AI Content Isn't The Problem. Publishing Without A Point Of View Is

We've reviewed a lot of sites that went quiet after a core update, and the pattern is boringly consistent. It isn't AI authorship that correlates with the drop. It's interchangeability.

Page after page that reads like it was assembled from the average of everything already published on the topic. Technically accurate. Reasonably structured. Entirely replaceable. A machine reading the web has ten thousand of those and no reason to prefer yours.

Meanwhile, plenty of sites using AI heavily are doing fine — because the AI is doing the assembly and a human is supplying the thing that makes the page worth reading.

What 'point of view' means in practice

It isn't having hot takes. It's having something on the page that could only have come from you:

• A position you'll defend — including what you don't recommend, and why.
• First-hand specifics: what happened on real jobs, what it cost, what went wrong, what you'd do differently.
• Your own data, however small. Twenty client results beat a stat you found in someone else's blog post.
• Local and sector reality — Australian regulations, seasonality, supplier constraints, what buyers here actually ask.
• Clear scope: who this is for and, just as usefully, who it isn't.

Where AI genuinely earns its keep

Used well, AI removes the drudgery around the thinking rather than replacing it. The highest-value uses we see are unglamorous.

Research synthesis: pulling apart what the top twenty results already cover so you can find the gap instead of adding the twenty-first restatement. Structural drafting: turning a messy voice memo from the person who actually knows the answer into an ordered outline. Editing: tightening, cutting hedging, flagging unsupported claims. Adaptation: taking one strong piece and reshaping it for different formats and reading levels without diluting it.

Notice what none of those are: 'write me a 1500-word article about X'. That prompt produces the average of the internet, which is precisely the thing that no longer earns attention.

A workflow that holds up

The version we run looks like this, and it's deliberately front-loaded with human input.

• Start with a question a real prospect asked, in their words. Not a keyword, a question.
• Interview the expert — often the business owner — for fifteen minutes. Record it. This is where the point of view comes from.
• Use AI to map existing coverage and identify what nobody is saying clearly.
• Draft structure with AI, then write or rewrite the sections that carry the argument yourself.
• Fact-check every number and claim against a primary source. AI is a confident liar about specifics.
• Add what only you can add: an example, a caveat, a cost, a photo of the actual work.
• Have someone who does the job read it before it ships. If they'd be embarrassed by it, don't publish it.

The volume trap

The tempting move once drafting is cheap is to publish more. It rarely works. Thin pages compete with each other for the same intent, dilute internal linking, and give both Google and LLMs a weaker overall read of the site's authority.

We'd take twelve genuinely useful pages a year over a hundred adequate ones, every time. The twelve get cited, linked and remembered. The hundred get crawled and forgotten.

The test to apply before publishing

One question, and it's harsher than it sounds: if a competitor published this exact page under their logo, would anything be lost?

If the answer is no, the page isn't finished. Not because AI wrote part of it — because nothing in it is yours.