The grounding page: one URL that states your facts

An engine answering a question about your company assembles the facts from whatever it retrieves: your site, a directory entry written three years ago, an old funding announcement, a review site summary nobody has touched since launch, a competitor's comparison table. It has no reliable way to know which of those is current. Where the sources disagree, the answer inherits the disagreement. A grounding page is the direct response: one canonical URL, plainly written, stating what is true, so retrieval has an authoritative version to lift from.

The problem is contradiction, not absence

Most companies have their facts on the web several times over, in several vintages. A price that changed, a product that was renamed, a founder who left, an office that moved, a positioning statement from an earlier stage of the business. None of those pages disappear when the fact changes. An engine choosing between them is not choosing between truth and falsehood. It is choosing between two plausible sources, and the older one is often on a domain with more history than yours.

Publishing the current version in one obvious place, phrased the way the question would be phrased, gives retrieval something to prefer and gives every third party a single thing to copy.

What the page states

• What the company is. The category, in one sentence, using the words your market uses rather than the words your positioning deck uses.
• What it sells. Named products or services, and what each one does.
• Who it serves. Customer type, industry, company size, the job the buyer is doing.
• Where it operates. Countries, regions, languages supported, and any place it explicitly does not serve.
• Pricing posture. Whether prices are public, the model in use, and the entry figure if you publish one.
• Founding facts. Founding year, legal name, headquarters, ownership status.
• Named people. Founders and executives, with roles.
• What it is not. The adjacent category you do not serve, the product you are routinely confused with, the same-name company in another industry. This item does the disambiguation work described in entities and how an engine resolves your brand.

How to write it

Write short declarative sentences that name their own subject. A sentence that begins with "it" or "the company" loses its subject the moment it is lifted out of the page, which is exactly what extraction does. Put the fact first and the elaboration after, the same answer-first pattern that makes any passage quotable, set out in how to get cited by AI engines.

Two rules matter more than the rest. Use specific dates and figures rather than ranges and hedges, because a specific claim can be attributed and a hedge cannot. Then drop the marketing register entirely. Adjectives do not survive extraction as facts. A sentence claiming a leading platform for growing teams contains nothing an engine can attribute to you, while a sentence naming your founding year, your products and the countries you sell in contains three facts a reader can check.

State a last-reviewed date on the page in plain text. It tells a reader and a retrieval system when the facts were last checked, and it makes a stale page visibly stale rather than silently wrong.

How to wire it up

• Link it from the homepage and the about page. A page nothing links to is a page crawlers reach late and rank low.
• Include it in the sitemap with an accurate lastmod value.
• Attach Organization markup here or on the about page. Google's Organization documentation lists no required properties, recommends adding as many relevant ones as apply, and recommends placing them on the home page or a single page that describes the organization. Use sameAs to list the external profiles you control.
• Make the third-party profiles agree with it. Same name, same description, same category, copied rather than rewritten.
• Keep it eligible. Per Google's AI features guide, a page must be indexed, crawlable and snippet-eligible to appear in AI Overviews and AI Mode, and no special markup or AI file substitutes for that. A grounding page behind a noindex tag does nothing.

Keeping it true

Put the page on a review cadence and record the date each time, whether or not anything changed. Quarterly is enough for most companies, and any pricing change, product rename, funding event or executive change should trigger a review immediately rather than waiting for the cycle.

The second half of the job points outward. When a fact changes, the third-party records repeating the old version are what an engine will still find. Work through the profiles, directories and partner pages that describe you and submit corrections, starting with the ones that rank for your brand name. The general mechanics of updating published content are covered in freshness, and what the markup does elsewhere in structured data.

What it does not do

Publishing a grounding page does not obligate any engine to read it, prefer it or repeat it. No major engine documents a mechanism for nominating a source of truth about yourself, and none publishes how it weighs a first-party statement against a third-party one. Anyone claiming a page like this guarantees accurate AI answers is describing something no vendor has documented.

What the page does is narrower and still worth the hour it takes. It makes the correct version available in one retrievable place, and it removes the contradictions that originate on your own property. Availability and consistency are the parts a publisher controls. Selection is not.