The AI Practitioner went live on 14 August with two of its pages hidden from every search engine on the internet.
Both pages existed. Both were built, styled and linked into the navigation structure. We told the crawlers not to look at either of them.
As of this morning, they are on. The site is complete and the business is open. This is the part of that story worth telling, because the decision behind it is one most people face and almost nobody writes down.
What was hidden, and how
Two lines in the site’s robots.txt file, sitting underneath a comment I wrote at the time:
Hidden until the 1 September full launch: /blog has no real posts and /book has no booking integration. Remove these two lines when both ship.
There is a matching filter in the site’s build configuration that keeps both paths out of the sitemap, so nothing points at them from anywhere a search engine can reach.
Belt and braces, on purpose. A stray link into a half-built page is the kind of thing you discover six weeks later in an analytics report.
Why an empty blog is worse than no blog
The reasoning is simple enough to state in a sentence: a half-built thing does more damage visible than invisible.
Picture the visitor. They have read the services page, they are mildly interested, and they click through to the blog because they want to know whether the person behind this actually knows anything. They find three placeholder posts and a “coming soon”.
They have now learned something specific about how this business operates, and it is not the thing I wanted them to learn. The same goes double for a booking page that does not book. Asking someone to take an action and then not honouring it is worse than never asking, because you have spent their attention and given nothing back.
An absent page costs you an opportunity. A broken page costs you the benefit of the doubt. Those are not the same price.
The part that is not a strategy
Here is the bit that would be easy to leave out.
Mid-August was the original target for the whole site. Everything. The blog and the booking page slipped, and when the date got close the choice was between launching without them or holding the entire site back another fortnight.
Neither of those is a triumph. I did not plan a staged launch in July and execute it beautifully in August. I ran out of time on two pages and had to decide which half to ship.
I am labouring this because the version of the story where it was all deliberate from the start is the version that usually gets told, and it is not much use to anyone. Choosing which half to ship is the decision most people actually face on a project with a date attached. It comes up late, under pressure, and there is rarely a good option, only a less bad one.
The useful question in that moment is not “what did I promise”. It is “which of these does more harm in public than it does missing”.
What the fortnight bought
Three things, and they are the reason the delay was worth taking rather than fudging.
Two blog posts written properly, rather than five written quickly. The blog is the canonical version of everything that goes out on social, which means it carries the evidence, the caveat and the worked example that a short post has to cut. Writing that in a hurry produces something worse than silence.
Booking wired to a real calendar with a real price against it. The 1:1 session is $175 and it now puts a real slot in a real diary. Payment is arranged separately rather than taken at the point of booking. That sentence took longer to make true than it takes to read.
Analytics running from the first day of real traffic. Cloudflare Web Analytics went on at soft launch, before there was anything much to measure, which means the first month of the full site is measurable rather than guessed at. Turning measurement on after you want an answer is how you end up with an opinion instead.
One thing we did not hide
The same robots.txt file that hides two pages goes out of its way to invite a specific set of
visitors: the crawlers behind ChatGPT, Claude, Perplexity, Google’s AI results, Apple and Bing. Each
one is named individually and allowed explicitly.
That is deliberate too, and it is the opposite instinct to the one above. People increasingly find a business by asking an assistant rather than searching, and being readable by those assistants is a choice you make in a config file months before it pays off. Naming them one at a time rather than relying on a blanket rule is defensive: a wildcard is easy to undo by accident, and a named list is hard to break without noticing.
Hide what is not ready. Invite what is. The two decisions look opposite and come from the same place.
What is live today
Home, About, Services, FAQ and Contact have been up since 14 August. As of this morning the blog is open and the booking page takes bookings.
The business behind it is Groundframe Ltd, trading as The AI Practitioner, and today is its first day of clean trading. That is the other thing that happened on 1 September, and it is why this date rather than any other.
If you have a project with a date on it and two components that will not make it, the question is not how to get everything done. It is which of the unfinished parts is worse in public than absent. Ship the rest. Say which bits are missing. Nobody has ever thought less of a business for that, and plenty have thought less of one for a page that promised something it could not do.
The site is at theaipractitioner.ai. The 1:1 session, From Curious to Capable, is now bookable at $175.