The short version
- Optimisation scores measure similarity to what already ranks, not quality. Those are different things and sometimes opposed.
- A perfect score means you resemble the existing top ten. It does not mean you are better than them.
- Plugin traffic lights in Yoast and Rank Math grade sentence length and keyword placement — useful as a proofreading nudge, meaningless as a target.
- The only durable advantage is content a competitor cannot produce: your data, your testing, your judgement.
Every content optimisation tool presents a number, and numbers are persuasive. A writer looking at 68/100 will make changes until it reads 85, and will believe that made the article better. Often it did the opposite, and understanding why requires knowing what the number is computed from.
These scores are built by analysing pages that currently rank for a query and measuring how closely your draft resembles them statistically. That is a genuinely useful signal for catching omissions. It is a terrible target, because maximising similarity to the average competitor is the opposite of differentiation.
What the scores actually measure
Look at the right-hand column. Two rows are genuinely useful, two are weak, and three are noise. A composite score that blends all seven and presents one number cannot distinguish between "you have missed an important subtopic" and "your sentences are longer than average" — and it is the second that is easiest to fix, so that is what gets fixed.
The damage patterns
- Unnatural repetition to raise term frequency, which makes prose worse and is explicitly discouraged by Google’s guidance.
- Padding to reach a word count modelled on competitors who were themselves padding.
- Fragmented arguments from chopping sentences to satisfy a readability heuristic.
- Subheading inflation that breaks an explanation into disconnected sections.
- Averaged-out voice — the specific thing that would make someone share the article being smoothed away because it does not resemble the top ten.
The last one is the real cost and it is invisible in any score. Optimisation tools cannot measure insight, so insight is the first thing sacrificed when a writer is optimising toward a number under deadline.
The way to use them that works
- Write the piece with the tool closed. The draft should be shaped by the subject, not the scoring model.
- Open the analysis and read the missing-term list as a set of questions rather than a checklist.
- Add what is genuinely missing. A term you cannot justify covering is a term to ignore.
- Stop well before a perfect score. If further gains require changes you would not defend to an editor, you are done.
- Judge the final piece the way a reader would: would anyone send this to a colleague?
The organisational version of this problem is worse than the individual one. Once a score becomes a target in a content team — a minimum before publication, a number in a freelancer brief — every writer optimises toward it because that is what they are being measured on. The scoring model then quietly becomes your editorial standard, and nobody decided that. If you use these tools across a team, be explicit that the score is diagnostic and that no one is judged on it.
What the score cannot see
Original data. First-hand experience. A genuinely better explanation. A useful table nobody else built. Those are the properties that earn links, citations and return visits — and they register as zero in every optimisation tool on the market, because no competitor page contains them.