Google AI Mode cited 36% of 81 brand-new pages within 24 hours of publication, while ChatGPT search cited 10% of the same pages and needed 30 days to reach 42%, according to Semrush’s December 2025 test. Perplexity has no comparable public benchmark, so the honest planning number for a mid-market company is one to four weeks per engine, with week four as the first fair moment to judge.
Those figures come from a strong domain, and yours may be slower. What transfers is the shape: Google is quick and fickle, ChatGPT is slow and sticky, and the question “when will it appear” has three parts, only one of which you control.
How fast do Google AI Mode and ChatGPT cite a new page?
Semrush published 81 FAQ-style pages in September 2025 on a domain with an Authority Score of 84, then queried Google AI Mode and ChatGPT search daily for 30 days using its Enterprise AIO tool. It is the cleanest public stopwatch I could find, and it is a single domain, which is worth saying before anyone builds a plan on it.
Google AI Mode had cited 29 of the 81 pages (36%) after 24 hours and 45 (56%) after seven days. It then wobbled. At its peak 48 pages (59%) were cited, and by day 30 only 21 (26%) still were. ChatGPT search began at 8 pages (10%) after a day, reached 14 (17%) after a week and 34 (42%) after a month. In Semrush’s words, once ChatGPT cited a page, the page kept the citation.
So the two engines reward different patience. Google gives you an early look and then reshuffles it, a pattern I covered in the piece on citation volatility. ChatGPT gives you a late look that tends to hold.
There is a practical reading for a company that publishes a handful of pages a month. If a page is uncited after one day, that says almost nothing, since even on Google’s fast lane 64% of Semrush’s pages were still uncited at the 24-hour mark. If it is uncited after 30 days on ChatGPT, roughly 58% of Semrush’s pages were in the same position, which means the page is in the majority, not in trouble. The sensible decision point sits later than most weekly reports assume, and the sensible reaction to an early blank is to check the door before rewriting the page.
Why does the wait differ between ChatGPT, Perplexity and Google?
Because the wait is mostly a crawl schedule, and each engine keeps its own. Google states the rule plainly in its Search Central documentation: a page must be indexed and eligible to be shown in Search with a snippet to appear as a supporting link in AI Overviews or AI Mode, and there are no additional requirements. It also says crawling can take anywhere from several days to several months, depending on how often its systems decide a page needs refreshing. Your page is sitting in a queue you cannot see, which is a different problem from being ignored.
Perplexity documents one number: changes to your robots.txt for PerplexityBot may take up to 24 hours to be reflected. Beyond that, the best public evidence is a single server-log case from The GEO Lab, April 2026. A page published on April 18 was crawled on April 30, twelve days later, and cited the same morning. One page is an anecdote, and the authors say plainly they cannot produce a universal number. The useful part is the mechanism: the delay sat between publishing and crawling, not between crawling and citing.
OpenAI’s published guidance for publishers is about access rather than speed. It tells site owners to allow OAI-SearchBot in robots.txt and says any public website can appear in ChatGPT search. It gives no timetable, which is itself information.
Publishing is a date you choose. Being cited is a date someone else’s crawler chooses.
Press and earned coverage run on a slower clock still. Stacker, a newswire that sells distribution and so has an interest in the result, tracked 245 stories across six AI platforms for four weeks after distribution and reported in July 2026 that Google’s products spike early and settle at a slightly lower level, while Claude compounds: 118 of the 132 stories that registered there grew between week one and week four. Its March 2026 study of 87 stories across 30 brands and eight platforms found a median 239% lift in citations over brand-owned content alone, measured over 30 days. Read the vendor caveat, then read the direction. Third-party mentions raise the ceiling, and they take weeks to register. I go further into that in the piece on PR and AI visibility.
What can a CEO control to shorten the wait?
Only the inner orbit: whether the page can be found, whether the bots are allowed in, whether the answer comes first, and whether the page is kept current. The outer orbit is patience. Four moves this quarter, none of which needs a marketer. Notice what is absent from the list: no new file format, no special markup. Google’s documentation says outright that you do not need machine-readable AI text files or special schema to appear in its AI features, so anyone selling you a shortcut through the crawl queue is selling you something the platforms say does not exist.
Open the door
Ask whoever runs your site to confirm robots.txt allows Googlebot, OAI-SearchBot and PerplexityBot, and that your ten most important pages show as indexed in Search Console. Google’s own condition for AI features is indexing plus snippet eligibility.
Write down ten buyer questions
Pick the ten questions your best customers asked before they bought. Record today which engines cite you for each. Without a baseline, every future result is an opinion.
Set the clock at week four
Check weekly if you must, but judge only at four weeks, and again at twelve. Both the Semrush and Stacker data show the picture is still moving at week four on some engines.
Lead every page with the answer
Put the direct answer and its number in the first two sentences, then keep the page dated and current. Be the answer, not the ad.
The part nobody wants to hear
Most companies will read a number like 42% at 30 days, nod, and change nothing. The gap between knowing and doing has little to do with knowledge. The findings above fit on an index card, and the four moves take an afternoon of a technical colleague’s time and a spreadsheet.
What is missing is an owner. Marketing has tools that report on AI visibility and no operating system that decides who checks them, when, and what happens next. Strategy tends to belong to whoever has ten free hours, which in a company of 200 people is no one. So the baseline never gets written, the week-four review never happens, and in month three someone concludes that “AI doesn’t work for us” from a test that was never run. What gets measured gets cited, and what nobody owns does not get measured.
Publish on a Monday, measure on the fourth Monday after, and refuse to draw conclusions from anything in between.