Answer
How long does it take to show up in ChatGPT?
The short answer
There are two clocks and they run at different speeds. Anything ChatGPT answers with a live web search can change as soon as your page is crawled and the sources it reads are updated. Anything it answers from memory does not change until the model is retrained. Nobody outside OpenAI can give you a date, and anyone who quotes one is guessing.
The two clocks: retrieval and training
Retrieval is the fast clock and training is the slow one, and which clock you are on is decided by the question, not by you.
When ChatGPT answers with search, it fetches pages at the moment you ask. OpenAI runs OAI-SearchBot for that, documented as the crawler used to surface websites in ChatGPT's search features. Change the page, let it be fetched again, and the raw material of the answer has changed. That is a crawl cycle, and it is short.
When ChatGPT answers from what it already knows, it is drawing on a model trained on a corpus gathered by GPTBot, documented as the crawler that collects content which may be used to train the foundation models. That corpus is fixed at the point the model is built. Nothing you publish today alters a model that already exists.
Most real answers are a blend, which is why the same business can be described correctly and incorrectly in two consecutive replies.
What genuinely moves in weeks, not months
Everything on the retrieval clock, which is everything you can edit and everything you can get someone else to edit. We are deliberately not putting a number of days on it, because the number depends on crawl frequency we do not control and cannot see.
- Your own pages. Correct facts, plain text, one version of the truth.
- Your robots.txt. If it is blocking the retrieval crawlers, nothing else on this list can matter.
- Your listings. Google Business Profile, directories, association pages, review platforms. These are usually recrawled more often than your website.
- Structured data. A JSON-LD block stating your name, services, area and contact details.
What that buys you is that the next time an engine looks, it finds the right thing. It does not buy a place in the answer, because being readable and being chosen are different problems.
What is out of your hands entirely
The training clock, the crawl schedule, and the model's own choices about who to name.
You cannot make a model retrain. You cannot request a recrawl of your site by an AI crawler the way you can nudge Google Search Console. You cannot see when GPTBot or OAI-SearchBot last visited unless you read your own server logs, which is genuinely worth doing and is the only direct evidence available to you. And you cannot make an engine prefer you over a competitor with ten years of press behind them, on any timeline, by editing your own site.
This is the honest boundary of the work. Everything inside it is worth doing and none of it comes with a date.
Why our own study cannot tell you a timeline yet
Because it is one day. Our 300-business scan was run on 30 July 2026, and every question was asked once, with no repetition runs. It is a snapshot, and we say so in the study's own caveats.
A timeline claim needs the same businesses measured repeatedly over months, with a record of what each one changed and when. We do not have that yet. When we do, it will be published with its denominators like everything else, and if it shows that the work moves nothing we will publish that too.
This is why you should be suspicious of a median figure quoted in days. We have seen agencies cite one. Ask them for the sample size, the measurement dates and the raw data. If those three are not available, the number is decoration.
What to promise a boss who wants a date
Promise a measurement cadence and a baseline, not an outcome date. That is a commitment you can keep and it is more useful than a guess.
Say it like this. We will measure the named rate across both engines this week, that is the baseline. We will fix the things we control, and here is the list. We will re-measure on the same questions every month and report the movement, up or down, with the transcripts. We will not forecast a date for the first mention because nobody honestly can.
If a boss or a client cannot accept that framing, the honest answer is that this work may not be for them yet. We wrote the full version of that argument, against our own commercial interest, in is AI visibility worth it.
Where to start
A baseline is the only thing that makes any later claim about timing checkable, including ours. The free scan gives you one across both engines, dated, with the answers stored.
Keep reading
Related questions
Answer
ChatGPT has wrong information about my business
There is no correction form. Here is what does work.
Read the answer →Answer
How do I know if AI visibility work is working?
What to measure, and what movement actually means.
Read the answer →Learn
Is AI visibility worth it?
An honest answer from a firm that sells the work.
Read the answer →Learn
The Receipts Engine
The method, the formula and the limits, in full.
Read the answer →See what AI says about your business.
Start with a free scan. It is the fastest way to find out whether AI recommends you or your competitor, and you keep the report.