Citation Trends: reading first cited, ramp-up and decay for each page

A page that earned twelve citations last month tells you little by itself. Did it earn them all in one week and then go quiet? Is it climbing? Was it cited constantly last quarter and now only occasionally? Those stories lead to very different decisions, and a flat count can't tell them apart.

That's what a time-based view of citations is for. Here is how we use it.

The metrics worth knowing

Promptwatch's Citation Trends classifies each citation by content type and source type, and then follows the citation rate over time. For each page it reports four timing measures: when the page was first cited, how many days it took to ramp up, its peak rate, and its decay half-life.

Each is a small piece of a page's life story:

First cited tells you how long a new page took to be noticed at all.
Ramp-up days tell you how quickly it gained traction once noticed.
Peak rate shows the best it ever did.
Decay half-life shows how long it takes to lose half of that rate.

We don't have sourced benchmarks for what a good half-life looks like, and we won't make one up. The comparison that counts is your own pages against each other.

Comparing pages in your own library

Sort your pages by decay half-life and look at both ends. Pages with a long half-life are holding their ground. Pages with a short one rise and fade fast.

Then ask what separates them. In our experience the durable ones tend to be those that answer a question completely and don't depend on a date or a news hook. The fading ones are often trend pieces, pages tied to a specific statistic or announcement, or pages where a competitor later published something stronger. That's a pattern we've noticed in client work, not a law, so check it against your own data.

Content type and source type

Citation Trends labels what kind of content is being cited and what kind of source it comes from. That helps when you're choosing what to produce. If your cited pages are overwhelmingly comparison pages and your how-to content gets nothing, there's a signal about what engines pull from you for the prompts you track.

Likewise, source type shows whether engines cite you from your own domain, from forums, from video, or from third-party publications. A brand cited almost entirely from its own site might be underexposed elsewhere. A brand cited mostly via others' pages might be leaving its own pages too thin.

Using ramp-up to set expectations

Teams new to this want to know how long until a new page gets cited. We can't give you a number, because we have no sourced figure and it varies by topic, engine and site. What we can do is measure your own site's ramp-up days across several pages and use that as your internal expectation.

After five or six pages you'll have a rough sense of how long things take on your domain. That's more useful than any industry average, and it stops you from judging a new page after two days.

Spotting a refresh candidate

A page whose citation rate peaked and has been sliding is a refresh candidate. Check the following before touching it:

Is the page still crawlable? Look for errors in crawler logs.
Did the facts change? Prices, features, rules, dates.
Did a competitor publish a better answer to the same question?
Did a third-party source that used to support the page change?

If the answers point to staleness, refresh it. Our content refresh playbook covers how. Promptwatch's Content Refresh feature gives diff-style update suggestions for existing pages, which saves a lot of rereading.

Spotting a protect candidate

The opposite case matters too. A page with a strong, long-lived rate is an asset. Protect it. Don't rewrite it for fun, don't change its URL without a redirect, and put a note on it saying why it matters. We've seen teams "tidy up" top-cited pages and lose them.

Put those pages on the Page Tracker in Promptwatch, which follows specific URLs, so you notice quickly if one drops.

Pair the trend with traffic

Citation rate alone doesn't prove value. Join it to visitor data. Promptwatch's 360 Insights combines crawler logs, citation clicks and AI-referred visits per page, and its Topic Insights point to where pages underdeliver for their topic. A page with a rising citation rate and no visits might be cited as background without being clicked. A page with a modest rate and strong conversions might deserve more attention than its citation count suggests.

Reading the chart without over-reading

Citation rates are noisy. A dip for one week isn't a decay. Use enough history, compare the same prompt panel, and avoid changing the panel mid-stream. If you add forty new prompts, the baseline shifts and the trend line lies to you.

Also remember that the engines change. A sudden drop across all pages and all models together is usually something on the engine side or in your access settings, not a page-level issue. Check crawler logs before you blame the content.

A monthly routine

Once a month, open Citation Trends and do four things:

List the five pages with the shortest half-lives.
List the five with the longest.
Note any new first-cited pages.
Decide one refresh and one protect action.

Put those on the Action Board so they're owned. Then repeat next month with the same panel.

If you're serious about this, track it in Promptwatch, because the timing metrics per page are built in and sit alongside crawler logs and visits. You'll stop asking "how many citations?" and start asking "which pages are fading and why?", which is a question you can do something about.