How Often to Update a Page So ChatGPT and AI Assistants Keep Citing It

Is there a proven update interval for AI citations?

No, nobody has published a proven interval for keeping AI citations. Parse, which sells an AI-visibility tool, says research gives no universal number of days after which content stops earning citations, and no universal share of text that must change for AI to notice a refresh (directional). We mark every figure from a company selling AI-visibility tools or services as directional.

A review schedule exists to catch a stale fact before an assistant repeats it. Bing's search team wrote in May 2026: "In search, stale content degrades ranking. In grounding, a stale fact produces a misleading response."

What review schedule fits each type of page?

Match each page's review interval to how fast its facts change and how many citations it earns. The schedule below is our rule of thumb, built from the sources here; calibrate it against your own tracking.

Start with the few pages that carry the load. Parse reports that 80% of a brand's AI citations come from 15% to 20% of its pages (directional). Ahrefs' refresh guide also starts with the top 20% of pages by traffic.

For data pages built to be quoted, see how to build a statistics page that becomes AI's default source. To give these reviews fixed slots next to new pages, see our AEO editorial calendar guide.

What does the recency evidence actually show?

The evidence shows a modest pull toward newer pages. In our reading, freshness works as a tiebreaker between pages that answer a question equally well, and it does not shut older pages out.

Ahrefs' study of AI citations covered 17 million citations across 7 AI search platforms and found cited URLs were on average 25.7% fresher than URLs in organic results. The average cited page was still 2.9 years old.

Seer Interactive, a search marketing agency, found in July 2026 that 75% of the pages LLMs cite were updated in the last year (directional). A sizeable share of those fresh pages had been first published years earlier and stayed fresh by being maintained (directional).

Lab tests show the same pull. An arXiv paper on recency bias in LLM rerankers injected fresh dates into passages and moved the Top-10's mean publication year forward by up to 4.78 years. Between equally relevant passages, a date alone reversed the model's preference by up to 25% on average.

Do ChatGPT, Perplexity, Gemini and Google AI Overviews weigh freshness the same way?

No, each engine treats freshness differently, and Ahrefs and Seer Interactive disagree about where Gemini sits.

• ChatGPT. Ahrefs found ChatGPT cites URLs 393 days newer than Google's organic results in its in-text references and 458 days newer in its citations. Seer puts 73% of the content ChatGPT cites as updated in the last year (directional). Freshness can only help when ChatGPT runs a search, and in our own check below it answered from memory; see when ChatGPT searches the web instead of answering from memory.
• Google AI Overviews. Ahrefs found AI Overviews was the only platform not preferring fresher content, citing pages 16 days older than organic results on average. Google says its AI features rely on core Search ranking systems to retrieve relevant, up-to-date pages, and it runs "query deserves freshness" systems for queries where fresh results are expected.
• Perplexity. Perplexity's in-text citations averaged 1166 days since publication in Ahrefs' data. In Seer's data only 65% of the content Perplexity cites was updated in the last year (directional).
• Gemini. Ahrefs measured Gemini's citations at 1118 days old on average, older than ChatGPT's. Seer's data instead ranks Gemini freshest, with 78% of cited content updated in the last year (directional).

The Gemini split likely reflects different years, samples and measures. Our reading for AI Overviews: they inherit Search's judgment of which questions need fresh pages.

What should trigger an update between scheduled reviews?

Update early when something changes the answer your page gives, or when your tracking shows the page losing ground. The calendar catches slow drift; triggers catch events.

Parse recommends review triggers matched to each claim's lifetime, calling this an editorial workflow and not a measured schedule (directional). These are the triggers we watch:

1. A cited fact changed. The table above lists the usual examples for each page type. IndexNow's guidance names content, pricing and availability as examples of meaningful changes worth submitting.
2. Your tracked citations dropped. A page missing from answers it used to win needs a look, ideally caught by citation monitoring alerts and not months later.
3. A newer competing page appeared. Ahrefs found Perplexity and ChatGPT order in-text references from newest to oldest, so a fresher rival on the same question can push your page down the list.
4. The season is coming. Use the lead time in the table above.
5. New data arrived. See the statistics row of the table above. Gander, which sells an AI-visibility tool, found ChatGPT added a year to 23.4% of the 23,973 fan-out queries it examined, even when the user gave none (directional).

What counts as a real update, and what is only a new date?

A real update changes what the page says. A new date or a cosmetic layout change does not count.

Google's sitemap documentation draws that line: main content, structured data and links are generally significant changes, and a copyright date is not. Google's helpful content self-assessment asks whether you change page dates to seem fresh when the content has not substantially changed.

Ahrefs reports that Google's John Mueller has warned against updating publish dates without corresponding content changes. Its refresh guide adds that Google can detect superficial updates and may ignore them entirely.

The problem is common. The arXiv authors note that most page "updates" are minor edits that leave the substantive text untouched yet reset the "Last updated" timestamp.

For which passages to rewrite and how to make them quotable, see the content refresh guide.

How do you signal an update to search engines and AI crawlers?

Show the change on the page itself, keep your sitemap's lastmod accurate, and ping engines through IndexNow.

On the page

Google recommends a prominent visible date labelled "Published" or "Last updated".

Its publication dates guidance also recommends datePublished and/or dateModified in Article or BlogPosting structured data, matching the visible date.

In the sitemap

Google uses lastmod if the value is consistently and verifiably accurate.

That rule comes from Google's sitemap documentation. Bing's sitemap guidance for AI-powered search, from July 2025, says Bing uses lastmod to prioritize URLs for recrawling and to skip unchanged ones.

• Use full timestamps. Bing asks for ISO 8601 lastmod values with both date and time.
• Never stamp the build time. Bing warns against setting lastmod to the sitemap's generation time unless the page content actually changed.
• Expect daily checks. Bing typically revisits submitted sitemaps at least once per day, its July 2025 post says.

With IndexNow

Bing's July 2025 guidance recommends IndexNow to notify Bing and participating engines instantly when a URL is added, updated or removed. Ahrefs' refresh guide notes IndexNow does not reach Google, and argues it can speed AI visibility because ChatGPT uses Bing's index.

How long until AI assistants see the updated page?

The wait depends on the engine and on whether you notified it. We found no engine below that commits to re-reading an edited page by a given date.

• Google AI Overviews and AI Mode. Google's recrawl documentation says crawling can take a few days to a few weeks. A page must be indexed and snippet-eligible to appear as a supporting link, so in our reading AI features see an edit only after Search re-indexes it.
• Bing. Bing's July 2025 guidance says freshness signals directly influence how quickly updates show up in its results and AI answers.
• ChatGPT. OpenAI's crawler documentation says OAI-SearchBot surfaces sites in ChatGPT's search features, and that a robots.txt change can take about 24 hours to reach search results. We found no stated timing for content edits.
• Claude. Anthropic says Claude-SearchBot crawls the web to improve search result quality; we found no published refresh timing.
• Without a ping. The IndexNow FAQ says crawler-based discovery of updates can take days or even weeks, and that a submission does not guarantee immediate indexing.

When should you leave a page alone?

Leave a page alone when its facts hold, it still earns citations at its usual rate, and looking dated would cost you nothing.

Seer suggests asking exactly those three questions before leaving evergreen content untouched (directional). Parse says an evergreen explainer with no AI citations and no AI Overview overlap does not need a quarterly refresh (directional).

An older page can still be the best answer. Seer calls Perplexity the engine where strong older evergreen content still earns citations (directional), and the arXiv authors warn that LLM rerankers may undervalue authoritative older material.

Frequency is no shortcut either. Ahrefs says: "Low-quality, irrelevant content that's updated every day will not have a magic positive effect on your visibility in AI search."

How do you know whether an update worked?

Track the AI answers themselves, since traffic and rankings alone will not show a change inside them.

Parse recommends recording repeated answers to a fixed question set before the edit, then repeating them on the same engines afterward (directional). Ahrefs suggests checking rankings 1 to 3 weeks after an update, which in our reading fits Google's recrawl window.

For Google's own AI features, Google points site owners to the Generative AI performance report in Search Console.

In our client reports every AI answer is kept word for word with its exact prompt, so anyone can paste the prompt into ChatGPT or Claude and reproduce it. For prompt sets and scoring, see how to measure your AI visibility.

How we tested this

Before writing, we put this article's main question to ChatGPT and Claude, read every page they cited in full, and built sections around the searches they ran. Both ran through the APIs with a web search tool available, each model deciding whether to use it.

• ChatGPT did not search. It answered from memory and cited no sources, with advice close to our sections on real updates and sitemap signals.
• Claude ran a single search. It cited Ahrefs' freshness guide alongside AI-visibility vendor blogs, an agency blog and a newsletter. The top-ranked page was about a month old; the rest dated from the spring.
• Claude set a much tighter cadence than ChatGPT. It was close to the table above, and Claude flagged that most of its figures came from vendor blogs and named Ahrefs' study as the strongest source.
• Neither named OnlyAEO. We had no page on this question yet.

The cadence an assistant recommends depends on which pages it happens to read, so lean on primary documentation. We left out the vendor decay figures Claude repeated that we could not check against a published page. These were single API runs, not the consumer apps.

Where does OnlyAEO fit?

OnlyAEO is an AEO platform that tracks where AI names your brand and creates the content to change it. Its managed plans include monthly AI visibility analytics, fact-checked articles published on the client's site, and a website health check so every AI engine can read the site.

The part that keeps facts current is the Context Graph: a single record of each client's facts and claims, where a change made once, such as a price, applies to all new content. Answer pages go on the client's own domain with guardrails so they do not hurt existing search presence.

When choosing a partner for this work, ask:

• Can they show the answers? Look for reproducible reports like those described above.
• Do they watch triggers? Ask which of the triggers above they track, and how.
• Do they refuse date-only refreshes? See what counts as a real update, above.

More questions about updating pages for AI citations

Do sitemap priority and changefreq values help crawlers find updates?

Not on Google: its sitemap documentation says Google ignores priority and changefreq values.

Will requesting a recrawl again and again speed things up?

No. Google says requesting a recrawl multiple times for the same URL won't get it crawled any faster. IndexNow asks sites to wait at least 5 minutes between submissions for frequently updated pages.

Can engines rely on a server's Last-Modified header to judge freshness?

Rarely reliable, in Parse's data: of 8,215,946 citations from ChatGPT Search and Google AI Mode, only 116,379, about 1.4%, passed its Last-Modified reliability filters (directional).

Do the pages AI cites month after month stay fresh?

Mostly. Seer Interactive found pages cited in all four months of its study, 45% of dated pages, were 68% fresh with a median age around six months (directional).