Here is the mistake I see most often, and it is an understandable one.
Somebody wants to write about a topic. So they look at what is already ranking, read the top few results, and produce something covering the same ground a bit better. More thorough, better written, nicer to look at.
And nothing happens.
The reason is straightforward once you see it. You have told the machine something it already knew.
Google has a term for this in its own guidance on AI features: commodity content. Material that adds nothing which was not already available somewhere else. Google’s position is that avoiding it is the single biggest long-term factor in whether you show up, ahead of any technical tweak.
Which is worth sitting with, given how much is currently being sold on the basis that the answer is a file in your root directory or a bit of markup. It is not. The answer is having something to say.
The concept worth having
There is a term for this, and it is a useful one to carry around: information gain.
It is the gap between what is already out there and what you are adding. Not how good your writing is, not how comprehensive it is. What is in it that was not available before.
A page with high information gain contains something that could not be found elsewhere. A page with none is a better-organised copy of the consensus.
For a search engine, the second sort is a reasonable thing to rank if it is genuinely clearer than the alternatives. For an AI deciding whose page to quote, it is close to worthless, because it already has that information from somewhere else. There is no reason to prefer your version.
So the question to ask before writing anything is not “is this good?” It is “what is in this that is not already known?”
If the honest answer is nothing, you are writing something that will sit there being perfectly correct and completely ignored.
What actually creates it
The good news for a small business is that you are sat on rather a lot of this without realising it.
Your own numbers. What a job actually costs on the Fylde coast in 2026. How long a boiler genuinely lasts round here. What proportion of the houses you work on have the same problem. Nobody else has these figures because nobody else does your job in your area.
What happened when you did the thing. Not advice about how a job should go. An account of how one actually went, including the bit that went wrong and what it cost to put right.
The patterns only you have noticed. The thing that always turns up in properties of a certain age. The mistake customers repeatedly make before they call you. The question you get asked every single week that nobody has written down properly.
Your opinions, where you actually hold them. Which materials you refuse to use and why. What you think is a waste of money. Where you disagree with the received wisdom in your trade. A genuine position is information. Balanced hedging is not.
Every one of those is unavailable to a machine unless you write it down, because it did not happen to anybody else.
Case studies, which are wildly undervalued
This is the version of it that most businesses skip, and I think it is the single biggest missed opportunity going.
A case study is information gain in its purest form. It is an account of a specific job, with specific circumstances, specific decisions and a specific outcome. It cannot be generated, it cannot be scraped from a competitor, and it cannot be found anywhere else.
It also does something rather useful when somebody is deciding whether to hire you. An AI asked to recommend a business needs evidence. “They say they are experienced” is not evidence. “Here is a job they did, here is what it involved, here is what it cost and how long it took” is.
What makes one worth writing:
Be specific. The town, the property, the actual problem, the actual timescale. Vagueness defeats the point of the exercise.
Include real numbers. What it cost, how long it took, what changed afterwards.
Say what was difficult. A case study where everything went perfectly reads like marketing. One that mentions the complication and how it was handled reads like an account, and it is far more persuasive to a human as well.
Piet Baudoin, who runs Poldermarketing in the Netherlands, adds a rule I agree with in his guide to a startup’s first content: do not publish a case study until you have all three parts, the real situation, what was done about it and how it turned out. If the job is not finished yet, wait for it.
And write one per job type rather than one per client. You are trying to be the obvious answer to a particular kind of question.
The bit where I have to be honest about myself
I would be a fraud writing this without admitting something.
I have published a great deal about being findable, and almost none of it contains original data of my own. I have quoted other people’s research all the way through, which is fine as far as it goes, but by my own argument it is not information gain. It is me being a well-organised summary of things somebody else found out.
I know exactly why. Original data is harder. You have to actually run something, measure it, and write up what happened, including the parts that did not work. Summarising is quicker and it looks nearly as good.
That is precisely why so few people do it, and precisely why it is worth doing.
So I am going to fix mine, and it will take a few months, because the only honest way to produce it is to do the work and wait. If you are in the same position, that is the answer for you too. There is no shortcut and the shortcut is the whole point.
What this means practically
Before you write anything, run one test.
Could somebody get this from an AI without visiting your website?
If yes, think again. You are producing something that will be perfectly accurate and entirely unnecessary.
If no, you have got something, and the more specific and local and first-hand it is, the more it is worth.
That does not mean never explaining anything. Explanations are useful, particularly to humans who are new to a subject. But do not expect an explanation of a known thing to be the piece that gets you recommended. It will not be, because the machine can already explain it.
The thing that gets you recommended is the thing that only you know.
Which is, admittedly, more work. It means paying attention to your own jobs, writing down what you learn, keeping track of your own numbers, and being willing to have an opinion in public.
But it also means competing on something your competitors cannot copy, which is a considerably better position than trying to write a slightly better version of a page that already exists.
Want to work out what you know that nobody else does? Get your free audit, or email hello@inkon.co.uk. I read every message myself.