A new domain waits months for Google. It does not wait for ChatGPT.

Everything you have read about search assumes you will be patient for a year. That advice was written for a different situation to the one you are in, and following it costs you the one channel where being new is not held against you.

Run the free check

Google is cautious with a domain that has no history, which is why a new site waits months for meaningful search traffic. An AI assistant answering a question retrieves and summarises passages that address it, and does not weigh domain age the same way, so a new product can be cited within weeks. On kurli.me, a domain registered on 27 July 2026, ChatGPT sent 56 visitors in the first twenty-five days while Google Search sent 23.

Why Google is slow, and why that is reasonable

When your domain is three weeks old, Google knows almost nothing about it. There is no history of people finding it useful, no established pattern of other sites referring to it, no record of what it is about. From Google's position, an unknown domain making confident claims is exactly what a low-quality site looks like, and the safest response is to wait and see.

This is not a punishment and there is no trick that removes it. It is a reasonable policy applied to a situation where you happen to be indistinguishable from the thing it is designed to catch. The only cure is time plus evidence, and time is the part you cannot compress.

So the practical question is not how to make Google fast. It is what you do during the months while Google is deciding.

Why an assistant behaves differently

When somebody asks an AI assistant a question, the system finds passages that address the question and composes an answer from them, citing what it used. The selection is driven by whether a passage answers the question well.

Domain age is not the organising principle of that process in the way it is for a young site in classical search. A clear, specific, self-contained paragraph that directly answers the question is a good candidate whether it was published two years ago or last Tuesday.

This is the asymmetry, and it is the single most useful thing a new builder can understand. The disadvantage that dominates your first year in Google is substantially reduced in the channel almost nobody is working on.

The evidence we have

We watch this on a domain we know well, because guessing about it would be worthless.

Kurli.me was registered on 27 July 2026, which is a matter of public record in WHOIS. In the first twenty-five days of that domain's existence, ChatGPT sent it 56 visitors and Google Search sent 23.

Both numbers are small, and saying so is the honest part. Twenty-five days into a domain's life nothing is large. What matters is the ratio and the direction: at the moment when Google has the least reason to trust a site, the AI assistant was already sending more than twice as many people.

One case is one case, and we would not build a strategy on it alone. What makes it worth acting on is that the mechanism explains the result, rather than the result being unexplained.

What this changes about the order of work

Conventional advice for a new site is roughly: fix the technical basics, publish consistently, build links, wait. That sequence is not wrong and it takes a year to pay.

For a product that needs users before it runs out of enthusiasm, the sequence worth running is different in emphasis rather than in content. The same technical basics matter, because an assistant cannot cite a page it cannot read. The same publishing matters. What changes is what you optimise the writing for, and how quickly you expect to see anything.

Concretely: write pages that answer one question each, put the answer near the top in a paragraph that survives being lifted out of context, be specific rather than promotional, and make sure the right crawlers are allowed in. Those are the things that make a passage usable, and they pay much sooner than link building does.

The mistake that closes the channel entirely

There is one error common enough to be worth naming here, before anything else on this page is acted upon.

A great deal of advice published in the last two years told site owners to block AI crawlers, on the reasoning that these systems consume your content and give nothing back. Many builders copied a robots.txt block from one of those posts.

The problem is that the bots are not interchangeable, and the ones that train models are different from the ones that fetch pages to cite in answers. OpenAI documents GPTBot as the crawler for content that may be used to train foundation models, and OAI-SearchBot separately as the one used to surface websites in ChatGPT's search features. Blocking the second one removes you from the answers. Anthropic and Perplexity draw the same distinction with their own separate agents.

So a blanket block, applied by somebody who wanted to opt out of training, quietly switches off the citation channel too. Read your robots.txt today. This is covered in detail on the crawler page in this section.

What it does not do

It should be said plainly, because overselling this would be its own kind of dishonesty.

The numbers at this stage are small. An AI assistant sending you visitors in your first month is sending tens, not thousands. If you need volume immediately, this is not the channel that provides it.

It also does not replace Google. Google remains the larger channel over any long horizon, and the work you do to become citable happens to be most of the work that eventually earns Google's trust anyway. These are not competing strategies.

What it does is give a new product a way to reach real users during the period when the alternative is waiting. That is worth a great deal when you are deciding whether to keep going.

Where to start in this section

The pages below take each part of this in turn, in the order that makes sense to work through them.

  • Check whether the AI crawlers can currently reach you, since a single wrong line makes everything else pointless.
  • Ask the assistants about your own product and read what they say, which tells you what they currently know.
  • Learn what makes a passage citable, and rewrite your best existing page to that shape.
  • Set up tracking, because most AI referral traffic arrives misfiled and you will otherwise conclude nothing is working.

Every part of this, in detail

Each section below is a full guide of its own, on one specific piece of the problem.

  • The Kurli case: We keep being asked whether the AI channel is real for a brand new product or just something consultants talk about. Here is the smallest honest answer we have: one domain, one month, two numbers.
  • Let the right crawlers in: A lot of advice published recently told site owners to block AI bots. The advice treated them as a single group. They are not, and the difference decides whether you can be cited in an answer at all.
  • How it picks: Somebody types a question that your product answers perfectly, and the assistant names three competitors. Understanding why is more useful than being annoyed by it, because most of the reason is fixable.
  • Ask about yourself: This is the cheapest diagnostic available to you. It costs twenty minutes, needs no tools, and the shape of what comes back points directly at what to change.
  • Write what can be lifted: This is the most useful shift in thinking available to anybody writing for the AI channel, and once you see it you cannot unsee it in your own drafts.
  • Perplexity: If you are going to spend a week understanding one AI surface properly, and you have a product that launched recently, this is the one that will repay it fastest.
  • Track the traffic: A builder tells us the AI work produced nothing. We look at the analytics and the Direct bucket has grown steadily for two months. These are frequently the same finding, read two different ways.

Part of a larger guide

This page is one part of Vibe coding SEO. The other parts:

Questions people ask

Can a brand new website get traffic from ChatGPT?
Yes, and typically faster than from Google. An assistant composing an answer selects passages that address the question rather than weighting domain age the way classical search does for a young site. On kurli.me, registered 27 July 2026, ChatGPT sent 56 visitors in the domain's first twenty-five days against 23 from Google Search.
Why is Google so slow for a new domain?
Because it knows almost nothing about you. There is no history of people finding the site useful and no record of what it is about, which from Google's position makes an unknown domain hard to distinguish from a low-quality one. The response is to wait and observe, and there is no safe way to shorten it.
Does blocking AI crawlers hurt me?
It can close the citation channel entirely, and many site owners have done it by accident. The bots are not interchangeable: OpenAI documents GPTBot as the training crawler and OAI-SearchBot separately as the one that surfaces sites in ChatGPT's search features. A blanket block intended as a training opt-out also removes you from the answers.
Is AI search traffic large enough to matter?
At the beginning it is tens of visitors rather than thousands, so it will not solve a volume problem. Its value is that it reaches real users during the months when the only alternative is waiting for Google, which is often the difference between continuing with a product and abandoning it.
Should I work on AI search instead of Google?
Not instead of, and the choice is largely false. Most of what makes a page citable by an assistant, which is clarity, specificity, direct answers and clean markup, is also what eventually earns Google's trust. The difference is emphasis and how soon you should expect to see anything.
How do I know if AI assistants are sending me traffic?
Usually you do not, because a large share of it arrives without a referrer and lands in the Direct bucket in analytics. Setting up a channel group that captures the assistant domains as referrers is worth doing before you conclude the channel is not working.

See what the assistants can currently read

The free Get Found Check reads your domain the way an AI crawler does and reports whether you are reachable, readable and citable right now.

Run the free check Have Licheo do it for you