Readability
How much effort does each sentence cost?
- Reading ease Flesch, or its Russian adaptation
- Sentence length average and the long tail
- Paragraph shape
- Scannable structure
- Long-word load
- Padding and filler
- Passive voice English
Free · pasted text never leaves your browser
Your text has three readers.
One is tired. One is a machine. One is a person deciding whether to stay. Paste a text or a page address and see how each of them gets on with it — line by line, with the exact sentences to fix.
- ReadabilityCan a tired person get through it?
- MachinesCan an AI lift a clean answer out of it?
- PeopleWill someone want to keep reading?
How it scores
Three readers, twenty-five questions.
Every check is something you can see in the text. Each one shows what it measured, on which sentences, and what to do about it. Checks that do not apply — a passive-voice test on Russian, page signals on pasted text — are dropped and the rest are re-weighted, never guessed.
Machines
If a system lifts one passage, does it still make sense?
- Passages that stand alone the one most human text fails
- Answer-first opening
- Quotable facts figures, dates, names
- Structure a parser can follow
- Unverifiable claims
- Questions answered directly
- Sources and attribution
- Enough to quote · Dated and current
- Page-level signals URL mode only
People
Would someone read the second sentence?
- The first line pulls
- Stock phrases and buzzwords
- Concrete, not abstract
- Speaks to a person
- Sentence rhythm
- Examples and stories
- Curiosity and tension
- A clear next step
What this cannot tell you
- It does not predict rankings or citations. The machine score measures properties that make text easier to lift and check. Whether a given assistant cites a page depends on things outside the text.
- It reads patterns, not meaning. A pattern match can be wrong: an adjective can look like a passive verb, and a text can be wrong in a way no formula sees. Every flag shows the words it fired on, so you can overrule it.
- It is not an AI-writing detector. Stock phrases make writing generic whoever wrote it.
- The overall number is a harmonic mean. The weakest reader pulls it down on purpose: text that suits machines and bores people is not good text.
Questions people ask
Does my text leave my browser?
Pasted text is analysed in your browser and is not sent to Landingpeak. If you fetch a page instead, our server downloads that public page and returns its text to you; the host and word count can appear in server logs. If you email yourself the readout, only the scores and the names of the biggest fixes are sent — never the text.
Will a high score get my page cited by AI assistants?
No, and no honest tool can promise that. The Machines score measures things that make text easier to lift and verify: a clear opening, passages that stand alone, concrete figures and named sources. The field guide separates what is measured from what is speculation.
How is the overall score calculated?
Each reader's score is a weighted average of its checks, with inapplicable checks dropped and the weights re-normalised. The overall number is the harmonic mean of the three, so a lopsided text cannot score well.
Which languages work?
English and Russian in full. Other Latin-script languages get the language-neutral checks — sentence length, structure, figures, dates — and the word-list checks are marked not applicable.
Why does well-written text score low on Machines?
Good prose leans on what came before: “This means…”, “It also…”. A reader has the previous paragraph; a retrieval system lifting one passage may not. The usual fix is small — name the subject again at the start of each section.
Is this an AI-writing detector?
No. It does not try to work out who or what wrote a text. The stock-phrases check flags wording that makes any writing generic.
Checking the words is half of it. The free site scan checks the page around them — structured data, crawler access, speed — and this guide explains how the two fit together.