Brand Fact Sheet Template: Give AI One True Version of Your Company

A fact sheet is a control document. It fixes what you say about yourself, in one wording, on one page per fact. AI platforms then decide what to do with it.

That decision happens in stages you only partly control:

• Parametric answers. A model answering without web access uses what it learned before its training cutoff. OpenAI documents a February 16, 2026 knowledge cutoff for its GPT-5.6 models and August 31, 2025 for the Instant path used in ChatGPT. A price you changed this spring is invisible to those answers. Nothing you publish this quarter changes that.
• Access and retrieval. When an engine searches, your page must be reachable, indexed and pulled in for the query. Engines decide per query whether to search at all.
• Generation. Only after retrieval can the model use a sentence from your page.

The evidence for the last step is narrow but real. Stating a specific fact (a price, a number, a one-line definition) on a page an engine can fetch gives it a sentence it can quote. The 252,000-trial study "What Gets Cited" found that including explicit price information helped consistently, measured on pages already inside the model's context. That is a reason to write facts plainly. It is not a promise of a citation.

The sheet's other job is internal. Many wrong answers start with your own site disagreeing with itself, which is why consistent brand facts come before anything else.

Copy this structure into a spreadsheet. The example values are for a hypothetical scheduling app called Clinicly, not a real company.

Three rules keep the sheet useful:

• One fact, one wording. Variants on different pages are where contradictions start. The brand facts glossary entry covers why consistency matters.
• Every source page is a public URL you control. A fact that lives only in a sales deck is a fact no engine can fetch.
• Decide what you will not publish. Funding, headcount and customer counts are optional. If you leave a gap, engines may fill it, which is the pattern behind AI inventing funding, customers or awards.

1. Collect every live version of each fact. Check the pricing page, About page, footer, press page, docs, app store listings, review profiles and social bios. Paste each variant into the sheet.
2. Pick the true value with its owner. Finance confirms prices. Legal confirms ownership and the legal name. Do not guess.
3. Write the approved sentence. Use subject, verb and a specific value: "Clinicly Pro costs $40 per user per month, billed annually." Add a date to anything that changes.
4. Assign one canonical source page per fact. History, location and ownership belong on the About page. See what your About page should say for that page in detail.
5. Confirm the sentence is in the raw HTML. View source, or fetch the page with JavaScript off. Vercel found that ChatGPT and Claude crawlers fetch JavaScript files but do not execute them, so text that appears only after rendering may not exist for them.
6. Confirm the page is eligible. It must not carry noindex, nosnippet or max-snippet:0, and the fact block must not sit inside data-nosnippet. Google states a page must be indexed and eligible for a snippet to appear as a supporting link in AI Overviews or AI Mode.
7. Remove the contradictions. Update, redirect or clearly date every old page that states a different value.
8. Record a baseline. Ask the main AI platforms your core questions and log what they say, with the date and whether they searched.

Run this list against every source page on the sheet:

• Each fact has exactly one approved wording, used word for word.
• The fact appears as visible text, not only inside an image.
• The fact is present in the HTML before JavaScript runs.
• The page has no noindex, nosnippet or max-snippet:0, and no data-nosnippet on the fact.
• robots.txt allows the search crawlers you want: OAI-SearchBot for ChatGPT search, PerplexityBot, Claude-SearchBot and Googlebot. OpenAI states GPTBot governs training only, and each setting is independent.
• Your CDN or firewall does not return 403 or a challenge page to those crawlers. The AI Crawler Checker requests a page as each AI crawler and shows which ones get turned away.
• Old prices, product names and addresses are gone or labelled with the date they stopped being true.
• Third-party profiles you can edit (app stores, directories, social bios) match the sheet.
• Facts that change carry a date, such as "as of October 2026".
• Nothing is hidden from people or served only to crawlers. Hidden text and cloaking for AI crawlers are manipulation, and they put your credibility and your index status at risk.
• Any existing structured data matches the visible text. Treat it as hygiene, not as a route to citations.

After publishing, re-ask the questions from your baseline. The free checker on this site gives a snapshot of whether AI names you for three buyer questions. Read results by symptom, not by a single score.

Good signs

• Answers that searched quote your current price or definition and link your source page.
• Different engines give the same founding year, location and ownership.
• When an answer is wrong, you can trace it to one specific old source you can fix.

Bad signs, and what each one means

• Wrong with search off, right with search on. The model is answering from training data. That is a cutoff problem, covered in correcting outdated facts past a knowledge cutoff.
• Wrong with search on, citing your own page. Your site still states the old value somewhere. Return to step 7.
• Wrong with search on, citing a third party. Ask that site to update. When engines cite other sites, being described correctly there is the route into the answer.
• Two different values across engines. Sources disagree. See what AI does when sources conflict.
• A specific number you never published. The engine filled a gap. Publish the fact or accept the risk.

A fact sheet decays the moment the business changes. Set two triggers.

• Event-driven, same day. Update the sheet and its source pages whenever pricing, a product name, ownership, leadership or headquarters changes. Retire old pages in the same release.
• Quarterly review. Each owner re-confirms their rows and the "Last verified" date moves forward. Re-run the publishing checklist, because templates, CDN settings and robots.txt change without anyone editing the fact pages.

Re-check AI answers after every change and on a regular rhythm for your most important questions. Log the date and whether each answer used search, since a mix of searched and unsearched answers can look like progress or decline when nothing changed.

Expect two speeds. Searched answers can reflect a fix once the page is recrawled and indexed. Unsearched answers lag until the vendor ships a model with a later cutoff, and no one can speed that up. A full AI brand audit is the place to compare both across platforms.