The Canonical Facts Page: One Page That Gives AI Something to Cite

TL;DR

One dated page on your own domain, 14 fields of plain-text company facts including what your product does not do. It makes the correct answer easy to cite, and cannot make a model look for it.

A canonical facts page is one page on your own domain that states every checkable fact about your company in plain text, with a date on it. It exists so that a retrieval system looking for your founding year, your category, your current plan names or what your product does not do finds one unambiguous answer instead of inferring one. It does not stop a model guessing. Retrieval is conditional, and training data has no correction channel.

Written against Google Search Central documentation, the Organization and SoftwareApplication structured-data references, Google's AI features guide and the Gemini and Claude search-tool documentation, all retrieved 2026-09-18. Structured-data requirements change; pin your reading to that date.

Key Takeaways

The page is for retrieval, not for buyers. Google's own best-practice list for AI features begins with "Making sure that important content is available in textual form" (Google Search Central, AI features and your website, retrieved 2026-09-18).

No markup is required to appear in AI features. Google states: "You don't need to create new machine readable files, AI text files, or markup to appear in these features. There's also no special schema.org structured data that you need to add" (same page, retrieved 2026-09-18).

Add Organization markup anyway, for search. It has "no required properties", and Google "does not guarantee that features that consume structured data will show up in search results" (Organization schema reference, retrieved 2026-09-18).

Date it visibly. Google asks you to "Add a user-visible date to the page and feature it prominently", with dates "consistent between user-visible and structured data, accurate, and not in the future" (Influence your byline dates, retrieved 2026-09-18).

The page cannot force anything. A model "analyzes the prompt and determines if a Google Search can improve the answer" (Gemini API, Grounding with Google Search, retrieved 2026-09-18) — so publishing the fact makes a correct answer possible, not certain.

What a canonical facts page is, and what it is not

One URL. Plain sentences. Every fact a buyer, a journalist or a retrieval system might need about your company, stated once, dated, with no marketing compression.

It is not your About page, which tells a story, nor a press kit, nor a pricing page built to convert. Those pages state facts as a side effect of doing something else, which is why they contradict each other — and your own site contradicting itself is the most common reason AI describes a company wrongly (7 reasons AI describes your product wrong).

The test for whether a sentence belongs here: could someone check it, and would they get the same answer next quarter? If yes, it belongs, with the date it was last confirmed.

Why it helps, and the limit you should state out loud

Three documented mechanics make this page worth writing.

Text is what gets read. Google's best practices for AI features list "Making sure that important content is available in textual form" and "Making sure your structured data matches the visible text on the page" (AI features and your website, retrieved 2026-09-18, SOURCED). A fact inside an image, a PDF or a JavaScript tab is a fact you have not published.

Search may reach pages that do not rank for your name. Both AI Overviews and AI Mode may use a "query fan-out" technique — "issuing multiple related searches across subtopics and data sources" (same page, retrieved 2026-09-18, SOURCED). A page that answers who owns X or what does X not do can be retrieved by a search you never considered (ANALYSIS).

Retrieval is conditional, and that is the limit. Google's grounding documentation describes the model analysing the prompt and determining "if a Google Search can improve the answer" (Gemini API, Grounding with Google Search, retrieved 2026-09-18, SOURCED). Anthropic's web-search documentation says triggering "is steerable through your system prompt" and offers max_uses "to cap the number of searches for each request" (Claude web search tool, retrieved 2026-09-18, SOURCED) — a ceiling, not a floor.

So the honest claim is narrow: this page makes the correct answer available and easy to quote. It cannot make a model look, and it cannot reach what a model absorbed in training. The routes that are documented, and the ones that are not, are set out in AI misinformation about your brand.

The 14 fields

#	Field	Why it is on the page
1	Registered legal name	The entity name in contracts and filings, which is often not the name you trade under
2	Trading names and former names	A rename leaves years of pages using the old one
3	Founding date	Guessed more often than any other company fact
4	Headquarters, and any other locations	Where are they based is a buyer question and a compliance question
5	Category, in ordinary words	If you never state it, the model infers one from whichever pages yours resembles
6	What the product does not do	The sentence almost nobody publishes, and the one that stops the worst errors
7	Products and current plan names	Exact strings, including capitalisation
8	Prices, each with an as-of date	Undated prices are what get quoted back to you two years later
9	Integrations and supported platforms	Frequently asserted from an old roundup
10	Headcount, as a band	A band you will still be able to confirm beats a precise number you will not
11	Leadership, with titles	Changes silently and gets repeated for years
12	Ownership, funding or public status	Answers are they a startup, or owned by someone
13	Certifications and audit status, including what you do not hold	Overstating this is a legal problem, not an SEO one
14	Press and correction contact	Gives a publisher somewhere to send a correction request

Two rules for the whole table. State the uncertainty where one exists — figure as of Q2 2026 is a fact; a stale precise number is not. And never publish a certification you are working towards as though you hold it.

The skeleton

Paste this, fill it in, and keep the headings stable so the anchors and quotations stay valid.

Plan	Price	What it includes
Free	$0	...
Team	$29 / user / month	...

# Company facts — Example Corp Last updated: 2026-09-24. Maintained by: [email protected] ## Identity - Registered legal name: Example Corporation Ltd - Trading name: Example - Former names: Formerly Widget Labs (renamed 2024-03) - Founded: 2019-04-16 - Headquarters: Bristol, United Kingdom. Other offices: none - Ownership: Privately held. Series A, 2023. Investors: ... ## What we are Example is a [category] for [audience] that does [function]. ## What we do not do - We do not [the thing people assume] - We do not [the adjacent category we are mistaken for] - We are not available in [region], and we do not offer [feature] ## Products and plans (prices as of 2026-09-24) Plan Price What it includes Free $0 ... Team $29 / user / month ... ## Integrations Supported today: ... · Not supported: ... ## People - Name — Chief Executive (since 2021) ## Compliance - Hold: ... - Do not hold: ...

## Corrections Something wrong on this page, or in an article about us? [email protected]

Dating it, and keeping it true

A page like this decays into a liability the moment it goes stale, so the maintenance rule matters more than the markup.

Show the date. Google asks site owners to "Add a user-visible date to the page and feature it prominently", and says dates should be "consistent between user-visible and structured data, accurate, and not in the future" (Influence your byline dates in Google Search, retrieved 2026-09-18, SOURCED).
Re-confirm on events, not on a calendar alone. A price change, a plan rename, a launch, a leadership change, a funding round, a rename, a new region. Each one is a trigger to update this page first, before the marketing pages (ANALYSIS).
Re-read the whole page quarterly and move the date only when you have checked every line. A lastUpdated bump with no check asserts freshness you did not verify.
When a price changes, fix the other surfaces too. The full sweep is in AI is quoting your old pricing.

Where to link it from

Treat it as a destination, not a buried page: link it from the footer sitewide, from the About page, from the docs landing page, from your press page, and from the profile links on accounts you control, and add it to your sitemap. Inside the page, link out to the primary sources for facts that live elsewhere — your filings, your status page, your security page.

The markup question, answered honestly

Add Organization structured data because it is cheap, correct, and machine-readable for classic search. Do not buy anything sold on the promise that markup improves AI visibility.

Google's Organization reference states there are "no required properties; instead, we recommend adding as many properties that are relevant to your organization", and the same documentation set says Google "does not guarantee that features that consume structured data will show up in search results" (Organization schema reference, retrieved 2026-09-18, SOURCED). Useful properties for this page include legalName, alternateName, foundingDate, numberOfEmployees, email, telephone, logo and sameAs.

For AI features specifically, Google is explicit: "You don't need to create new machine readable files, AI text files, or markup to appear in these features. There's also no special schema.org structured data that you need to add", and "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary" (AI features and your website, retrieved 2026-09-18, SOURCED). The same reasoning killed a lead magnet we had planned — see why Google removed the FAQ rich result.

Two traps if you mark up your product or prices on this page.

SoftwareApplication rich results require a rating or a review: "You must include one of the following properties: aggregateRating" or a review (SoftwareApplication structured data, retrieved 2026-09-18, SOURCED). If you do not publish ratings, the markup is still valid data and simply will not produce that rich result.
A past date in priceValidUntil can suppress the listing: "Your listing may not display if the priceValidUntil property indicates a past date" (merchant listing structured data, retrieved 2026-09-18, SOURCED). Either maintain the date or leave the property out.

Here is the JSON-LD object. Publish it inside a JSON-LD script block in the page head — the object itself is all you need to paste into your template or CMS field:

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Example",
  "legalName": "Example Corporation Ltd",
  "alternateName": "Widget Labs",
  "url": "https://example.com/",
  "logo": "https://example.com/images/logo.png",
  "foundingDate": "2019-04-16",
  "email": "[email protected]",
  "description": "Example is a [category] for [audience] that does [function].",
  "sameAs": [
    "https://www.linkedin.com/company/example",
    "https://github.com/example"
  ]
}

Keep every value identical to the visible text above it. Google's own best-practice line asks for structured data that matches the visible text on the page, and a mismatch is the one way this markup can actively hurt you (ANALYSIS).

How to tell whether it is doing anything

Ask the questions a buyer would ask, in several phrasings, and record the date, the model as the product names it, whether search was used, whether the answer cited anything, and whether it cited you. The prompts are in 40 prompts to check what AI says about your brand; the recording method, so two runs are comparable, is in how to find out what ChatGPT says about your company.

Two outcomes are worth separating. If answers start citing this page, it is working. If answers stay wrong but never cite anything, the error is upstream of retrieval and no amount of publishing will move it (ANALYSIS).

How This Guide Was Sourced

Written and maintained by the LogicBalls editorial team (logicballs.com). Disclosure: LogicBalls builds AI writing tools. This is a page you write once, by hand, and no product is needed for it.

AI involvement. AI assisted with research and drafting. Every quotation above was matched against the linked documentation page as fetched on 2026-09-18, and the structured-data requirements were read in Google's current references rather than carried from an earlier draft.

Sources. Google Search Central: the AI features guide, the Organization, SoftwareApplication and merchant listing structured-data references, and the byline dates guide. Provider search documentation: the Gemini API's Grounding with Google Search page and Anthropic's web search tool page. Google's "Who, How, and Why" guidance on people-first content informed the identity fields.

What is not claimed. We publish no measurement of how often a facts page changes an AI answer, no timeline for a change to appear, and no ranking benefit. There is no downloadable template — the skeleton above is the template, and it is on this page because we would rather publish the thing than a form in front of it.

What could not be verified. No provider documents preferring a company's own domain over a third-party page when both are retrievable, so the ordering advice here is reasoning, not documentation, and is marked as such.

No LogicBalls telemetry is used in this guide. Every figure above is external and linked.

Frequently Asked Questions

Is this just an About page?

No. An About page is written for people and usually tells a story. This page is written to be quoted: short declarative sentences, one fact each, dated, including the facts that are commercially unglamorous.

Will it stop AI getting us wrong?

No, and any page promising that is overselling. It makes the correct answer available and easy to cite. Whether a model retrieves it is the model's decision, and what it learned in training has no correction route.

Do we need schema markup on it?

Add Organization markup for classic search — it has no required properties and costs nothing. For AI features Google says plainly that no special markup or machine-readable file is needed.

What about an llms.txt file?

Google's changelog entry of 15 June 2026 says these files "aren't needed for Google Search (and won't negatively or positively impact your visibility or rankings)", adding that "it's fine if you want to maintain these files for other services or systems that use them" (Search Central documentation updates, retrieved 2026-09-18). Publish one if another service you use consumes it, not as an AI-visibility tactic.

How often should we update it?

On every event that changes a fact, and a full re-read quarterly. Move the visible date only when you have checked every line, because an unverified date claims a freshness you have not earned.