How I use AI, and where I refuse to.
I sell marketing advice. If you cannot tell whether that advice came from fifteen years of doing the work or from a chatbot given a keyword, my opinion is worth nothing to you. So rather than claim I never touch these tools, here is exactly how I use them and what I will not use them for, on my own site and on yours.
Parts of this website were drafted with AI assistance, working from my own notes, experience and client results. Every claim, opinion and number in it is mine, checked by me, and I am accountable for all of it. What AI never does here is invent a fact, a review, a result or a recommendation.
I would rather tell you that plainly than claim a purity nobody in this industry currently has. If something here reads hollow or turns out to be wrong, tell me and I will rewrite it or take it down.
Why this page exists
Since 2023 the internet has filled up with marketing articles that say nothing. They are grammatically perfect, structurally tidy, and completely empty. They exist to occupy a keyword, not to help the person reading. Anybody who has searched for business advice recently has waded through dozens of them.
That flood has made honest expertise harder to spot, and it has made buyers rightly suspicious. Publishing this policy is the only way I can think of to be checkable rather than just reassuring. It also matters commercially: I optimise client sites for answer engines, and I will not recommend standards to a client that I do not hold myself to.
Where I do use AI
I use it the way I would use a very fast, very well read assistant who has never actually run a business. Useful for gathering, structuring and checking. Useless for judgement, and never left unsupervised.
Pulling together what has been published about a Google update, or summarising a long documentation page so I can find the part that matters. I then verify anything I intend to state as fact against the primary source.
Turning my notes, transcripts and client reporting into a first draft in a sensible order. The material and the thinking are mine. I then edit every line, cut what is padding, and put my name to the result.
Asking what a sceptical reader would object to, or what I have left out. This is genuinely the most valuable use, and it changes what I publish more often than I expected.
Schema markup, spreadsheet formulas, bulk keyword grouping, spotting patterns across a few hundred rows of search data. Machine work, done by a machine, then sense checked.
Where I will not use it
These are not preferences. They are lines, and they apply to client work as firmly as to my own site.
Nothing goes out on the strength of a prompt. If I could not defend every claim in a piece to a client on a call, it does not get published. If I have nothing worth saying on a topic, the honest answer is to publish nothing at all.
Every review on this site was written by the named person. Every figure in a case study came from the client’s own reporting. Inventing either is fraud, and no phrasing softens that.
No generated images of people, premises, staff or work. Photographs of my clients’ businesses are photographs. Illustration and diagrams are fair game and are labelled as such.
Emails, review responses and messages from me are from me. You will never get an automated answer wearing my name, and you will never be handed to a bot mid conversation.
Your customer lists, sales figures and commercially sensitive documents do not get pasted into a consumer AI tool. Where a tool is used on client data it is one that does not train on inputs, and it is agreed with you first.
What every article has to survive
Before anything gets published on this site it has to pass four questions. Most first drafts fail at least one.
If I have not run it for a real client, I do not write about it as though I have. Where I am reasoning rather than reporting, I say so in the text.
Specifics are what separate experience from summary. If a draft contains no figures, no examples and nothing checkable, it is not finished.
An article that lists every option with equal weight has helped nobody decide anything. If I think something is a waste of money, I write that it is a waste of money.
Yes, always. Articles that withhold the useful part to force an enquiry are advertising with a byline. Plenty of people read these and do the work themselves, and that is fine.
Corrections, updates and getting things wrong
Search changes constantly, so advice that was right in 2024 can be wrong now. Every article carries the date it was last reviewed, not just the date it was first published. When I find something out of date I either update it and change the review date, or I remove it.
If I get something factually wrong and you tell me, I will correct it and say in the article that it was corrected. I would rather have a visible correction than a quietly edited page.
What this means if you are a client
Content produced for you follows the same rules, with one addition: you always know which is which. Anything drafted with AI assistance is flagged to you before publication, and nothing goes on your website without your sign off.
There is a practical reason beyond ethics. Google’s guidance is about whether content is helpful and demonstrates real experience, not about how it was typed. Generic generated content fails that test, gets outranked by businesses doing the work properly, and takes your credibility with it. The cheap route costs more in the end.
If another agency cannot tell you plainly how your content is produced, ask them again until they can. It is a fair question and the answer tells you most of what you need to know.
The awkward ones, answered
So was this website written with AI?+
Yes, in part, and I am not going to pretend otherwise on a page about transparency. The pages here were drafted with AI assistance from my own notes, my client work and my results, then edited and approved by me line by line. The distinction I care about is not who typed it. It is whether a real person with real experience stands behind every claim in it, and whether anything was invented. Nothing was.
Will you use AI on my content, and will you tell me?+
Probably yes for drafting and research, and always openly. Your content is built from your knowledge, your customers’ questions and your results, then edited by me and signed off by you before anything is published. What I will not do is take a keyword list, generate forty articles from it and invoice you. That damages the site it is published on, and I would rather lose the work than charge you for it. If budget is tight we publish less and make it count.
Does Google penalise AI content?+
Not for being AI written as such. Google’s stated position is that it rewards helpful content demonstrating real experience and expertise, however it was produced, and acts against content made primarily to manipulate rankings. In practice most bulk generated content falls squarely into the second category, which is why it stops working.
Is my business data safe when you use AI tools?+
Your commercially sensitive data does not go into consumer AI tools. Where AI is applied to client material it is through tools that do not train on inputs, and it is agreed with you beforehand. If you would rather it was not used on your account at all, say so and it will not be.
You optimise for AI search. Isn’t that a contradiction?+
No. Answer engine optimisation is about making genuinely good information easy for AI systems to find, understand and cite. That works best when the content is real, specific and clearly attributed to somebody accountable. Being transparent is not in tension with AEO, it is a requirement of it.
Want marketing you can put your name to?
Thirty minutes, no charge and no pitch. I will look at where you show up now and tell you honestly what I would fix first.