When you publish a new page, how long before an AI engine can cite it? The answer depends entirely on how a given engine sources its information — and the two paths have wildly different timelines.
This article is about first indexation: the lag between publishing and the moment new content can surface. That’s distinct from your ongoing update cadence (content freshness strategy) and from how long a whole AEO program takes to move (how long does AEO take).
Two clocks, two very different speeds
AI engines get facts from two mechanisms, and each has its own latency:
- Retrieval (RAG / live search). The engine queries a search index or fetches the live web at answer time. New content can surface as soon as it’s been crawled and indexed — days to a few weeks.
- Training (the model’s baked-in knowledge). Facts learned during model training only change when a new model version ships, which can be many months after the content existed. See what is RAG for brands for how the two interact.
Most modern assistants blend both: live retrieval for current questions, baked-in knowledge for general ones. Your new page reaches the retrieval clock quickly and the training clock slowly.
The retrieval clock: days to weeks
For retrieval-based answers, latency is essentially crawl-and-index time plus the engine’s freshness preference. The sequence:
- Discovery. A crawler finds the URL via your sitemap, internal links, or external mentions.
- Crawl. The AI bot or the underlying search index fetches the page. See the AI crawlers guide for which bots matter and how to admit them.
- Indexing. The content becomes a candidate for retrieval.
- Surfacing. The engine chooses to retrieve and cite it for a relevant query.
Realistically, expect anywhere from a few days to several weeks for a new page to become citable in retrieval-based answers — and longer for it to consistently win over established sources.
What speeds the retrieval clock up
- Submit and keep an accurate XML sitemap.
- Link the new page from already-indexed, frequently-crawled pages.
- Ensure AI bots aren’t blocked in
robots.txt. - Serve clean server-rendered HTML so content isn’t hidden behind JavaScript.
- Earn an external mention or two — they accelerate discovery and trust.
The training clock: months, and out of your hands
Content that only lives in a model’s training data can’t change until a new model version is trained and released. You cannot force this. New facts about your brand will reach baked-in knowledge over model release cycles, not on your publishing schedule — which is why brand entity signals compound slowly. This is part of how LLMs learn about brands.
The practical takeaway: don’t rely on training-time knowledge for anything time-sensitive. Optimize for the retrieval clock, where you have leverage, and let the training clock catch up over model updates it’s not worth waiting on.
Why your page hasn’t shown up yet
If a freshly published page isn’t being cited, common causes in rough order:
- Not yet crawled — too new, or poorly linked/no sitemap entry.
- Crawled but not preferred — indexed, but the engine still favors established sources for that query.
- Blocked — AI bots disallowed, or content requires JS to render.
- Wrong expectation — you’re testing a query the engine answers from training knowledge, not live retrieval.
Diagnose which clock the query runs on before concluding the content failed. Asking the same engine a clearly current question (“latest…”) usually triggers retrieval and is a fairer first test.
Set realistic checkpoints
A reasonable expectation for new content:
| Milestone | Typical window |
|---|---|
| Crawled by AI/search bots | Hours to ~1 week |
| Citable in retrieval answers | Days to a few weeks |
| Consistently preferred over rivals | Weeks to months |
| Reflected in model training knowledge | Multiple months / next model |
Check at these intervals rather than refreshing daily. Judging a page in the first 48 hours almost guarantees a false negative.
Frequently Asked Questions
How long after publishing can an AI engine cite my new page?
For retrieval-based answers, typically a few days to several weeks once the page is crawled and indexed — assuming bots aren’t blocked and the page is well linked. Consistently winning over established sources takes longer, often weeks to months.
Why do some answers ignore content I published months ago?
That query is likely being answered from the model’s baked-in training knowledge rather than live retrieval. Training knowledge only updates when a new model version ships, which can take many months and isn’t something you can accelerate.
How can I speed up first indexation?
Keep an accurate sitemap, link the new page from frequently-crawled existing pages, ensure AI crawlers are allowed in robots.txt, and serve server-rendered HTML. A couple of external mentions also accelerate discovery and build trust faster than internal signals alone.
How soon should I check whether new content is working?
Check at staged intervals — around a week for crawling, a few weeks for retrieval citations — rather than daily. Testing within the first day or two almost always produces a false negative because the content hasn’t been crawled and indexed yet.