A new page waits a median of 10 days for its first citation when AI engines already cite its domain, and 33 days when they don't. Launch pages from trusted sources were cited in the next weekly snapshot. Last year's pages held on for weeks after the new ones arrived.
Research
106 weekly snapshots from February to November 2026. Every number below is explained in the sections that follow.
The same questions, asked to AI engines every week, with every cited page recorded.
Every week, Brandlight asks the same set of questions to the AI engines each industry tracks and records every page the answers cite. A page counts as cited in a snapshot when it is among the 1,000 most-cited pages in its industry that week. Its first citation is the first snapshot where that happens.
For publish dates we read each page's own markup, then checked it against the page's citation history and, where it exists, the public web archive. Snapshots are weekly, so the fastest lag this study can show is "cited in the next snapshot".
It depends on where the page lives. Pages on domains the engines already cite were picked up 2.4 to 3.1 times faster in every industry.
Research
"Domain already cited" means the engines cited other pages from the same domain in that industry during the first 4 weeks of the window. Pages published in the last 8 weeks of the window are left out, so every page had the same time to be cited.
Consumer electronics moves fastest. A new page from a tech publisher the engines already cite was cited within a week 42% of the time, and within 2 weeks 73% of the time. Health insurance moves slowest: 19 days at the median even on already-cited domains, and 45 days on new ones.
Across all 3 industries, 64% of new pages on already-cited domains were cited within 2 weeks. On new domains, 26% were. After 8 weeks the gap narrows to 88% against 70%.
Research
All 3 industries pooled.
| Cited within | Domain already cited | New domain |
|---|---|---|
| 1 week | 36% | 12% |
| 2 weeks | 64% | 26% |
| 4 weeks | 75% | 44% |
| 8 weeks | 88% | 70% |
Google AI Overviews was among the three fastest engines in every industry. Claude and Perplexity had the two lowest shares in every industry.
Research
Each engine is judged only on pages where it ran in every weekly snapshot of the page's first 8 weeks. Blank cells: the engine does not run in that industry's panel.
| Engine | Consumer electronics | Health insurance | Personal care |
|---|---|---|---|
| ChatGPT | 48% | 18% | 10% |
| Microsoft Copilot | 17% | 12% | |
| Google AI Overviews | 43% | 14% | 13% |
| Gemini | 45% | 10% | |
| Google AI Mode | 42% | 13% | 11% |
| Perplexity | 33% | 9% | 4% |
| Claude | 34% | 8% | 1% |
Google AI Overviews cited 13% to 43% of new pages within 2 weeks, depending on the industry. Claude and Perplexity had the two lowest shares in each industry. Consumer electronics runs much faster than the other two industries for every engine, so compare engines within a column.
Given 8 weeks, Claude and Perplexity reached 8% to 54% of the same pages, depending on the industry. The full list of engines measured: ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews, Microsoft Copilot and Perplexity.
Around a flagship smartphone launch, every launch page with a reliable publish date was cited in the first weekly snapshot after it went live.
Research
Days counted from the launch event (day 0). Consumer electronics, 22 markets.
| Page | Went live | First cited | Days to first citation |
|---|---|---|---|
| Brand newsroom launch release | Day 0 | Day 4 | 4 |
| Brand newsroom launch release (regional) | Day 1 | Day 4 | 3 |
| Tech news site: camera explainer | Day 1 | Day 4 | 3 |
| Independent camera lab test | Day 12 | Day 18 | 6 |
| Brand newsroom feature | Day 19 | Day 25 | 6 |
| National newspaper review | Day 19 | Day 25 | 6 |
| Tech news site: head-to-head camera test | Day 24 | Day 25 | 1 |
The brand's newsroom release was cited 4 days after the launch event. Coverage published over the following 24 days followed the same pattern, each piece cited in the snapshot after it appeared: 1 to 6 days in every case.
Head-to-head comparison articles went from almost nothing to 4.8% of all citations in the category, then faded.
Research
Consumer electronics, 22 markets. Comparison pages set the competitor's new flagship against the existing one. Each bar is one weekly snapshot.
Before the competitor's launch, these comparison pages drew 0.2% of citations or less. Tech publishers then wrote head-to-head reviews, and each of the largest was cited 8 to 9 days after it went up. They entered at #2, #3, #4 and #6 among every page cited in the category that week. The largest drew 4,379 citations in its first snapshot.
The peak came on day 32. By day 60 the comparison pages were down to 1.1% of citations, 4.4 times below it.
Slowly. The 2027 Medicare Star Ratings came out on 8 October 2026. 31 days later, AI still cited pages about 2026 far more.
The release date is fixed and public, which makes it a clean test. New pages about 2027 started earning citations within 10 days, and their share kept growing. Pages about 2026 kept most of theirs.
The same timing shows up elsewhere in health insurance. KFF published its annual employer health benefits survey on 22 October 2026. It was cited 10 days later and reached 490 citations in a single weekly snapshot.
Most new citations are gone within a week. Pages on the brand's own site stay longest and social posts fade fastest, in all 3 industries.
Research
Pages that entered an industry's most-cited set after at least 4 weeks away. "Brand's own site" is the site of the brand each industry panel tracks.
| Industry | Brand's own site | Other sites | Social and forums |
|---|---|---|---|
| Consumer electronics | 35% | 21% | 8% |
| Health insurance | 28% | 20% | 10% |
| Personal care | 33% | 14% | 9% |
We followed 5,294 pages from the week they entered an industry's most-cited set. 37% were still cited the following week, and 18% were still cited 8 weeks later. Part of the brand's advantage comes from questions about the brand itself, which keep drawing on its own site.
Best-of lists arrive slowly and stay. A new best-of list took a median of 38 days to be cited in health insurance and 45 days in personal care, against 18.5 and 28 days for guides. Once in, 25% to 31% of best-of lists were still cited 8 weeks later, among the highest of any format. Videos fade fast, about as fast as social posts: 6% to 10% were still cited after 8 weeks. In consumer electronics, reviews arrive fastest, at a median of 6 days.
In health insurance and personal care, about 7 in 10 "new" publish dates are refresh dates. Consumer electronics, where launches create genuinely new pages, sits far lower.
Research
Cited pages whose stated publish date falls inside the study window and that the public web archive holds. A date counts as a refresh if the page was cited before it, or appears in the archive more than 7 days before it.
| Industry | Refresh dates |
|---|---|
| Health insurance | 72% |
| Personal care | 69% |
| Consumer electronics | 24% |
Publishers change the publish date when they refresh a page, and brands do the same on their own sites. All 5 of the launch brand's product pages we checked now show a publish date months after AI engines first cited them.
Outside fast-moving categories, the date in the markup mostly records the last refresh. To know when a page first appeared, check its citation history or an archive.
The same pattern in three unrelated industries, and events whose dates nobody in the panel controls.
It repeats across industries. The gap between already-cited and new domains appears in consumer electronics, health insurance and personal care, with medians of 8, 19 and 13 days on known domains against 20, 45 and 40 on new ones.
The event dates are external. A launch event, a competitor's launch and a government ratings release happen on dates set by someone else, so the before and after are fixed in advance.
Volume swings are taken out. Event figures are each page group's share of all citations in that week's snapshot. Weeks when engines cite fewer pages overall, such as the drop we reported in The Source Collapse, move every page together and leave shares comparable.
Dates are checked, not trusted. Any page cited before its stated publish date is treated as refreshed and left out of the timing figures.
Five conclusions from the evidence above.
The same kind of page was cited 3.3 times faster at the median on an already-cited domain. Before launching a new microsite, check whether the content can live on a domain your category's answers already draw on.
Even on already-cited domains, the median wait ran from 8 days in the fastest industry to 19 days in the slowest. Content for open enrollment, a launch or a seasonal peak has to go live weeks before it.
31 days after a new edition was out, AI engines still cited the previous year's pages 3.4 times as much. Update or redirect the old page when the new one goes live.
After a competitor's launch, head-to-head articles reached #2 among all cited pages within days of going up. The publishers who write them are the ones the engines already cite, and they write them in the first weeks after a launch.
Only 18% of newly cited pages were still cited 8 weeks later. Pages on the brand's own site and best-of lists were the most likely to still be there. Count how long a citation lasts as well as when it arrives.