document the accuracy improvement. Share it with your board.

The PDF prospectus is not dead. But it is no longer sufficient. The companies that publish for both human analysts and AI retrieval systems are the ones whose narratives survive intact into the next era of capital markets.

#AI IR#PDF prospectus#machine-readable#structured data#annual report AI#investor deck#AI-native disclosure#IPO filing#S-1 optimization

FAQ.RELATED

Frequently Asked Questions About This Topic

RELATED.ARTICLES

What is AEO? A Complete Guide for Listed Companies

Answer Engine Optimization (AEO) ensures AI systems correctly retrieve, cite, and describe your company. This guide cove...

READ

How ChatGPT, Perplexity, and Gemini Describe Your Stock

When an investor asks five different AI platforms the same question about your company, the answers vary wildly. This ar...

READ

ALL ARTICLES FAQ