How often should you update a listicle to keep it cited by AI search? Our refresh routine

Every client with a "best of" page eventually asks how often it needs updating. We would like to hand over a number. We do not have a verified one, and neither does anyone who answers with a flat "every 90 days." So here is the routine we use instead, built on what changes and what our data shows.

Start with what goes stale

The right cadence depends on how fast the facts in the list move. For a list of software tools, prices and plan limits change often, and products disappear. For an evergreen how-to list, steps change when a tool updates its interface. For a list of local service providers, the details change slowly.

We tell clients to sort each list into one of three speeds:

• Fast: tools, pricing, products. Spot-check monthly, full review quarterly.
• Medium: vendor and service lists. Review quarterly.
• Slow: evergreen guidance. Review twice a year, or when something obviously changes.

These are our working suggestions, not findings from a study. We set tighter cadences for AI-tool lists because the products move quickly.

A real update, not a new date

The most common shortcut we see is changing the "last updated" date without changing anything. We do not do this, and we advise clients against it. Readers can tell, and engines can compare content between crawls. A real update means:

• Re-verifying each claim at its source, usually the vendor's own pricing or documentation page.
• Dropping entries that no longer exist and adding notable newcomers.
• Rewriting the verdict if the order has changed.
• Putting a date beside each fact we checked.

We time refreshes with data, not instinct

Each page has a citation life. It gets picked up, ramps, peaks, and fades. If you can see the fade starting, you can refresh before the page loses ground.

We use Promptwatch for this. Its Citation Trends view reports, per page, the first cited date, ramp-up days, peak citation rate and decay half-life. When we publish or refresh a list, we register the URL in the Page Tracker and watch it for the target prompts. A few months in, the decay half-life tells us roughly how long that page holds before it needs attention. That number belongs to that page, on those prompts, which makes it far more useful than a general rule.

We also track rival lists for the same prompts. If a competitor's refreshed list overtakes ours, we want to know the same week.

Refreshing efficiently

Content Refresh in Promptwatch gives diff-style suggestions for updating an existing page, so we can see proposed changes against the current text instead of rewriting from scratch. Our editors review every change and verify every fact before publishing. We do not hand fact-checking to a tool.

For clients on the Business plan ($579 a month), the Content Agent can plan and write against visibility gaps on a schedule, with a review inbox or a hands-off mode with a quality check. For listicles, we use the review inbox. A wrong price on a list that engines quote is a bigger problem than a slow update.

Scoping notes: Essential ($95) includes prompt tracking and content gap analysis. Professional ($245) adds the knowledge base and AEO articles. The automated Content Agent is on Business.

The routine we hand to clients

1. At publication: log the date, sources and target prompts.
2. Weekly: check whether the page is cited, using the Page Tracker.
3. Monthly: spot-check the highest-volatility facts, such as pricing.
4. Quarterly: full review, rewrite where needed, refresh the visible date.
5. Event-driven: acquisitions, pricing changes and big launches trigger a check within days.

Common mistakes

Date-only updates. Covered above. Avoid.

Updating everything equally. Put effort where the prompts are most valuable.

Updating without looking at the cited sources. If the engine cites a rival's list, read it. Often it covers an entrant or a fact we missed.

Treating a generic rule as truth. If a blog gives you an interval with no source, it is a guess. Ours included, until your own decay data replaces it.

What we tell clients it cannot do

No refresh schedule guarantees a citation. Engines weigh sources differently, answers vary between runs, and models update. What a good routine does is keep the facts right, which is worth doing regardless, and give you data to learn how quickly your own pages lose ground.

For your own lists

Pick your three most valuable lists. Log their prompts. Track them for two months. Look at when each one starts to fade. Then set a schedule that fits what you see, tighter for fast-moving topics and looser for slow ones.