Commercial

How we use AI, and where the gate is

Yes. An AI agent drafts a large part of what we publish, including pages on this site. Nothing reaches a client site or a client account without a named human reading it and signing it. That is the entire policy. The rest of this page is where the edges are.

The short answer, in full

Does your agency use AI to write content is now a question with only one honest answer, and we would be suspicious of any agency in 2026 that claimed a flat no. The useful question is not whether a tool was involved. It is where the gate is, who holds it, and what happens to the draft that fails inspection.

Our gate is one person and it is a bottleneck on purpose. That is why this site ships in batches rather than appearing overnight.

Where the agent is allowed to work

• First drafts of pages, answers and guides, written against a brief that fixes the facts, the prices, the terms and the claims it is permitted to make.
• Research and structure: finding the questions people actually ask, grouping them, and proposing the one page that answers one of them completely.
• Mechanical work at volume: schema, internal links, redirect maps, metadata. The parts where consistency across hundreds of pages beats flair, and where a human is reliably worse by the hundredth page.
• Checking our own output. A large part of the tooling behind this site exists to fail a page that breaks a rule, including rules about prices, client names and claims we are not allowed to make.

The editorial gate

The agent drafts. Marco publishes. There is no third state and there is no queue that drains itself.

In practice that means a named person reads every page before it goes live, and the same name sits at the bottom of it with the date he last read it. Where a page states a fact, a number, a price or a term, the agent is structurally forbidden from marking that fact as sourced. Only the human can set that flag, and the publishing pipeline refuses to ship a page where it is missing. The constraint is in the code, not in a promise.

This is slower than the alternative and it is the only part of the process that makes the rest of it safe to use at volume.

Where it is not allowed near your account

• No autonomous spend. An agent does not change a bid, a budget or an audience. A person does that, in your ad account, with your money.
• No publishing to your properties. Nothing posts to your website, your Google Business Profile or your ad account without a human action behind it.
• No client data in a prompt beyond what the specific task requires, and nothing of yours used to train anything.
• No invented proof. Client names, reviews, volumes and results come from a register a human verified, or they do not appear at all. That rule is why this site publishes fewer numbers than its competitors.

Why this beats the realistic alternative

The alternative is not a specialist writing every word from scratch. At our price the real alternative is a junior producing five hundred words an hour on a subject they learned about this morning, or a content mill billed at a markup. We have read the output of both. Neither is more accurate than a tightly briefed draft that a specialist then rewrites.

What decides the quality is the brief and the review, not the tool. A vague brief produces generic copy whether a person or a model writes it. A brief that fixes the facts, forbids the claims and names the reader produces something specific either way. The review is where it gets a voice, and the voice is the part a model cannot be trusted with.

Search engines agree, roughly. What gets penalised is content with no purpose and no accountability, not content with a tool in its history. The way to stay on the right side of that is to publish fewer pages that each answer a real question, with a named human behind every one.

What you can audit

Every page carries a byline and a last reviewed date. Ask us how any specific page was produced and we will tell you which parts were drafted, what the brief fixed, and what changed in review.

If you would rather no AI assistant touched your account at all, say so before we start. We will tell you honestly what that changes about the volume and the price, and if the honest answer is that it does not work at our floor, we will say that instead of agreeing and quietly doing it anyway.