Acclaira

Chapter Fifteen

Freshness is a habit, not a date field

AI answers skew towards recent content, which tempts people into faking dates. That does not work. What works is a small, boring maintenance routine.

Chapter 15 of 26 · 2 min read

The analogy

A restaurant reprints its menu with today's date but serves the same tired dishes. Nobody is fooled twice. The restaurant next door changes three dishes a season and quietly becomes the one people recommend.

What the evidence supports

There is a real recency effect, and it is more modest than the industry claims. Ahrefs' analysis of around 17 million citations across seven AI platforms found cited pages averaged 1,064 days old, against 1,432 days for traditional organic results — a meaningful skew towards fresher content, but note that both averages are measured in years. TIER B

You will also see claims that content under 30 days old earns "3.2× more citations", or that a refresh programme moved a citation rate from 12% to 47%. TIER C I could not trace a method behind either. Ignore them.

The honest summary: recency helps at the margin; it does not override relevance or evidence. A three-year-old page with a unique number will beat a one-week-old page of generic advice.

What actually counts as a refresh

Changing the date is not a refresh. It is worse than nothing, because if it becomes a site-wide pattern it is trivially detectable and it destroys the credibility of your dates generally. A refresh means at least one of:

  • New data, new figures, or updated prices.
  • A regulation, process or requirement that has changed.
  • A new section answering a question you have started being asked.
  • Removal of something that is no longer true.

When you do make a substantive change, then do update both the visible "last updated" date on the page and the dateModified field in your Article schema, and keep them consistent with each other.

The bulk-restamping trap

Several publishing tools and SEO plugins will helpfully set dateModified to the build date across your entire site every time you deploy. Your whole archive then claims to have been updated this morning. This is the single most common way a well-intentioned site ends up with dishonest dates, and nobody notices because no human reads schema. Check what your build actually stamps.

A maintenance routine that fits a real business

1
Identify your twenty pages that matter

The ones tied to revenue. Ignore the rest.

2
Review five of them each quarter

Twenty pages, four quarters, five a quarter. Each one gets looked at annually. That is a sustainable rhythm, not a project.

3
Ask three questions per page

Is anything on it now false? Is there a number that should be this year's? Is there a question we now get asked that this page does not answer?

4
Change something real, then update the date

In that order. Always in that order.

5
Push the change to Bing via IndexNow

So the update reaches ChatGPT's retrieval layer without waiting for a crawl.

If you only do one thing

Put a recurring quarterly appointment in the calendar to review five pages. The routine matters far more than the individual edits.

Read the whole book

Answering Machines is 26 chapters on being the answer when your customer stops searching and starts asking. Free, ungated, and every figure is sourced and confidence-rated.