The Source File: what your organisation looks like to a machine
A people-first website is only half the story. This experiment renders an organisation the way a machine tends to: as a structured entity with a type, a name, attributes and links. We annotate that "source file" line by line to show how each field can shape AI visibility, entity clarity and source trust.
Introduction
We write web pages for people: paragraphs, images, a friendly tone. But AI systems increasingly relate to organisations as entities — structured things with a type, a name and a web of attributes and relationships. It can help to picture that as a source file: a compact, machine-readable description of who you are.
The file below is illustrative. It is not a literal export from any specific AI system, and the internal workings of those systems are not public. But it is a useful lens: if a machine had to summarise you in a few clear fields, what would it write — and is that what you would want it to say?
The source file
Illustrative structured data using the Schema.org vocabulary. Numbered markers correspond to the field notes below.
Field notes, line by line
Declares what kind of thing this is — here, a Person. Classifying the entity clearly is the first step in helping a machine reason about it rather than guess.
The canonical label for the entity. Keeping the exact same name everywhere supports disambiguation — telling you apart from others who share it.
Associates the entity with topics. This is where relevance lives: it signals the subjects for which this person might be a sensible answer.
A concrete, verifiable fact. Specifics like dates are easier to corroborate than adjectives, and corroboration is commonly associated with greater confidence.
Links the entity to its other authoritative profiles. These connections help a machine reconcile one identity across the web — a key part of building trust.
A short, plain-English summary. This is often the most quotable line — the sentence a system may lean on when it has to describe you in one breath.
How a machine reads this
Read top to bottom, the file moves from identity (what and who) to attributes (what they know, where they work) to relationships (the other sources that corroborate them). That progression mirrors how clear information tends to be assembled: establish the entity, describe it, then connect it to the wider web.
Structured data like this does not replace good writing, and it is not a guaranteed ranking factor. What it can do is make the facts explicit, so they are less likely to be inferred incorrectly. The same discipline applies to ordinary prose — which is exactly what the AI-readable About page guide is about.
The same entity, decoded
Translated back into a sentence, the source file above reads as something a person — or an AI answer — could say plainly:
Tom Mason is the founder of AwarenessAI, founded in 2025, who works on AI visibility, generative engine optimisation, source trust and entity clarity. He studied at Lancaster University, is a member of UKAI, and can be found at tommason.co and on LinkedIn.
When the structured file and the human sentence agree, there is little room for drift. When they conflict — or the file is missing — the machine has to fill the gaps from elsewhere, and that is where misrepresentation creeps in.
Writing your own source file
You do not need to hand-write JSON to benefit from this thinking. The aim is simply to be able to fill in each field clearly and consistently — in your structured data and in your plain content alike:
Frequently asked questions
What is a "source file" in the context of AI visibility?
It is a way of picturing how an AI system tends to represent an organisation: as an entity with a type, a name, attributes and links to other sources. It is an illustrative concept, not a literal file inside any specific AI system.
Does structured data guarantee better AI visibility?
No. It can make key facts explicit rather than leaving them to be inferred, which may help, but it is not a guaranteed ranking factor and cannot correct inaccurate or inconsistent information elsewhere.
Do I have to add code to my website?
Not necessarily. Structured data helps, but the same clarity can be expressed in ordinary content. The thinking behind the source file matters more than the format.
Key takeaways
- Machines tend to see organisations as structured entities, not prose.
- Type, name, topics, facts, links and a clear summary all carry meaning.
- When your structured data and your writing agree, there is little room for drift.
- Structured data can help, but it does not guarantee visibility or fix bad information.
Related GEO Lab resources
Definitions of entities, structured data, schema and source trust.
Express the same clarity in your plain content.
References & further reading
- Schema.org — the structured-data vocabulary used in the example.
- Google Search Central — introduction to structured data.
- Tom Mason — independent research on AI source trust and entity interpretation.
Last reviewed: June 2026 · An experiment in the GEO Lab knowledge library.
