What works for getting content cited by AI search engines: rewriting pages, publishing new content, or earning mentions
Rewrite existing pages, publish new content or earn mentions on other sites? How we decide which lever gets an AI citation, using Promptwatch data on client programs.
Clients usually arrive with a favorite answer to this question. Some are sure they need more content. Others want a PR push. A few think a rewrite of the top ten pages will do it. We try to retire all three opinions and let the data pick, because the right lever changes from prompt to prompt.
We do not lean on a single industry study for this. The research we have seen is mixed in method, and it measures averages, not your site. So we diagnose each prompt.
The diagnostic
For each prompt a client cares about, we ask three things in order.
Can the bots reach our pages? If AI crawlers hit errors, nothing else matters. We check this first through Agent Analytics in Promptwatch, which logs real-time, IP-verified AI crawler visits with error tracking and a crawl-to-citation path. Crawler logs start on Professional at $245 a month.
Who is cited for this prompt, and where do they live? Citation analytics show cited pages and domains. We sort them into three buckets: the client's own pages, competitors' pages, and third-party pages such as Reddit, YouTube, LinkedIn, review sites and news.
Does the client have a page that should have won? Page Tracker lets us watch specific URLs, ours or a competitor's, and compare.
The buckets give us the lever.
When we rewrite
If the client's own page is cited but loses position, or is crawled and never cited, the page is close. We use Content Refresh, which gives diff-style update suggestions, then tighten the page around the exact question: a direct answer first, then the evidence. We keep the URL.
Citation Trends help us set expectations. They report, per page, when it was first cited, the ramp-up days, the peak rate and the decay half-life, so we can tell a client whether a refreshed page is likely to hold its gains or fade.
When we publish
If nothing from the client's site appears and the cited sources are competitors' or neutral brand pages, there is a real gap. Content gap analysis gives us coverage scores against the prompt list. For clients on Business or Enterprise, the Content Agent can plan, write and publish against those gaps to WordPress, Webflow or Framer, with a review inbox so nothing goes live unapproved. We edit every draft. The Action Board also gives a prioritized to-do list, and the weekly Discovery run suggests up to five evidence-backed actions.
One caution from our own practice. Agency plans do not include the Content Agent, so when a client wants scheduled publishing we move them to Business, or we write pages ourselves and use the data to brief them.
When we go earn mentions
If third-party pages dominate, the work is outside the site. The offsite mentions view shows where a brand is named in third-party pages without a link, and the Reddit, YouTube and LinkedIn citation reports show which communities the engines trust for that topic. Then the plan is human work: contributing to the right threads honestly, getting included in the comparison pages that already rank, pitching original data, or partnering. No tool does this for you, and we say so.
Test one lever at a time
We pick a subset of prompts, change one lever, and leave a matched group alone. Because answers vary run to run, we compare steady daily panels across weeks, not single checks. Promptwatch's A/B period comparison makes this easy to report.
What published case studies suggest
Promptwatch's own customer case studies are worth reading, as long as they are attributed correctly. Crisp scaled to 5 to 10 articles a day and reported AI-referred traffic converting at twice the rate of traditional channels. TreeHouse, which runs Pararius and Huurwoningen.nl, reported AI search reaching about 25% of its organic conversions, 99% from ChatGPT, after fixing crawler access and automating weekly reporting. Those are those customers' results. The pattern we take from them is the order of operations: access, then fit, then volume.
Our advice
Do not pick a lever by habit. Measure the gap, choose the lever, and measure again. That sequence is why we run these programs on Promptwatch: the same platform that shows the gap also helps close it and report the result.
No approach guarantees a citation. The engines decide what to cite and change how they do it without notice.