The Verified Source Pack: How to Become a Source AI Agents Actually Trust

Think about what happens when someone uses an AI assistant to research a product, compare service providers, or check whether a business is right for them. The AI doesn't browse your website the way a human does. It queries, validates, and summarises. If it can't find your information in a form it can confidently use, one of three things happens: it hedges with vague answers, it pulls facts from third-party sources that look more structured, or it skips your business entirely.

Most SMEs are invisible to agents right now. Not because of poor SEO, but because they have never shipped a machine-readable version of their official truth.

Key Takeaway: AI agents need structured, verifiable facts, not pages. A Verified Source Pack is the infrastructure layer that packages your business's official truth in a form agents can trust and cite. Building one now puts you ahead of competitors who are still optimising only for human readers.

Written by Derek Chua, digital marketing consultant and founder of Magnified Technologies. Derek works with SMEs on SEO, GEO, and AI-era content strategy, and has applied these principles across client sites in professional services, healthcare, and e-commerce.

Why Schema Markup Isn't Enough Anymore

Schema markup tells search engines what your page is about. That solved the interpretation problem: is this a business, a product, a FAQ? Structured data reduced guesswork for crawlers.

AI agents have a different job. They don't just interpret pages. They make recommendations, summarise options, and sometimes execute on a user's behalf. To do that reliably, they need facts that don't wobble. They need to know: what are your actual prices? What is your return policy, including the edge cases? What is your warranty, and what does it exclude? What can someone expect when they contact support?

If those facts are spread across five different pages, inconsistent, or buried in marketing language, the agent is guessing. And agents don't like guessing. The risk of being wrong about your business's policies is too high.

The shift is simple but significant. Pages persuade humans. Schema clarifies pages. Verified Source Packs package truth for agents.

What a Verified Source Pack Actually Is

A Verified Source Pack is a structured, machine-readable document (or set of documents) that contains your business's official operational truth. It's not a content marketing piece. It's closer to a data feed for your business, the same way a product catalog feeds an e-commerce platform.

The concept was articulated by SEO strategist Duane Forrester in a March 2026 piece, drawing on patterns already emerging across agentic AI platforms. The core argument: brands that publish structured, verified, official truth in a discoverable format will be the sources agents trust first.

There are four components.

What you're stating (the content)

This is the operational truth your business would stand behind in public. For a services business, it includes what you offer, what you don't, your process, your pricing structure or ranges, your service area, your credentials, and your key constraints (who isn't a good fit, what scenarios require different treatment). For an e-commerce business, it includes products, pricing rules, inventory behavior, shipping policies, returns, and warranty terms.

The constraint layer matters more than most businesses realise. If you don't explicitly state what you exclude or what requires special handling, an AI agent fills the gap with assumptions. Those assumptions may be wrong. Getting explicit about boundaries is as important as getting explicit about capabilities.

How it's structured (the format)

Machine-readable means a machine can parse it predictably. For most SMEs, this starts with clean JSON, a single-page summary document in plain text, or a structured FAQ that covers the key facts. The point is not technical sophistication. The point is predictable parsing.

More advanced implementations use an API endpoint that returns the pack index, which lets agents query your information directly rather than scraping it from pages. That's a developer project, not a marketing one. But even without an endpoint, publishing clean structured documents under a stable URL on your domain is a meaningful step forward.

How it's verified (provenance)

Provenance answers the question: how does an agent know this is genuinely your information, and that it's current? At a basic level, this means hosting the pack on your own domain with HTTPS, maintaining a version number and last-updated timestamp, and keeping it current when your business changes.

More advanced provenance involves signing the document so an agent can verify it hasn't been modified by a third party. You don't need to implement this immediately, but building the habit of versioning and timestamping is free and valuable from day one.

How it's found (discoverability)

A Verified Source Pack that no one can find is just an internal document. Host it under a stable, predictable path on your domain (something like yourdomain.com/agent-info or yourdomain.com/.well-known/source-pack). Link to it from relevant pages: your About page, your Services page, your Contact or Support section. Include it in your sitemap.

You may also encounter references to /llms.txt, a proposal that helps AI systems navigate your site. Including a pointer to your pack from llms.txt is a useful additional hint. Note that llms.txt is not a formal standard backed by major AI providers yet, so treat it as a directional signal rather than a guaranteed ingestion path.

What This Looks Like for a Services SME

For most SMEs, you don't need an API or cryptographic signing to start. Here's a practical first version.

Create a single document (start with plain text or a clean JSON file) that covers your business truth:

• What your business does and who it's for
• What it doesn't do or who isn't a good fit (the constraints)
• Your service process, step by step
• Your pricing structure (ranges are fine if exact prices vary)
• Your credentials, certifications, and professional memberships
• Your service coverage (geographic or sector-specific)
• Your contact and escalation paths
• A clear date stamp and version number

Host it on your domain. Link to it from your About and Services pages. Add it to your sitemap. Update it whenever something changes.

That's version one. It's not perfect, but it's already more structured than what most of your competitors have published.

At Magnified: What We've Observed Across Clients

At Magnified, when we audit client sites for AI visibility, the gap we see most often is not technical. It's truth fragmentation. A services firm might have its guarantee stated on one page, its process described differently across three service pages, and its pricing entirely absent from anywhere machine-readable. A clinic might have its consultation process buried in a 2019 blog post.

When we work with clients to consolidate this into a single structured document and publish it clearly, the impact is two-fold: it helps AI systems surface more accurate information about the business, and it forces the business itself to clarify things it had left vague. That second benefit is underrated.

The businesses that invest in this now are building infrastructure. The ones that wait are building a catch-up problem.

The Practical Starting Point This Week

If you want to take one action after reading this, open a new document and write down your business's constraints. Not what you do. What you don't do, who isn't a fit, and what situations require different handling. Most businesses have never written this down in one place. Doing so is the first step toward building something agents can actually trust.

For a broader starting point on how AI systems read (and misread) your site, see What AI Actually Sees When It Visits Your Website. For the technical layer that supports this, the Schema Markup Guide for SMEs covers the structured data foundation you should already have in place before building a Verified Source Pack.

If you want help auditing your current AI visibility and building a structured source pack, talk to the Magnified team.

---

Frequently Asked Questions

What is a Verified Source Pack? A Verified Source Pack is a structured, machine-readable document published on your domain that contains your business's official facts: what you offer, your policies, your credentials, your pricing structure, and your key constraints. It gives AI agents a trusted, parseable source of truth about your business, reducing the chance they pull inaccurate information from third-party sources or skip your business entirely.

Does my business need a developer to build one? A basic Verified Source Pack does not require a developer. You can start with a well-structured plain text or JSON document published under a stable URL on your existing website. Developer involvement becomes valuable later if you want a live API endpoint that agents can query directly, but that is not required to get started.

How is this different from schema markup? Schema markup tells search engines what type of thing a page is, and labels key attributes like name, address, and opening hours. A Verified Source Pack goes further: it packages your complete operational truth, including policies, constraints, process details, and pricing ranges, in a form agents can retrieve and reason with. Schema clarifies pages. A Verified Source Pack packages truth for agents. Both are useful; they solve different problems.

Will AI agents definitely use my Verified Source Pack if I build one? There is no formal commitment from major AI providers to consume a specific pack format. What is clear is that AI agents prioritise structured, verifiable, domain-hosted facts over scraped page content when making recommendations. Building a pack now positions you ahead of the curve, in the same way that early schema adopters benefited before Google formally required it.

What should I update first in my business's operational truth? Start with your constraints. Most businesses have never written down, in one place, who isn't a good fit, what scenarios require special handling, and what your business explicitly does not do. This is the information agents are most likely to get wrong when guessing, and the most valuable information to make machine-readable first.