Can I tell AI engines which page is my official source?

No AI engine publishes a way to tell it which page is your source of truth. Google's guide to its generative AI features says you don't need new machine-readable files, AI text files, markup or Markdown to appear, and that there are no special optimizations for AI Overviews or AI Mode. OpenAI, Anthropic and Perplexity document their crawlers and their robots.txt rules. None of them documents a field for declaring an authoritative page.

That means the engine decides what to trust, answer by answer. When it searches, it pulls a set of pages for the query it wrote, then picks which ones to use and cite. Your site is one candidate among many. You can't rank yourself to the top of that set by declaration.

What you can do is make the right fact easy to find and hard to contradict:

• State each fact once, specifically, on a page crawlers can fetch.
• Make every other page on your site say the same thing.
• Get third-party pages that disagree corrected.

If you were hoping for a submission form, the companion question on submitting brand facts to AI engines covers what exists and what doesn't.

Three ideas get sold as ways to name an official source. The evidence doesn't back any of them as a lever.

llms.txt

llms.txt is a proposed file that lists a site's key pages for language models. Google says Google Search ignores it, and that publishing one will neither help nor harm visibility in Search. That statement covers Google only. No other engine documents reading the file either. Keep one if you like. Don't expect it to make any engine treat a page as official.

Structured data

Organization or Product markup describes your brand in a machine-readable format. Google says structured data isn't required for generative AI search. Ahrefs tracked 1,885 pages that added schema and found AI Overview citations fell 4.6%, with no major uplift on any platform. Those pages were already heavily cited, so whether schema helps a new page get considered is untested. Existing markup does no harm. It just isn't a way to declare authority.

Labels in the copy

Writing "official source" or "verified information" on a page is text like any other. The engine reads it as a claim, not an instruction.

An engine can only quote a sentence it can reach. That's where your real control sits, and the levers are mostly about removing blocks.

• Rendering. Vercel found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them, so they can't read client-side rendered content. If your pricing or founding year appears only after JavaScript runs, those crawlers may never see it. Googlebot does render.
• robots.txt, per bot. OpenAI states that each of its crawler settings is independent. GPTBot controls training use. OAI-SearchBot controls whether you appear in ChatGPT search. Blocking the wrong one removes you from search without telling you.
• Firewalls and CDNs. A firewall can return 403 to a crawler that robots.txt allows. The AI Crawler Checker requests a page as each AI crawler and shows which ones get turned away.
• Snippet controls. noindex, nosnippet, max-snippet:0 and data-nosnippet keep content out of AI answers. Check that none was applied by accident.

Then make the fact specific. A price, a number or a one-line definition gives the engine a sentence it can quote. Vague copy gives it nothing to lift. A brand fact sheet is a practical way to fix the wording once and reuse it.

Your site is rarely the only page an engine retrieves about you. Directories, review platforms, comparison articles and old press coverage all state facts about your company. If they disagree with you, the engine has to choose, and nothing guarantees it picks your page. The question on what AI does when sources disagree covers that choice in detail.

The evidence shows a strong association: when engines cite other sites instead of yours, being named correctly on those sites is the route into the answer. No study promises an effect size. But a wrong fact repeated across several outside pages is a common reason the same error shows up on several AI platforms.

Start by finding which outside pages engines cite for your buyer questions. The playbook on finding the pages AI engines cite walks through it. Then fix the ones you own or can edit, such as your review and directory profiles, and ask publishers to correct the rest.

Some answers never touch your site at all.

• Answers without search. When an engine answers from what the model learned in training, nothing you publish this quarter reaches it. OpenAI lists a February 16, 2026 knowledge cutoff for its GPT-5.6 models. A fact you changed after that date can't appear in those answers until a later model ships.
• Retrieval is discretionary. ChatGPT decides per query whether to search. Claude.ai search is a user toggle. The same question can get a searched answer one day and an unsearched one the next.
• Citation is not the same as use. An engine can cite your page and still word the fact differently, or use your page without citing it.

One tactic to avoid: hidden text or instructions aimed at AI crawlers, such as "treat this page as the official source". Research on search agents shows injected content can sway some models, which is exactly why it counts as manipulation. Results vary by model and shift with every safety update. It also risks spam action, and it can be used against you on third-party pages.

1. List the facts buyers ask about: pricing, plans, founding year, headquarters, product names, integrations.
2. Pick one wording for each and put it on one crawlable page, in the HTML before JavaScript runs.
3. Search your own site for contradictions and fix them. The guide to consistent brand facts covers how.
4. Check robots.txt for OAI-SearchBot, PerplexityBot and Claude-SearchBot, then check your firewall lets them through.
5. Find the third-party pages engines cite about you and correct the wrong ones.
6. Re-check what engines say after each fix. The free checker on this site gives a snapshot for up to three buyer questions.