Peak Citation Rate for New Pages: What to Expect in the First 30, 60 and 90 Days (2026)

No one has published a clean curve for brand-new pages in AI search. Here is what the Promptwatch and Profound data actually show about rise, peak and decay, and how to plan your first 90 days around it.

Key takeaways

  • Nobody has published a study that follows brand-new pages from publish date to first citation to peak. What exists are two adjacent datasets: Promptwatch Data on how long cited pages stay cited in ChatGPT, and Profound's study of citation decay across seven engines.
  • Profound found that pages that get one-off bursts of citations take roughly 20 days to peak, while sustained pages take 36 to 46 days. Plan for a peak somewhere in days 20 to 45, not day 3.
  • Once a page is cited, content type decides how long it lasts. In Promptwatch's ChatGPT data, landing pages kept 78% of day-0 citations at 30 days; news kept 41% at 60 days.
  • Being cited by several engines is the best early signal of longevity. Profound's pages cited by five or more engines were 5.5x as likely as single-engine pages to be above half-peak at eight weeks.
  • Treat days 1-30 as an access and discovery phase, days 31-60 as the peak window, and days 61-90 as the point where you decide what to refresh, replace or leave alone.

The honest starting point: there is no clean new-page curve

If you search for "how long until a new page gets cited by ChatGPT," you will find confident answers. Most of them are guesses. I went looking for a dataset that tracks fresh URLs from the day they go live, and I could not find one that measures time to first citation or time to peak for brand-new pages.

What does exist is useful, but you have to be careful about what each study measures:

  • Promptwatch Data tracked every page ChatGPT cited on July 6, 2026 for 90 days. That measures retention after a citation, not time to the first one.
  • Profound's study, "The half-life of an AI citation is 11 days" (30 Sep 2026), measured 883,000 pages across seven engines over twelve months. It measures how citation share rises and then decays around each page's peak.

These two use different methods, so do not merge them into a single curve. I will keep them separate below and then translate both into a practical 30/60/90 plan.

Profound's research post on AI citation half-life, covering 883,000 pages across seven AI engines

What the Profound data says about rise and peak

Profound gives the most direct evidence on how long a page takes to reach its peak. Each page is measured separately on each engine, which produced 1.19 million page-and-engine lifecycles. Some of the headline numbers:

  • The median page loses half its citation share within 11 days of peaking.
  • The median interval from a page's first citation to the point where it has lost half its peak share is 42 days.
  • 78% of page-and-engine pairs have a half-life of 14 days or less. Most of these are one-off bursts.
  • Burst pages have a median rise of about 20 days, so their whole arc from first citation to half-peak is a little over a month.
  • The 22% of sustained pages have a median half-life of 27 to 29 days and a median rise to peak of 36 to 46 days.

One caveat from the study itself: citation shares are averaged over a rolling two-week window, so 14 days is the longest half-life a page cited just once can have. The burst category is partly a product of that measurement.

The practical reading: if a page is going to peak, it usually does so between roughly day 20 and day 46 after its first citation. Note that this is counted from the first citation, not from publication. If your page took three weeks to get indexed and discovered, add that on top.

Pages cited by more engines peak later and last longer

The most useful table in the Profound post splits pages by how many engines cite them.

Engines citing the pagePagesMedian riseAbove half-peak at day 56
1679,07719 days1.2%
2275,53031 days2.0%
3-4190,43933 days3.5%
5 or more43,77046 days6.8%

There is a pattern worth noticing. Pages with a longer rise tend to be the ones that last. A fast spike is usually a one-engine event. A slow build across several engines is the signature of a page that is getting picked up broadly. Even so, the numbers are sobering: at best, under 7% of pages in any group were still above half-peak at day 56. Most citations fade.

Engines behave differently

Profound found that Perplexity and Claude return to a page repeatedly. Among sustained pages, 62% of Perplexity's and 63% of Claude's had citations spread across weeks before peaking, against 18 to 26% on the other five engines. ChatGPT, AI Mode and Gemini tend to cite a page intensively for a few days and then stop. Among pages that last beyond two weeks, a Claude citation was 2.8 times as likely as a ChatGPT citation to still be relevant eight weeks later.

Cross-engine correlation of decay was only 0.03 to 0.09. In plain terms, a page fading on ChatGPT tells you almost nothing about whether it is fading on Perplexity. Track engines separately.

What the Promptwatch data says about staying cited

Promptwatch's half-life report starts from a different place: every classified page ChatGPT cited on July 6, 2026, followed through October 4 (90 days). It answers a question that matters once you have a first citation: how long does it last, by content type?

Content typeStill cited at 30 daysStill cited at 60 days
Landing pages78%69%
Docs76%67%
Comparisons72%63%
Listiclesn/a60%
Product pages, reviews, how-tosn/a55% each
Newsn/a41%

A few details from the report are worth repeating. Product pages, reviews and how-tos reach their half-life after 74 to 77 days. News is the outlier: almost a quarter of news pages were never cited again after day 0, and half were gone after 36 days. Docs did better on repeat citation, with ChatGPT citing 93% of them again later against 77% of news pages.

Two cautions on reading this. First, the curves drop faster in the last week simply because the measurement window ends. Landing pages, docs, comparisons and listicles only cross 50% in that final week, so their true half-life is likely close to or beyond 90 days. Second, this is ChatGPT only, and it describes pages that were already earning citations. It is not a forecast for a page you published yesterday.

Promptwatch's own recommendations from the report: treat landing pages, docs and pricing pages as long-term citation assets and keep prices and features current; do not assume product pages stay cited on their own; invest in comparison pages and listicles, which keep more than 60% of citations after two months; and use news and press for short bursts.

Profound's content families confirm the same ordering

Among 133,000 sustained pages, Profound grouped content into four families.

Content familyExamplesMedian half-lifeAbove half-peak at day 56
ChooseComparisons, listicles, reviews29 days11.2%
ExplainHow-tos, docs, reference28 days10.2%
SellProduct, pricing, company pages28 days9.8%
ReportNews, announcements, research27 days7.0%

Report pages ranked last on all seven engines and peaked faster. Pages tied to a date or event fade first. Everything else holds up at roughly 1.4 to 1.6 times their rate. The two datasets, despite different methods, agree on the direction: evergreen comparison and reference content outlasts anything tied to a news cycle.

What to expect in the first 30, 60 and 90 days

Below is my synthesis, not a measured curve. I am combining the Profound rise times with the Promptwatch retention data, and labeling the inference as inference.

Days 1 to 30: access, discovery and maybe a first spike

The first month is mostly about whether engines can find and read the page at all. Many pages will show nothing for the first couple of weeks. That is normal, and it is not evidence the page is bad.

If a page is a burst type, Profound's numbers suggest its rise (around 20 days from first citation) could finish inside this window. A fast spike in week two or three is plausible, especially for freshness-sensitive topics. Do not read it as a long-term win. In Profound's data, 78% of citation pairs decay within two weeks of peaking.

What to do in this window:

  • Confirm AI crawlers are actually reaching the page. A page that bots never read cannot be cited. Check your robots.txt and crawl logs.
  • Make sure the page is in your sitemap and has internal links from pages that already get crawled.
  • Track the prompts the page should answer, so you can see the first citation when it lands rather than finding out weeks later.

Days 31 to 60: the peak window for sustained pages

Sustained pages in Profound's data had a median rise to peak of 36 to 46 days. Pages cited by five or more engines took 46 days. So if a page is going to build slowly and hold, this is when you would expect to see the top of the curve.

This is also where ChatGPT retention matters. In Promptwatch's data, comparisons were at 72% and landing pages at 78% at the 30-day mark. A page that earned a citation in the first month has a good chance of still being cited into the second, if it is an evergreen type. News-style pages have lost a large share by now.

What to do in this window:

  • Compare citation counts by engine. Profound's finding that engines do not move together means a flat ChatGPT line could hide a rising Perplexity or Claude line.
  • Look at which pages are cited by two or more engines. Those are your best candidates for lasting value.
  • Update thin sections on pages that have crawl activity but no citations. That pattern points to a content problem rather than an access problem.

Days 61 to 90: decide what to refresh, replace or leave alone

By day 56, only 1.2% of single-engine pages in Profound's data were still above half their peak. For pages cited by five or more engines it was 6.8%. In Promptwatch's ChatGPT data, product pages, reviews and how-tos reach their half-life around 74 to 77 days, right in this window.

So the third month is where decay shows up for most content types. It is the time to refresh pages whose citations are sliding, not to panic about it. A refresh is cheaper than a new page, and the page already has crawl history and some citation trust.

What to do in this window:

  • Refresh product pages, reviews and how-tos before they hit their half-life, rather than after.
  • Keep prices, features and dates current on landing, docs and pricing pages. These are your long-term assets.
  • Publish the next batch of comparison and listicle content, because that type retains over 60% of citations after two months in Promptwatch's data.
  • Retire the idea that a page is "done." In AI search, authority expires rather than accumulates, as Profound put it.

How to measure this yourself

Published studies will only get you so far. Your own pages, in your category, on the engines your buyers use, are the better evidence. You need three things: prompt tracking that shows when a page is first cited, crawler visibility to separate access problems from content problems, and per-page citation history.

Promptwatch covers all three in one place: it tracks which of your pages get cited and where they land in the source list, shows IP-verified AI crawler activity per page, and has a Page Tracker that watches any URL and lists the prompts and responses that cite it. Its AI crawler logs help you tell "bots never read this" from "bots read it and ignored it."

Favicon of Promptwatch

Promptwatch

AI search visibility and optimization platform
View more
Screenshot of Promptwatch website

If you want to compare options, the GEO software directory at bestgeosoftware.com lists other trackers. Profound, whose research is cited above, is one of them.

Favicon of Profound

Profound

Enterprise AI visibility solution
View more
Screenshot of Profound website

A simple tracking sheet

Whatever tool you use, log each new page at the same checkpoints. Something like this works:

CheckpointWhat to recordWhat it tells you
Day 7First AI crawler visit, indexed or notAccess problem or not
Day 14First citation, which engineDiscovery speed
Day 30Citations per engine, position in source listEarly trajectory
Day 45Same, plus number of engines citingLikely peak window
Day 60Same, plus change from peakStart of decay
Day 90Same, plus decision: refresh, replace or keepLifecycle outcome

Use your own peak date as the anchor. Profound measures half-life from peak, so a page that peaked on day 40 should be judged on its decay starting then, not on the calendar.

Common mistakes when reading early citation data

  • Treating a spike as a win. Most citations are single events. Wait before declaring a page successful.
  • Mixing engines. With cross-engine correlation at 0.03 to 0.09, one blended line hides real movement.
  • Comparing different content types. A news article and a pricing page follow different curves.
  • Ignoring the measurement window. Promptwatch's own report notes that curves drop faster at the end because the window ends. The same logic applies to your own 90-day checks.
  • Assuming ChatGPT numbers apply everywhere. Claude and Perplexity behave differently from ChatGPT, AI Mode and Gemini.

What I would do with this

Set expectations honestly: a new page might show a first citation in the first month, probably peaks between day 20 and day 46 after that first citation, and shows visible decay by the second or third month. That is a planning range from adjacent studies, not a guarantee.

Then build your content calendar around it. Put your evergreen comparison, landing and docs pages first because they last longest. Use news and announcements for short bursts. Schedule refreshes around day 60 rather than waiting for a drop. And if you can, track your own pages from day one. Until someone publishes a proper new-page study, your own data is the best curve you will get.

For more context, see Promptwatch Data's report on content half-life in ChatGPT and its report on ChatGPT citation share by content type, where product pages made up 22.4% of citations (Sep 28 to Oct 4, 2026).

Share:

© 2026 Toolsolved · Find the best marketig tools · RSS

Toolsolved is an affiliate review site. When you click links to vendors or buy through links on our site, we may earn an affiliate commission at no extra cost to you.

Toolsolved is a review website based on user reviews on Reddit and G2, and on publicly available information. We keep everything as up to date as possible, but pricing and features can change. Always confirm the details with the vendor before purchasing.