Citation inventory
Which of your pages currently earn citations, on which prompts, across which engines. You cannot protect a position nobody has written down.
A refresh cadence is the schedule that keeps citation-carrying pages current enough to hold their position. Retrieval-based engines favour recent content, so a page that earned a citation last quarter can quietly lose it to a newer competitor page saying much the same thing.
Typically ongoing, scheduled quarterly and reviewed monthly.
Retrieval-first engines re-fetch on nearly every query and prefer recent sources when several say similar things. Your page does not get worse — a newer one appears and wins the tie. Without a refresh cadence, citation share erodes gradually while nothing on your side appears to change.
This is the mechanism most GEO programmes overlook, because it is invisible in a monthly report. Nothing breaks. Traffic does not collapse. A citation you held for eight weeks simply stops appearing, and unless you are running a panel weekly you find out considerably later, if at all.
Perplexity is the clearest case because it retrieves fresh on almost every query, but the pattern holds elsewhere. Where several sources make comparable claims, recency is one of the tie-breakers, and it is the only tie-breaker you can influence on a schedule rather than by out-writing everyone.
The trap is that this invites the laziest possible response: change the date, republish, claim freshness. Engines evaluate the content, not the timestamp, and a page whose date moved while its substance did not gains nothing. It also trains your team to perform maintenance rather than do it.
A real refresh means the page says something it did not say before — a new figure, a changed recommendation, a correction, coverage of something that has happened since. That is more work than a date change and it is the only version that holds a citation.
The programme covers identifying which pages currently carry citations, scoring their decay risk, a scheduled refresh cycle prioritised by what you stand to lose, substantive updates rather than date changes, and monitoring that flags a citation loss within days of it happening.
Which of your pages currently earn citations, on which prompts, across which engines. You cannot protect a position nobody has written down.
Pages ranked by how exposed each citation is — how contested the topic is, how fast the underlying facts move, and how recently competitors published.
A schedule weighted by what you stand to lose rather than by page age, so effort concentrates on citations that are both valuable and at risk.
New figures, changed recommendations, corrections and coverage of what has happened since. Explicitly not date changes, which engines see through.
Panel monitoring that flags when a previously held citation disappears, so it is investigated within days rather than noticed at a quarterly review.
Which pages are no longer worth maintaining, consolidated or retired deliberately instead of being refreshed forever out of habit.
The cadence works on a rolling quarterly cycle: inventory the citations you hold, score each for decay risk, refresh the most exposed with substantive updates, and monitor weekly so a lost citation triggers investigation rather than waiting for the next review.
Every page currently earning citations, mapped to the prompts and engines it wins. This is the asset register the rest of the programme protects.
→ A citation inventory by page and prompt
How contested the topic is, how quickly its facts change, and how recently competitors published on it. High-value plus high-risk goes to the front of the queue.
→ A risk-scored refresh priority list
Each scheduled page updated with something genuinely new — a figure, a revised recommendation, a correction. If there is nothing new to say, that is a signal to consolidate rather than republish.
→ Updated pages with a change log
Weekly panel runs flag lost citations. A loss triggers a look at what displaced you, which frequently teaches more than the refresh schedule itself does.
→ Loss alerts and displacement analysis
The difference is whether the page now says something it did not say before. Engines evaluate content rather than timestamps, so a substantive update can recover a lost citation while a date change accomplishes nothing except making a maintenance report look busy.
| Substantive refresh | Cosmetic refresh | |
|---|---|---|
| What changes | Figures, recommendations, corrections | The published date |
| Effort | Hours of subject-matter input per page | Minutes, automatable |
| Effect on citations | Can recover a lost position | None — engines read the content |
| Effect on readers | The page is genuinely more useful | Erodes trust when readers notice |
| Selection method | Risk-scored by citation value | Oldest pages first, regardless of value |
| Sunset option | Consolidate or retire when exhausted | Refresh forever out of habit |
You need a refresh cadence when citations that you previously held have quietly disappeared, when your best-performing pages have not changed in over a year, or when nobody in the business can name which pages currently earn your generative visibility.
Refresh work is defensive and should be judged that way. Success looks like citation share holding steady rather than climbing, with recovered positions on individual pages. Clients who skip it typically see gradual erosion that nobody attributes correctly until a quarterly review.
Each answer is written to stand alone in 40 to 60 words — the shape an AI Overview or Perplexity citation lifts. Ships with FAQPage schema.
Quarterly for pages carrying valuable citations in contested topics, annually for stable reference material. Frequency should follow citation value and how fast the underlying facts move, not page age — refreshing the oldest pages first is the wrong prioritisation.
No. Engines evaluate the content rather than the timestamp, so a date change with no substantive update accomplishes nothing. It also erodes reader trust when someone notices a supposedly updated page saying exactly what it said last year.
Because retrieval-first engines prefer recent sources when several say similar things. Your page did not get worse — a newer competing page appeared and won the tie. Recency is the tie-breaker you can influence on a schedule.
The page says something it did not say before: a new figure, a changed recommendation, a correction, or coverage of something that has happened since. If there is genuinely nothing new to add, that is a signal to consolidate rather than republish.
Weekly panel monitoring flags it within days. Without that, losses are typically discovered at a quarterly review, by which point a competitor has held the position long enough to be entrenched and reclaiming it is considerably harder.
No. Pages carrying no citations and attracting no traffic should be consolidated or retired rather than maintained indefinitely. Part of this work is deciding what to stop doing, which is usually the recommendation clients least expect from an agency.
Thirty minutes with a senior strategist. We pull your live visibility while we talk and tell you plainly whether a content refresh strategy is what you need — or whether your problem sits somewhere else.