Home › AI Search › Citation ready content
How to Write Content an AI Will Cite
An answer engine is looking for a passage it can quote without needing the rest of the page, so the writing that wins is the writing that survives being lifted out of context. Six rules cover most of it: answer the question in the first sentence, keep each idea in a self contained unit, put a real named author on the page, say when it was last checked, include at least one piece of evidence that exists nowhere else, and be explicit about what your answer does not cover.
Published 26 August 2026 · Sources checked 20 August 2026
Rule one. Answer in the first sentence
Most business writing builds towards its point. Citation ready writing leads with it and explains afterwards. The test is simple: delete everything except your opening paragraph, show it to somebody who has never seen the page, and see whether they now know the answer.
Here is the change on our own home page, which failed this test badly until this year. The heading was the two words "Marketing Agency" and the opening line was "We help ambitious businesses rise, grow, and thrive through strategic marketing, powerful branding, and results-driven digital campaigns." That is a slogan. It tells an engine nothing it could quote in reply to a question, because it is not an answer to one.
Rule two. One idea per self contained unit
If a paragraph only makes sense because of the paragraph before it, it cannot be lifted. Write so that each section stands alone, with the subject named rather than referred to as "it" or "this". Sub headings are load bearing here, because they tell an engine where one idea ends and the next begins.
This is also why very long articles often underperform short ones in AI answers. Length is not penalised, but burying five answers inside three thousand words of continuous prose makes all five harder to extract than five clear pages would be.
Rule three. A real named author
Anonymous pages are harder to trust and easier to skip. Google's guidance on helpful content is explicit that of experience, expertise, authoritativeness and trust, trust is the most important. A named author with a real profile is the cheapest trust signal available, and the overwhelming majority of UK agency and business sites still publish everything under an account called admin.
Rule four. Say when you last checked
In a field that changes every few weeks, an undated page reads as stale whether or not it is. A visible last checked date, genuinely maintained, does two things: it tells a reader how much to rely on the page, and it gives you a reason to actually revisit it. Publishing a date you do not honour is worse than publishing none.
Rule five. One thing that exists nowhere else
This is the only rule a competitor cannot copy. Original evidence means your own numbers, your own test, your own survey, your own worked example, including the parts that did not go well. Everything else on a page can be researched and rewritten by anybody. First party evidence cannot.
It does not have to be a large study. A single honest figure from your own work, with the method stated and the date attached, is more quotable than a thousand words of general advice.
Rule six. Be explicit about scope
Say what your answer does not cover, and where the evidence is thin. Writers avoid this because it feels like weakness. It reads as the opposite. A page that says "this applies to UK service businesses, and we have not tested it on ecommerce" is more useful, and more obviously honest, than one that implies universal application.
It also protects you. Jono Alderson has made the point that AI systems do not simply repeat the story a business tells about itself, and can test that story against documentation, reviews, competitors, forums and pricing found elsewhere. An overreaching claim on your own site is now checkable in a way it was not five years ago.
Which pages to rewrite first
Not all of them, and not in the order you would expect. Start with the pages that already rank, because independent analysis puts ordinary search rank near the top of the factors associated with being cited, and a page already in the top ten is the most likely candidate for extraction. Rewriting a page nobody can find does not make it quotable, it makes it a well written page nobody can find.
- Pages already ranking in the top ten for a question, which are the nearest to being cited.
- Pages answering a definitional or comparison question, which is the shape of query assistants handle most.
- Your most commercially important service page, which is worth the effort regardless of where it ranks.
- Everything else, slowly, as part of normal maintenance rather than a project.
Does this make writing sound robotic?
It does if you apply it mechanically. The rules are about structure, not about voice. Answering first, keeping ideas separate and being honest about limits are qualities of good explanatory writing generally. They were what a decent editor asked for long before any of this existed.
Read next
Frequently asked questions
How long should an answer first opening be?
Sixty to a hundred words is the range we use. Long enough to answer properly, short enough to be lifted whole. If you cannot answer the question in a hundred words, the heading is probably asking two questions and the page should be two pages.
Do we need to rewrite everything we have published?
No, and trying to is how this becomes a project that never finishes. Rewrite the pages that already rank for a question, plus your most important service page. Leave the rest to normal maintenance.
Will an AI credit us if it uses our writing?
Often not. Research by Semrush found almost 62 per cent of citations were ghost citations, where the page was used as a source link but the brand was never named in the answer. Write for the outcome you can control, which is being useful enough to be used.
Should we add an FAQ section to every page?
Only where the questions are real ones customers ask, and only where each question appears on one page and nowhere else. A shared FAQ block repeated across dozens of pages is the exact pattern Google's scaled content abuse policy describes.
Not sure how AI assistants describe your business?
We run a hand checked AI visibility audit, scored out of 100 across five areas, with the weightings published so you can argue with them.
Book a 15 minute callSources
- Google Search Central. Creating helpful, reliable, people-first content. Updated 10 December 2025.Of experience, expertise, authoritativeness and trust, Google states that trust is the most important.
- Zyppy, Cyrus Shepard. AI Citation Ranking Factors Analysis. 7 May 2026.54 sources, 23 factors scored. URL accessibility and search rank scored highest.
- Semrush with Kevin Indig, Growth Memo. Why 62% of AI citations don't lead to brand mentions. 9 June 2026.3,981 domain appearances, 115 prompts, four engines. 61.7 per cent ghost citations.
- Jono Alderson. SEO vs GEO is the wrong question. 28 July 2026.On corroboration, and why scattered inconsistent mentions do not help.
- Google Search Central. Spam policies for Google web search. Updated 15 May 2026.Scaled content abuse and doorway abuse, in Google's own wording.
Every figure on this page was read on the source page listed above. This page is reviewed monthly and the last checked date is updated when it is.
