The short version
- Optimisation tools evaluate against ranking pages. AI writing tools generate text. Neither knows anything you do not tell it.
- Combining them naively — generate a draft, optimise to a high score — produces content statistically identical to the existing top ten and valuable to nobody.
- AI writing is genuinely useful for structure, first drafts and editing. It is not useful for producing information you do not have.
- The scarce input in both cases is something true and specific that you know. Neither tool supplies it.
These get grouped together as "AI content tools" and they do opposite things. One reads the competition and scores you against it; the other produces text from a prompt. Used together carelessly they form a closed loop that reliably produces the least differentiated content it is possible to publish.
Used deliberately, both are genuinely useful. The difference is whether a human with knowledge is supplying the substance.
The short answer
Fit for your workflow
Use AI for structure and drafting. Use optimisation for gap-checking. Supply the substance yourself.
An AI writing tool is good at organising material, producing a first draft from notes you provide, and editing for clarity. An optimisation tool is good at catching subtopics you have omitted. Neither can supply the specific, checkable, first-hand material that makes a page worth reading — and that is the only part that is genuinely scarce.
- You have expertise and no time
- AI drafting from your notes works well.
- You have neither expertise nor data
- No tool fixes that. Do not publish.
- Team with inconsistent output
- Optimisation tool for consistency.
- Editing and tightening a draft
- AI is genuinely good at this and underused for it.
What each is for
The fourth row carries the main risk. Asked for statistics, a language model will often produce plausible, precisely formatted numbers with no basis. Any figure that appears in a draft without a source you can check should be treated as fabricated until verified — which is a discipline, not a setting.
The closed loop to avoid
The failure pattern is mechanical. Generate a draft from a model trained partly on existing web content about the topic. Score it against the pages currently ranking. Edit until the score is high. The output is, by construction, an average of what already exists — and you have spent money to produce something a summary can reproduce for free.
What breaks the loop is introducing information from outside it: a number you measured, a test you ran, a customer conversation, a screenshot of your own data, a judgement you are willing to defend. See content for humans vs content for LLMs.
A workflow that keeps the substance yours
The arrangement that works puts the human where the information is and the tools where the labour is. Gather the material yourself — the data, the test results, the customer conversation, the opinion you are prepared to defend. Use an AI assistant to organise it and produce a draft. Edit it into something you would put your name on. Then run an optimisation pass purely as a gap check.
What makes this work is the ordering. Substance first means the tools are shaping something real rather than generating something generic. Reverse the order — generate, then optimise, then look for something to say — and you get the closed loop described above, which produces pages that satisfy neither the guidance nor a reader. See editorial quality vs optimisation score.
Where AI writing is genuinely strong
Turning messy notes into a clean structure. Tightening prose you wrote too quickly. Producing a first draft of something formulaic so you can spend your attention on the parts that are not. Catching inconsistencies across a long document. These are real productivity gains and none of them requires the model to know anything you do not.