Refresh or New Page? The AEO Decision Guide
Refresh the existing page when it already ranks or gets cited for the question's topic, when the gap is depth or currency rather than subject, and when the page's URL has inbound links or citation history — equity a new URL starts without. Publish new when the question is a genuinely distinct intent (different buyer, different job), when the existing page already answers a different question well (stuffing two intents onto one URL dilutes both), or when the old page's format can't carry the answer shape the question needs. The tiebreaker: one page per distinct question a buyer would ask in one sitting; never two pages competing for the same one.
Every content-gap review ends at the same fork: the question is real, a page half-covers it — do you deepen that page or write a new one? Teams default to 'new' because publishing feels like progress and editing feels like maintenance. For AI-answer visibility, that default is usually backwards: answer engines cite pages with retrieval history and link equity, and a fresh URL has neither on day one.
The decision table
| Signal | Points to refresh | Points to new page |
|---|---|---|
| Existing page's relationship to the question | Same topic, shallow or stale answer | Different question that happens to share keywords |
| Buyer and job behind the question | Same buyer, same decision | Different buyer or different stage |
| URL equity | Page has links, citations, ranking history | Page is weak with nothing to lose |
| Answer shape needed | Fits the page's existing format | Needs a format the page can't carry (tool page, template, comparison) |
| Risk of self-competition | New page would chase the same query | Intents are distinct enough to coexist |
The equity math most teams skip
A page that already gets crawled, ranks for adjacent queries, or appears in AI citations has paid its cold-start cost. A refresh inherits all of it the day it ships; a new page re-pays it over months. The exception that proves the rule: when the existing page is itself weak — no links, no impressions, no citations — there's no equity to inherit, and the refresh argument collapses. Check the numbers before deciding; the gut feel that 'the old page is fine' is usually wrong in both directions.
The cannibalization trap
The worst outcome isn't choosing wrong once — it's the archive that accretes near-duplicate pages, each half-answering the same buyer question, splitting links and confusing retrieval. Answer engines pick one source per claim; two of your pages competing for the same question means you've halved your own odds twice. Before any new page ships, search your own site for the question — if a page comes back that a reasonable buyer would accept as the answer's home, that page gets the update. And when you do refresh, update the date honestly only when the content genuinely changed — engines and readers both learn quickly which sites' dates mean anything.
Doing the refresh right
A citation-worthy refresh is surgical: put the direct answer where the page's structure promises it, update the claims that aged, add the missing sections as real sections (not a bolted-on FAQ that restates the intro), and keep the URL. Log the change — what was added, when, why — so movement in citations can be attributed. If the refresh rewrites more than half the page, ask if you are actually writing the new page under an old URL; sometimes you are, and that's fine — the URL keeps its equity either way.
FAQ
Does updating a page improve AI citations? — Refreshed pages with genuine additions are the standard path to citation improvement — the URL keeps its retrieval history while the content closes its gap. 'Updated' dates on unchanged content do nothing except erode trust in your dates.
How many questions can one page answer? — As many as one buyer would ask in one sitting on one topic. The practical ceiling is coherence: when the page needs two different intros for two different readers, it's two pages.
Should I delete the weak page instead? — If it answers nothing distinct and has no equity, consolidate it into the stronger page with a redirect. A thin page competing with your own better answer is pure liability.
Put this to work on your own website.
Fenn finds what your customers ask, drafts the articles and site fixes, and measures what ChatGPT, Claude, Gemini, Perplexity and Grok say about you — with every change waiting for your approval.