Content patterns that trigger LLM citations: claim and evidence structures

If we had to pick the single content pattern that most reliably moves an LLM citation rate, it would be claim and evidence. Stated as a claim. Followed by specific, verifiable evidence. Repeated throughout the page in dense sequence.

The pattern is not new. Journalism teaches it, legal writing teaches it, technical writing teaches it. What is new is the cost of ignoring it. Maritime marketing copy that drifts away from claim-and-evidence into general persuasive prose is not just weaker rhetorically; it is invisible to the retrieval and generation pipeline that increasingly decides who gets quoted.

What the pattern looks like in maritime

Bad: persuasive prose

Our deep expertise in technical management, built over decades of partnership with leading shipowners, allows us to deliver operational excellence across complex global fleets while ensuring full compliance with the most demanding regulatory standards.

Three claims, none falsifiable, no extractable evidence. An LLM reading this paragraph cannot quote anything specific.

Good: claim and evidence

We have provided technical management since 1994. As of Q1 2026, we manage 84 vessels across tanker, bulker and gas carrier fleets, all above 5,000 GT. We hold ISM DOC certification across all three vessel categories, audited annually by DNV. Vessels are registered under Cyprus, Marshall Islands, Liberia and Panama flags.

Same paragraph length. Six concrete claims, each backed by specifics. Every sentence is extractable as a standalone factual statement.

The difference in citation behaviour between these two paragraphs, on a service page that competes for “VLCC technical management” or “ship management for tanker fleets” prompts, is large enough that we have rebuilt entire site content programmes around the second pattern.

Where to deploy claim and evidence

Service page openers

The first paragraph of every service page should be claim and evidence. If a buyer can read those two or three sentences and know what you do, who you do it for and what scale you operate at, the parser can quote them and the page enters citation range.

Capability bullets

Lists work well in claim-and-evidence form. Each bullet starts with the claim and adds the evidence inline.

• Crew management for Filipino and Indian officers and ratings, drawn from our Manila and Mumbai offices, with 1,200 active seafarers as of 2026.
• Newbuilding supervision at Korean (HMD, HHI), Japanese (NYK, Mitsui) and Chinese (Yangzijiang, NTS) yards, with 14 supervisors currently on assignment.

FAQ answers

FAQ blocks are pure claim and evidence by design. The question is the prompt; the answer is the claim plus the evidence. This is why FAQPage schema is so heavily cited: it matches the structure LLMs are trained to reproduce.

Case studies

Case studies should be a sequence of claim-and-evidence pairs: the situation, the action, the result, each with concrete detail. “We retrofitted twelve bulkers for EEXI compliance between 2023 and 2025, achieving an average reduction of 14% in attained EEXI through engine power limitation and propeller modifications.” That sentence is citation-grade. “We helped our client achieve compliance” is not.

Why this works at the model level

LLMs are statistical reasoning systems built on examples of well-structured factual writing. When they generate a citation-bearing answer, they look for source passages that match the structure of what they are trying to say. A passage that is already in claim-and-evidence form fits that structure directly. A passage of persuasive prose has to be paraphrased, hedged or summarised, and the model often skips it in favour of a cleaner source elsewhere.

This is the same reason Wikipedia gets cited so heavily: the encyclopaedic style is claim and evidence by convention. You do not need to write like Wikipedia. You do need to match its structural discipline on the parts of your site that matter for AI search.

A test before you publish

Take any paragraph you have just written. Ask: if someone read just this paragraph, with no surrounding context, what specific facts could they extract and verify? If the answer is fewer than two, rewrite the paragraph. If it is three or more, you are probably in citation range.

The bar is not high. Most maritime marketing copy fails it. The brands that pass it consistently across their site outperform their competitors in AI search by margins that are difficult to overcome through other means.