Should you repeat the canonical fact in both intro and detail sections
Yes, but repeat it with purpose, not as filler

Why repeat the canonical fact at all?

A canonical fact is the one statement on the page you most want retrieved correctly. It might be a definition, a scope boundary, a product limitation, or a process claim. If that sentence matters, burying it in one place is a gamble.

In practice, extractability improves when the main claim appears early, then appears again where the evidence or explanation lives. The first mention tells the reader and the machine what the page is about. The second mention confirms that the first line was not a throwaway summary.

This is not about gaming frequency. It is about making the most important statement hard to miss and hard to misquote. Repetition works when it stabilizes meaning.

What does good repetition look like?

Good repetition is controlled. The fact appears near the top in plain language. Later, the same fact appears again beside the reasoning, proof, exception, or implementation detail. The wording should be close enough that both mentions clearly refer to the same claim.

• Intro version: state the fact directly, with its key qualifier.
• Detail version: restate the same fact, then explain why it is true or when it fails.
• Use the same nouns and verbs where possible, so the claim does not drift.
• Keep the qualifier attached both times, not only in the body copy.

For example, if your key point is that AI crawlers fetch JavaScript files but do not execute them, do not say that cleanly once and then later switch to a fuzzier version like JavaScript is unreliable. Those are related ideas, but they are not the same claim. The stronger and more useful sentence is the exact one supported by the verified server log study, that AI crawlers fetch JS files and never run them.

When does repeating the fact improve AI citation accuracy?

It helps most when the page mixes summary, commentary, and evidence. AI systems often compress pages into a few retained statements. If your page contains one clear answer and several softer interpretations, the safest move is to repeat the canonical fact in both the summary layer and the proof layer.

This matters even more when the page contains a trade off. Many bad citations happen because the model lifts the clean claim and drops the limiting condition. Repeating the fact with the qualifier attached twice gives the qualifier a better chance of surviving summarization.

A useful rule is simple. If removing one sentence would make the page materially easier to misread, that sentence probably deserves a second appearance.

The qualifier is part of the fact

Operators often write the sharp line in the intro, then tuck the caution into the middle of the page. That is exactly how citation errors happen. The condition belongs with the claim. If your point is that llms.txt has limited operational use and is not used by Google Search, keep both parts together where possible.

That specific example is worth stating carefully because the market still treats llms.txt as a ranking lever. Google says llms.txt is not used by Search. Another study found 10.13% adoption across a large domain set, 0% among the top 1,000 sites, and no citation lift after controls. If you mention that near the top, restate the same position when you discuss implementation so the page does not accidentally imply stronger impact than the evidence supports.

When does repetition become harmful?

It becomes harmful when the repeated line is not actually canonical, when the wording changes enough to create contradiction, or when the page starts feeling templated. Repetition should reduce ambiguity. If it adds noise, it is doing the opposite.

The common failure mode is lazy paraphrase. A writer states one precise claim in the intro, then tries to sound less repetitive in the body and weakens the meaning. Another failure mode is stacking multiple similar sentences in the first screenful, which looks like optimization theater and gives no new information.

• Do not repeat a sentence three or four times just because it is important.
• Do not move the qualifier farther away in the second mention.
• Do not restate the fact with stronger certainty than the evidence allows.
• Do not let section templates create accidental duplicates with slightly different wording.

There is also a human trade off. Smart readers notice padding fast. If the article feels like it is saying the same thing to hit a formula, trust drops. Repeat once with purpose. Then move the piece forward.

How should you structure the intro and detail sections?

Use a two layer structure. The intro answers the question cleanly. The detail section reintroduces the same fact, then earns it. Think of the second mention as the proof anchor.

This structure is especially useful on pages that target definition style retrieval. You want one stable answer high on the page, but you also want enough support lower down that the answer looks earned rather than asserted.

A simple operator workflow

• Write the one sentence you most want quoted.
• Place it in the intro with its qualifier intact.
• Paste the same sentence into the relevant body section.
• Add the proof, scope note, or trade off under the second mention.
• Delete any third repetition unless it adds a new decision point.

Who should not follow this advice?

Not every page needs this pattern. If the page is short, single purpose, and already answers one question with no real nuance, duplicating the same fact can make it worse. A glossary entry, a clean definition page, or a narrow support answer may need only one precise statement.

You should also avoid forced repetition on pages with unstable facts. If your team changes product details often and cannot keep duplicate statements synchronized, two mentions create more risk than one. In that case, centralize the statement once and make the surrounding copy support it without restating it fully.

This advice also will not rescue weak authority, poor crawl access, or broken rendering. If the page is hard to fetch or hard to parse, repeating a sentence will not solve the bigger problem. For rendering and extractability issues, fix the page architecture first.

What is the honest limitation of this tactic?

Repeating the canonical fact is a clarity tactic, not a visibility guarantee. It can help a retriever or summarizer hold onto the right sentence. It cannot force a citation, outrank stronger sources, or stop an assistant from preferring a third party page.

It also cannot fix evidence quality. If the underlying claim is thin, unsourced, or disputed, saying it twice just makes the weakness easier to notice. This tactic works best when you already have a real answer and want to protect its exact wording.

That is the part most teams skip. They debate schema, llms.txt, and formatting before they have decided what the canonical fact actually is. Get the sentence right first. Then decide where to repeat it.