Bradley Benner has been optimising local pages since 2010 and now sells white-label SEO fulfilment to other agencies. He has a rule for where AI-shaped writing belongs on a page, and it is the cleanest version of our own argument that we have heard from somebody outside this company.
“The structure of a page, the main optimization elements, that should be really for the models and for the bots. But the paragraph text between those main optimization elements should be for the human. It should be to convert the human, to answer their questions.”— Bradley Benner, Semantic Mastery
01Two audiences, two jobs, one page
Benner counts six main optimisation elements on a page: the URL, the titles, the headings, the subheadings, the media and the internal links. Those exist so a machine can work out who you are, what you do and where you do it. He optimises the SEO title as a strict entity statement for exactly that reason.
Then the H1 does something different.
“The H1 becomes a restatement of the entities that were referenced in the SEO title. And I say restatement, not repeat … The SEO title, the meta title is for the bots, the models. But the H1 is for the human visitor.”— Bradley Benner
02Where the generated drafts actually fail
A language model is good at the first job. Structure, titles, headings, consistent entity phrasing across a page — that is pattern work, and it is why generated drafts score well on audit tools and still do not sell anything.
The failure is in the paragraphs. That is where a reader decides whether the person writing has done the work, and it is the one part of the page that cannot be assembled from what already ranks. Our whole position on information gain is that the sentences between the headings have to contain something the model could not have known.
03What that means for a content brief
- Machine layer: URL, SEO title, H1, H2s, H3s, media, internal links. Consistent, entity-led, repetitive on purpose.
- Human layer: every paragraph in between. A named person, a real number, an objection answered, a limit admitted.
- The test: delete the headings. Does the remaining prose still tell a reader something they could not get from the first result they already read?
That is the split we build to, and it is why our drafts start with an interview rather than a SERP scrape, inside a human-in-the-loop workflow. A human on a recording says things a corpus average never will.
04Why this is getting more important, not less
“Because Google’s no longer the only kid on the block … we have to focus more on not just getting exposure across all of these different surfaces, but also in the copywriting and the marketing side, the actual direct response marketing side of things, the human element side of things.”— Bradley Benner
His read is that the arrival of AI search pushed the industry back toward traditional copywriting, because exposure across more surfaces is worth less if nothing converts once somebody arrives. That matches what we see in client drafts: the structural work is now cheap and universally available, so the paragraphs are the only place left to be different.
05The honest limit
This is a practitioner’s framework, not a controlled test. Benner is describing how he teaches page optimisation today, based on fifteen years of client work and his own fulfilment volume. We agree with it because it matches what we build, which is not the same as proving it.
What you can test yourself is narrower and takes an afternoon: take your last five published pages, strip the headings, and read only what is left. If it reads like a summary of the competition, the machine layer was the only layer you shipped.
Want content like this for your clients?
Human-led, AI-accelerated, expert-reviewed. We build SEO content that reads — and ranks — like a real person wrote it, because one did.