Tracking content freshness for AI search: the tool and routine we use
How we track content freshness for AI search on client sites, using citation decay per page and crawler logs, and the tool we run that routine on.
Clients ask whether they need to keep updating old content for AI search. Our honest answer is that we cannot prove a model prefers fresher pages. What we can do is watch whether our pages keep getting cited, and test whether refreshing helps. That turns a vague worry into a measurement.
What we track
Every page has a citation life. It begins to get cited, ramps up, peaks, and often fades. For freshness, we want per-page answers to a few questions.
When was the page first cited? How long did it take to ramp up? What was the peak citation rate, and how quickly has it dropped? Are AI crawlers still fetching it, or hitting errors? After we refreshed it, did the rate recover?
We also look at the competition: which pages are cited for the same prompts, and how recently do they appear to have been updated.
Why crawler logs belong in a freshness check
A page can look stale in the data when the real issue is access. A bot error, a redirect chain or a blocked path produces the same symptom, falling citations. Before rewriting anything, we check whether GPTBot, OAI-SearchBot, ClaudeBot or PerplexityBot are still fetching the page cleanly.
The tool we use
We run client programs on Promptwatch, and freshness work is one of the places it helps most. Citation Trends classifies every citation by content type and source type and shows citation rate over time, with first cited, ramp-up days, peak rate and decay half-life per page. It is the closest thing we have found to a freshness curve for AI search. Page Tracker follows specific URLs, ours or a competitor's. Agent Analytics shows IP-verified crawler visits with a crawl-to-citation path and error explanations.
For the fix, Content Refresh suggests diff-style updates to existing pages, content gap analysis shows missing topics, and the Action Board puts the work in priority order. Publishing goes to WordPress, Webflow and Framer. On Business plans, the Content Agent can plan, write and publish on a schedule, with a review inbox or a hands-off mode.
Pricing as of late October 2026: Explore is free (10 prompts, ChatGPT only), Essential $95 a month, Professional $245 and Business $579, with agency plans at $199, $399 and $799. Crawler logs start on Professional.
Other options
AthenaHQ offers an Action Center and content agent. Scrunch AI offers audits and bot reporting. Profound is enterprise-focused. On-page tools such as Surfer and Clearscope help with rewriting but do not track AI citations. We use these when a client already has them, and keep citation tracking elsewhere.
Our refresh process
- Sort pages by decay and by business value.
- Check logs on the worst performers to rule out access issues.
- Refresh with substance: new data, corrected claims, answers to the follow-up questions the fan-outs reveal.
- Re-publish and note the date.
- Watch the page's citation rate over four to eight weeks.
What a refresh will not do
Changing a date without changing the content does nothing useful, and we advise against it. A refresh also does not guarantee a citation, since models vary run to run. We report the outcome either way, because a refresh that did not help tells you where to look next.
If you only track one thing this month, track which of your top ten cited pages are losing citations.
What we report to clients
Each month we show three lists: pages gaining citations, pages losing them, and pages refreshed in the last period with their before and after rate. The third list matters most, because it is the only one that tests our own work. Some refreshes help, some do nothing, and clients value that we show both. It also keeps the program honest about where effort should go next.