Content refresh or new article? A decision table for winning AI citations
When a page is losing citations, rewriting it often beats publishing something new. A decision table, a refresh checklist, and the data that tells you which route to take.
Quick verdict
If you already have a page on the topic, refresh it before you write another. A new article on the same question competes with your own page, splits whatever authority you have, and leaves the weak page online. If you have no page, or the existing one answers a different question, write new. The data to decide comes from citations and crawler logs, and Promptwatch provides both, plus Content Refresh, which proposes diff-style updates for existing pages.
The decision table
| Situation | Evidence | Move |
|---|---|---|
| No page on the question | Content gap analysis shows low coverage | New article |
| Page exists, cited, citation rate decaying | Citation Trends shows decay | Refresh |
| Page exists, never cited, bots fetch it | Crawler logs show fetches, no citations | Rewrite structure |
| Page exists, never fetched | Logs show no visits or errors | Fix access first, then judge |
| Page covers the topic but for a different intent | Prompts differ from the page's question | New article, link the two |
| Page is thin but ranked well in classic search | Search Console strong, AI weak | Refresh with direct answers |
Reading the signals
Citation Trends reports, per page, first cited, ramp-up days, peak rate, and decay half-life. A short half-life tells you citations fade fast, which suggests the topic rewards freshness. A page that peaked and fell after a competitor published is a refresh candidate with a clear target: the page that replaced you, which Page Tracker can follow.
Crawler logs separate technical from editorial problems. Agent Analytics shows fetches by GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, and others, with error tracking and a citation rate per page. If a bot hits a 403 every time, no rewrite will help.
Topic Insights, part of 360 Insights, show where pages underdeliver for their topic. That is a ready list of refresh candidates.
A refresh checklist
A refresh is more than changing the date. Work through these:
- Open with the answer. Put the direct response to the question in the first two sentences.
- Update every number and name. Remove figures you cannot source. Add the current ones you can.
- Match headings to prompts. Rename sections so they read like the questions people ask.
- Add what the cited competitors have. If they include a table or a step list, so should you.
- Remove padding. Cut paragraphs that restate the heading.
- Check internal links. Point related pages at this one.
- Confirm the page is accessible. Test with logs after publishing.
Promptwatch's Content Refresh proposes the edits in a diff-style view, so reviewers can see what changes before accepting.
When a new article is the better call
Write new when the intent is different, when the page would exceed a reasonable length, or when you can own a distinct angle. If you do, link the new and old pages to each other and make clear which answers which question.
For gaps you intend to fill in volume, Promptwatch's Content Agent on Business and Enterprise plans, writes, and publishes against visibility gaps on a schedule with a review inbox or a hands-off mode. The Knowledge Base, available from Professional, keeps facts consistent across articles. Agency plans do not include these.
Publishing and re-measuring
Updates publish to WordPress through the official plugin, to Webflow including localized collections, and to Framer. Framer has no per-item publish: going live redeploys the whole site, so batch your Framer changes.
After publishing, wait three to four weeks, then compare:
| Measure | Before | After |
|---|---|---|
| Fetches per week | Log it | Log it |
| Citation rate | Log it | Log it |
| Position among sources | Log it | Log it |
| AI-referred visits | Log it | Log it |
The visitor analytics script, GTM template, or WordPress plugin records AI-referred visits and conversions. Without a baseline, you cannot tell whether the refresh helped.
What the evidence from customers shows
Promptwatch's published case studies are the customers' own. Crisp scaled to 5 to 10 articles a day and reported AI-referred traffic converting at twice the rate of traditional channels. That shows what is possible at volume. It does not mean refreshes produce a given lift, and we will not claim a number for them.
Plan notes
Essential at $95 a month includes content gap analysis and 360 Insights. Professional at $245 adds 5 AEO articles, the Knowledge Base (100 items), and crawler logs. Business at $579 adds 10 AEO articles, a larger knowledge base, and the Content Agent. Add-ons exist for extra AEO articles on Professional and Business.
Other tools in the ranking are worth a look for editing support. Surfer and Clearscope, for example, are content optimization tools, and the directory entries describe what each covers.
Who should pick which
| You are... | Recommendation |
|---|---|
| Sitting on an old library | Refresh the top ten decaying pages first |
| Entering a new topic | New articles, built from gap analysis |
| A lean team | Refresh monthly, new quarterly |
| A content-heavy brand | Business plan for the Content Agent |
| An agency | Agency plans for measurement, your own process for writing |
Refresh first, write second, and measure both. Promptwatch is the one place where the decay signal, the access check, and the edit live together.