The honest split is narrow but it is real. For one class of work the software genuinely runs the whole way with nobody touching it: it reads a site, works out the trade, writes the change, refuses it if it is defective, publishes it to the live page, rebuilds the sitemap and tells the search engines. 313 items have gone through that path.
What the software published without a person
| Meta description rewrite | 76 |
| Schema / JSON-LD to add | 76 |
| Page | 67 |
| Title tag rewrite | 61 |
| AISO refresh — FAQ schema | 14 |
| Lead magnet | 9 |
Everything else is a person with software helping, and it would be dishonest to describe it any other way. 5,736 finished items are sitting in a queue waiting for a human — and the work done on this network in the past two weeks was somebody writing generators and running them, not the engine acting on its own.
The distinction Tim is really asking about
The engine can produce at scale — it has for months. What it mostly cannot do is act: place an ad, post to an account, submit a listing, send an email. Producing is the easy half and we should stop presenting it as the whole.
Not a low number — none. No business, no customer, no payment attributable to anything this software did. Since measurement was fitted, across the whole network: 0 leads and 180 recorded events.
What is true, and why it still does not count
Two of the sites do receive real inquiries — HulloShips and OFS Thai. They arrive as email to a personal address, through a path nothing was measuring, and none of them can be traced to a thing the engine did. Claiming them would be exactly the move Tim is guarding against.
Traffic is not the missing piece either. The network takes about 39,771 real human page views a day, filtered of bots and scanners. The audience exists. It has never been asked to do anything, and until this week most of it had no way to respond even if it wanted to.
What would change the answer
One named business, one enquiry that arrived through a measured path, one amount they paid. That is the whole bar. It is low, and we have not cleared it.
The boundary is sharper than a percentage suggests, and it is worth stating as a line rather than a score.
Software does this alone
- Read a website and score it
- Work out the trade and load a knowledge base for it
- Write titles, descriptions, structured data, FAQs, whole pages
- Refuse its own work when it is defective
- Publish to a site we control
- Rebuild the sitemap and notify search engines
- Record that a form was filled or a number tapped
A person takes over here
- Anything needing an account — Google Business Profile, social, ad platforms
- Anything on somebody else's website
- Turning a brief into a finished article
- Any judgement about the business — price, positioning, what to spend
- Anything needing a body — a camera, a phone call, a handshake
- Deciding a thing is good enough to go out
The line is not where people assume. It is not "software writes, humans edit" — the software's writing goes out unedited on a narrow class of work. It is software can act on property we own, and almost nowhere else. Every remaining blocker is an account, a credential or a judgement.