Half of new pages took an estimated 42 days to get a first AI citation
How long until AI engines cite a new page: 493 pages, 8 sites, 4 engines (2026). 493 new article pages on 8 sites (LoudFace and 7 client sites), first published 17 April to 2 November 2026, AI answers read daily on up to 4 engines until 2 November 2026 across ChatGPT, Google AI Overviews, Perplexity and Gemini.
Data windowPages published 17 April to 2 November 2026; AI answers read daily to 2 November 2026
42 days
estimated wait until half of new pages had a first AI citation, counting pages not yet cited
| Value | |
|---|---|
| 7 days | 25.5% |
| 14 days | 32.2% |
| 30 days | 46.3% |
| 60 days | 56.7% |
Source: LoudFace analysis of Peec AI daily answers and CMS publish dates, 493 new pages on 8 sites, to 2 November 20263 Nov 2026
Key numbers
median wait among pages cited within 60 days, a cited-only figure
115 pages; a quarter waited 3 days or lessof new pages were cited by an AI engine within 7 days of publishing
462 pages on 8 sitesof pages watched for 60 days had no AI citation in that time
88 of 203 pages; 67 (33%) still had none on 2 November 2026of cited pages were first cited on the day any engine first read them
246 cited pages; it was 46.4% of the 323 pages readKey takeaways
- Counting the pages not yet cited, half of 493 new pages on 8 sites had their first AI citation within an estimated 42 days, three times the 14-day median among pages cited within 60 days.
- 25.5% of new pages were cited by an AI engine within 7 days of publishing and 46.3% within 30 days, on the questions their sites track.
- Of the 203 pages watched for 60 days, 43.3% had no AI citation in that time and 33% had none by the end of the data.
- Of the 144 pages watched for 60 days that had no citation by day 14, 38.9% were cited by day 60.
- On the 8 sites where both ran, Google AI Overviews cited 34.4% of new pages within 30 days and ChatGPT 26.4%, and the gap grew from 2.6 points at 7 days.
Ask how long a new page takes to be cited by AI engines and the published medians say one to two weeks. Profound put it at 6.81 days for ChatGPT and Claude, as MaxAEO reports it. MaxAEO's own tracking of 740 new pages on brands' sites put it at 13 days. Neither write-up says how pages that were never cited were handled. A median taken only over cited pages leaves those pages out, and the wait looks shorter. Semrush took a different route: it tested 81 pages on its own blog for 30 days on two engines and reported how many of them engines cited at several points in that month, with no median wait.
We followed 493 new pages on 8 sites from the day each went live, read every AI answer to the questions those sites track each day, and estimated a median wait that keeps the pages not yet cited in the count.
Counting the pages not yet cited tripled the wait here
Among the 115 pages cited within 60 days, the median wait was 14 days, close to MaxAEO's 13. A quarter waited 3 days or less and a quarter 28 days or more. Those 115 pages are 56.7% of the 203 pages we watched for 60 days.
A Kaplan-Meier estimate keeps every page in the count and treats a page not yet cited as still waiting on the last day of data. On it, the median wait for all 493 pages was 42 days, three times as long. On the same 203 pages alone it was 40 days. The ratio depends on where the cited-only median is cut: among the 110 pages watched for 90 days, the median among pages cited within 90 days was 18 days. By day 90 the Kaplan-Meier estimate reaches 64.8% of pages cited.
The 42 days pools very different sites. LoudFace's own site, 138 of the pages, had a Kaplan-Meier median of 19 days; the 355 pages on the 7 client sites, 50 days.
Counting only pages watched for the whole period, 25.5% were cited within 7 days, 32.2% within 14 days, 46.3% within 30 days and 56.7% within 60 days. Of the 203 pages watched for 60 days, 43.3% had no citation in that time. 21 of them were cited later, and 67 pages, 33% of the 203, had still not been cited when the data ended on 2 November 2026.
Which engines cited new pages
ChatGPT and Google AI Overviews ran on all 8 sites. Within 30 days, Google AI Overviews had cited 34.4% of new pages and ChatGPT 26.4%. The gap grew with time: 2.6 points at 7 days, 3.8 at 14, 8 at 30 and 11.8 at 60, by which point 7 of the sites had pages old enough to count.
Perplexity and Gemini ran on fewer sites, so their shares cover fewer pages and do not compare directly with the other two. Pages from 4 sites count for Perplexity, which cited 37.6% of 93 pages within 60 days. Gemini cited 47.3% of 110 pages from 3 sites in the same period.
| Value | Note | |
|---|---|---|
| ChatGPT | 31.5% | 203 pages, 7 sites |
| Google AI Overviews | 43.3% | 203 pages, 7 sites |
| Perplexity | 37.6% | 93 pages, 4 sites |
| Gemini | 47.3% | 110 pages, 3 sites |
Source: LoudFace analysis of Peec AI daily answers, pages watched for 60 days on the sites where each engine ran3 Nov 2026
ChatGPT and Google AI Overviews ran on all 8 sites, and pages old enough for 60 days came from 7. Perplexity and Gemini ran on fewer sites, so their bars do not compare directly with the other two.
Most cited pages were cited the day they were first read
An engine retrieves a page when it reads it while answering, and cites it when the answer names it as a source. Of the 246 pages ever cited, 61% had their first citation on the same day any engine first retrieved them. The median gap was 0 days, and three quarters were cited within 3 days of the first read. For the pages that got cited, the wait was mostly a wait to be read.
A read does not guarantee a citation. Of the 323 pages an engine read at least once, 150 (46.4%) were cited on the day of their first read and 77 (23.8%) were never cited. Of the 204 pages read within 30 days, 23.5% had no citation by day 30.
| Step | Count | Share of 337 |
|---|---|---|
| Watched for 30 days | 337 | 100.0% |
| Read by an AI engine | 204 | 60.5% |
| Cited by an AI engine | 156 | 46.3% |
Source: LoudFace analysis of Peec AI daily answers, 337 pages watched for 30 days3 Nov 2026
How it varies by site
The share of new pages cited within 30 days ran from 14.7% on a literacy-education site to 65.7% on an employment-law site. LoudFace's own site was at 52.2%. The sites differ in topic, tracked questions, engines and days of data, so read the spread as the range across these sites.
The literacy-education site had the slowest start: 0 of its 34 pages were cited within 7 days, yet 43.8% of the 32 watched for 60 days were cited by then.
What this means if you publish to be cited
In this sample, a page with no citation at 14 days was often cited later: of the 144 pages watched for 60 days that had no citation by day 14, 38.9% were cited by day 60. Judging a new page at 60 days gives it the time most first citations here took.
Plan for the pages that are not cited, too. On these sites about four pages in ten had no citation within 60 days, and a third had none by the end of the data.
Figure 4
New pages cited within 30 days, by site
In alphabetical order. Sites differ in topic, tracked questions and engines. Peec has data on 62.7% of days for the tutoring site and 60.2% for the supply-chain site, whose bar rests on 10 pages.
| Value | Note | |
|---|---|---|
| AI agents software | 29.4% | 17 pages |
| B2B marketing agency (LoudFace) | 52.2% | 90 pages |
| Employment law | 65.7% | 70 pages |
| Literacy education | 14.7% | 34 pages |
| Supply-chain software | 20.0% | 10 pages |
| Thyroid health | 57.6% | 33 pages |
| Trading education | 31.6% | 38 pages |
| Tutoring | 44.4% | 45 pages |
Source: LoudFace analysis of Peec AI daily answers and CMS publish dates; client sites named by industry3 Nov 2026
Method
How we measured it, and where it stops.
Read this before you quote the numbers. The sheet is the paper’s measurement record: what was sampled, over which window, on which engines, and how.
- Sample
- 493 new article pages on 8 sites (LoudFace and 7 client sites), first published 17 April to 2 November 2026, AI answers read daily on up to 4 engines until 2 November 2026
- Window
- Pages published 17 April to 2 November 2026; AI answers read daily to 2 November 2026
- Engines
- ChatGPT, Google AI Overviews, Perplexity, Gemini
- Published
- 3 November 2026
Question. After a page is published, how many days pass before an AI engine first cites it in an answer, and what share of new pages are cited within 7, 14, 30 and 60 days? We wrote the method down and recorded its hash before collecting any data. Three amendments made afterwards are listed below and in the published method file.
Sites and pages. The 8 sites LoudFace writes for whose AI answers are tracked daily in Peec AI and whose pages live in a Sanity CMS: LoudFace's own site and 7 client sites (employment law, literacy education, tutoring, thyroid health, AI agents software, trading education, supply-chain software). We took every published article document of each site's article type: 922 documents. The publish date is the CMS publish field, as a UTC day.
New pages only. A page counts when it was published on or after the first day its site has Peec data, and when its publish date is no more than 3 days before the document was created in the CMS. An earlier publish date marks an imported or back-dated post, which was not new that day; this removed 405 documents. 18 duplicate copies of pages left by one site's import were merged into their page, dated by the earliest copy. 5 pages published after the window ended were left out, and 1 page whose first citation came before its publish day was left out because it did not go live that day (amendment 2, prompted by seeing that page in the results). That leaves 493 new pages.
AI answers. Peec AI records, for each tracked question, engine and day, which pages the answer retrieved and cited. We read every record for each site's own domain: 36,241 rows of page, day and engine. ChatGPT and Google AI Overviews ran on all 8 sites. Perplexity ran on 5, but on one of them only before any page in the sample was published, so its shares cover 4. Gemini ran on 4. The date of a first citation is the first day Peec recorded one.
Matching. A record belongs to a page when its address is on the site's domain, under the site's article path, and ends in the page's slug, ignoring www, case, query strings and a trailing slash. Matching on the last path segment was amendment 1: one site puts a category folder before the slug, which the first rule missed. 3 recorded addresses were cut off mid-word and match no page.
What counts. Retrieved: the engine read the page while answering (Peec's retrieval count above zero). Cited: the answer cited the page (citation count above zero). Days to first citation is the first day with a citation on any engine minus the publish day.
Analysis. The share cited within h days uses only pages whose publish day plus h is on or before 2 November 2026, so every page in a share was watched for all h days; the 60-day share therefore covers older pages than the 7-day share, from 7 of the sites. A per-engine share also counts a page only when that engine has data for the page's site on at least one day between the publish day and the publish day plus h (amendment 3); the CSV gives each engine's first day of data after each page's publish day. The 42-day median is a Kaplan-Meier estimate over all 493 pages, with a page not yet cited counted up to the last day of data. The 14-day median is the plain median among the 115 pages cited within 60 days, the figure a median gives when pages not yet cited are left out. On the 203 pages watched for 60 days alone, the Kaplan-Meier median is 40 days; with a 90-day cut, the cited-only median is 18 days. These two figures and the medians for LoudFace's own site and the client sites were added after the first draft, at the reviewers' request, to show how the result depends on those choices; the frozen method did not name them. Site figures are shown only for sites with at least 10 pages watched for 30 days, under their industry. The analysis script re-derives every number from the raw rows and gives the same output on every run.
Limitations
Only tracked questions. Each site tracks a set of questions in Peec AI, and the sets may have changed during the window as questions were added and retired; we did not read the question history. A page cited in answers to questions nobody tracks is invisible here, and a question added after a page was published delays the first day that page could be seen cited. These shares are a floor for citation anywhere, and a page counted as not cited may have been cited elsewhere.
Recorded days only. The first citation is the first day Peec recorded one. An engine can answer the same question differently through the day, so a page may have been cited earlier in answers Peec did not record.
Our sites, our pages. All 8 sites are LoudFace or its clients, and the pages were written to be found in search and AI answers, mostly for questions the sites already track. They are not a sample of the web, and results for sites that publish without that aim may differ.
Missing days. Peec has data on every calendar day for 5 of the 8 sites, on 91.4% of days for the AI agents software site, and on only 62.7% for the tutoring site and 60.2% for the supply-chain site. A missing day can only delay the first sighting, so on those sites the waits may read longer than they were.
Engines differ by site. Perplexity's shares cover pages from 4 sites and Gemini's from 3 or 4 depending on the period, so they rest on fewer pages and do not compare directly with ChatGPT and Google AI Overviews.
Publish dates. The publish day is the CMS publish field. A page that went live on a different day, or whose address changed after publishing, is measured from the wrong day or loses its earlier citations. We removed back-dated posts and the 1 page cited before its publish day, but cannot rule out every case.
Different pages at each horizon. Each share covers only pages watched for the whole horizon, so the 60-day share is about older pages than the 7-day share; that is why two sites' 60-day shares sit below their 30-day shares. The Kaplan-Meier estimate uses all pages but assumes a page's chance of a first citation does not depend on when it was published.
Earlier studies. We read the Semrush and MaxAEO studies directly; Profound's figure is as MaxAEO reports it, without its method. Neither MaxAEO nor that report says whether pages never cited were in the median, so we compare their medians with our cited-only 14 days and do not claim theirs left those pages out. MaxAEO also notes that Profound's clock may start from an existing address instead of a publish day.
No causes. The study measures when citations arrived. It does not compare topics, formats or page quality, and nothing in it shows that any change to a page would make it cited sooner.
Versions
- Version 1.0 · 3 Nov 2026CurrentFirst edition: 493 new pages, AI answers to 2 November 2026.
- Version 1.1When the paper is re-run, the new version is listed here and this page links to it.
Data and citation
- Per-page data (CSV, one row per new page)Client sites are named by industry only; page addresses are removed.
Bukva, A. (2026, November 3). How long until AI engines cite a new page: 493 pages, 8 sites, 4 engines (2026). LoudFace Research. https://www.loudface.co/research/how-long-until-ai-engines-cite-a-new-page
Questions
What readers ask about this paper
- How long does it take for a new page to be cited by AI engines?
- In our sample of 493 new pages on 8 sites, counting pages not yet cited, half had a first AI citation within an estimated 42 days. Among pages cited within 60 days the median wait was 14 days, and 25.5% of all new pages were cited within 7 days.
- Does every new page get cited by AI engines?
- No. Of the 203 pages we watched for 60 days, 43.3% had no citation in that time on the questions their sites track, and 67 pages (33%) had none by the end of the data.
- Did some AI engines cite more new pages than others?
- On the 8 sites where both ran, Google AI Overviews cited 34.4% of new pages within 30 days and ChatGPT 26.4%. Within 60 days Gemini cited 47.3% of 110 pages from 3 sites and Perplexity 37.6% of 93 pages from 4 sites; they ran on fewer sites, so those figures do not compare directly.
